REF·DB

Working reference for a mechanical design engineer — UK aerospace & defence. Standards, drawing callouts, GD&T, materials and cost, in one place.

METT+00:00:00

REVv1.0.0 · 2026-08-14

SCALENTS · THIRD ANGLE

MODE

ORIENT

SELECT

SIZE

VERIFY

DETAIL

FEATURES

MAKE

SPECIFY

ZONE 1

Overview

A working reference for solid rocket motor mechanical design — — of tools, data and the reasoning behind them. Nothing here needs a network.

PREDICTOR LAB — PREDICTIONS LIVE APART FROM THE REFERENCE DATA Work package estimator for hours and cost from your own history, Weibull reliability, Monte Carlo uncertainty, regression with prediction intervals, and the general predictor bench. Kept outside the eight phases so a fitted model is never mistaken for an allowable.

THREE WAYS IN — PICK WHICHEVER MATCHES YOUR QUESTION You do not need to know the structure to use this. Search if you know the word for what you want. Start a task if you know the job but not the tool. Follow the route if you want the whole workflow in order.

Start a task

The questions this site gets asked most. Each one lands you on the tool, not on a menu.

SYSTEMS STATUS · SELECT TO NAVIGATE · CTRL+K SEARCH ALL · G+NUMBER JUMP (G05, G28) · [ ] STEP ZONE · { } STEP PHASEALL STATIONS GO

The route through

The zones are ordered by when you use them, from understanding the material through to releasing the drawing. You will not work through them in order every time — but this is roughly the sequence a component actually follows.

THE ONE-PARAGRAPH DRAWING PHILOSOPHY A drawing is a contract, not a picture. Every dimension and note should answer one question: “how will the inspector accept or reject this?” If a requirement can't be measured, it can't be enforced. Dimension the function, tolerance only as tight as the function needs, and state the standard you're working to (in the UK, that's BS 8888 in the title block).

FASTEST WAY TO USE THIS SITE Ctrl/Cmd-F works on every page, and the Standards page has its own live filter. Each design guide ends with a copy-ready drawing callout in a dashed box.

Your session

Everything you set is kept in this browser and can be handed to someone else as a link.

nothing changed yet

THE LINK CARRIES THE CALCULATION, NOT A SCREENSHOT A shareable link encodes only the values you changed, so it stays short. Whoever opens it lands on the same zone with the same inputs and can carry on from there — change a number and send it back. That is a design review conversation rather than a pasted image.

Your work is also saved in this browser automatically, so closing the tab does not lose it. Nothing is sent anywhere — this file has no network access at all.

Design record

Everything you have set across the site, collected into one block you can paste into a design note or a review pack.

Press BUILD RECORD to capture the current state of every calculator.

WHAT THIS IS A transcript of your inputs and this site's outputs at one moment. It is a working note, not a released document — every figure carries the same screening-value caveat as the zone it came from. It is here because a calculation you cannot get out of the tool is a calculation you will redo.

⚠ Reference aid only. Values are typical industry guidance for early design. Always verify against the current issue of the cited standard, your company DRM, and programme-specific requirements before release.

ZONE 60

Standards & specs library

The documents that get invoked on UK aerospace & defence drawings, grouped by what they do. Filter by category or type anything — number, topic, keyword.

REFERENCE	WHAT IT IS	CATEGORY	WHY YOU'D REACH FOR IT

WHERE TO GET THEM BSI Knowledge (bsigroup.com) for BS/ISO — many employers have a subscription, ask before buying. DEF STANs are free from the UK Defence Standardization site (search “DStan”). ASME from asme.org, SAE AMS/AS from sae.org. MMPDS via your stress office. Never design from a random PDF of unknown issue — the issue date changes requirements.

DESIGN NOTES — DECISIONS, TRAPS AND CHECK SEQUENCES

APPLICABLE STANDARDS — IN ONE LINE Cite the standard, its issue and the clauses invoked — an undated reference silently tracks whatever was published last.

Order of precedence is normally contract, then programme specification, then company procedures, then national and international standards. Check the contract — it usually says so explicitly.

Standards a UK A&D mechanical designer meets most

Confirm the current issue and what your contract invokes before relying on any of these.

Standard	Covers
BS 8888	UK technical product documentation — invokes the ISO GPS set
ISO 1101 / ISO 5459	Geometrical tolerancing / datums
ISO 8015	GPS fundamentals — the independency principle
ISO 286	Limits and fits
ISO 2768	General tolerances — use with care, see Drawing frame & notes
ASME Y14.5	US GD&T — differs from ISO in defaults
ISO 965 / ISO 5855	Metric threads / aerospace MJ threads
ISO 3601	O-rings and housings
VDI 2230	Bolted joint calculation
NASA-STD-5020	Threaded fastening systems, spaceflight
MMPDS	Metallic material allowables
ECSS-E-ST-32 series	Space structures, incl. pressurised hardware
NASA-STD-5019 / 5009	Fracture control / NDE for fracture-critical parts
EN 9100 / AS9102	Aerospace QMS / first article inspection
EN 10204	Material inspection certificate types

TRAP ISO and ASME look alike on the page and are read differently. The same feature control frame can mean different things under ISO 8015 independency and ASME Rule #1. The drawing must say which system governs.

DESIGN RULE Tailoring is legitimate, silent tailoring isn't. If you're not meeting a clause, record the deviation with its justification and get it agreed.

ZONE 11

Material selector

Pick a family, tick what the part needs, then sort any column. — materials covering aluminium, titanium, nickel and cobalt superalloys, stainless, high-strength case steels, magnesium, refractories, carbon/ablatives, composites and polymers.

Selection chart

An Ashby-style property chart over all — materials. Choose an application preset or build your own: pick the two properties that matter, decide which families to include, apply limits, then overlay a performance index to rank the survivors and draw its selection line.

HOW TO READ THE SELECTION LINE With an index active on its natural axes, the dashed lines are contours of constant index value — every material on a given line performs identically for that requirement. Moving the line towards the top-left improves performance, so anything lying above and left of the bold line beats your current best candidate. This is the whole point of an Ashby chart: it shows you not just the winner, but everything within reach of it, and what you would trade to get there.

Chart axes

APPLICATION PRESET

PERFORMANCE INDEX

X AXIS

Y AXIS

LOG X LOG Y GUIDE LINE

BUBBLE SIZE

COLOUR BY

LABELS

SEARCH NAME

Include & limit

APPLICATION DOMAIN

MATERIAL FAMILIES

SERVICE CONDITION

DESIGN TEMPERATURE — 20 °C (derates strength & fatigue)

FRACTURE, FATIGUE & LAYUP

CRACK ORIENTATION

TOUGHNESS BAND

SURFACE CONDITION (fatigue ka)

LOADING MODE (kb, kc)

SECTION SIZE mm (kb)

RELIABILITY (ke)

MEAN STRESS MPa (Goodman)

COMPOSITE LAYUP

COMPOSITE KNOCKDOWN

PROPERTY LIMITS

MIN SERVICE TEMPERATURE — 0 °C

MAX DENSITY — 25.0 g/cm³

MIN YIELD STRENGTH — 0 MPa

MAX COST — no limit

FILTERS

WELDABLE ONLY HIDE FAMILY-DEFAULT DATA HIDE EXCLUDED

DATA PROVENANCE — READ BEFORE YOU TRUST A RANKING Values are typical room-temperature figures for screening and ranking, not design allowables. Strength is 0.2% proof for metals, failure stress for ceramics and ablatives, and unidirectional fibre-direction failure stress for composites — so the three are not directly comparable, and the ranking table now says which applies. Fracture toughness is indicative only: it varies strongly with condition, orientation and section thickness. Temperature-dependent properties are quoted at ambient unless the material exists only hot. Take design values from the governing material specification, an allowables source such as MMPDS or CMH-17, or your own test data.

Ranking & comparison

Top candidates by the selected index. Tick up to four to compare them side by side.

DESIGN NOTES — DECISIONS, TRAPS AND CHECK SEQUENCES

MATERIAL SELECTION — IN ONE LINE Design to the allowable at temperature, for the exposure time, in the load direction — never to room-temperature typical yield.

The selector's job is to stop designs being sized on the wrong number. For anything that sees heat, the governing figure is strength at temperature after the actual exposure time.

Screening data

Yield is the specification minimum for a common product form. Retention is an approximate fraction of room-temperature yield on short exposure. Screening only — pull the real curve and basis before any analysis.

Alloy	Min yield MPa	150°C	300°C	500°C	ρ g/cm³
Al 6082-T6	260	0.75	0.20	—	2.70
Al 7075-T73	≈430	0.80	0.25	—	2.81
Ti-6Al-4V annealed	≈830	0.85	0.70	0.55	4.43
17-4PH H1025	1000	0.92	0.82	see note	7.80
316L annealed	170	0.85	0.72	0.60	8.00
Inconel 718 aged	1030	0.95	0.90	0.85	8.19

TRAP Precipitation-hardened steels over-age permanently above their ageing temperature, and 17-4PH also embrittles with long exposure in roughly the 300–450°C range. Neither recovers on cooling. Keep service temperature well below the ageing temperature, or pick a different alloy.

Selection order that works

Filter on temperature capability for the exposure duration, not the peak alone.
Filter on environment — galvanic pairing and stress-corrosion susceptibility.
Then optimise on specific strength. Doing this first is how 7075 ends up in a hot joint.
Check the product form. A forging allowable is not your allowable if you machine from plate.

WATCH High-strength aluminium is susceptible to stress-corrosion cracking in the short-transverse direction under sustained tension. For 7075 specify -T73 or -T76 over -T6 and accept the strength penalty; for 2024 use a -T8xx temper.

ZONE 12

Weighted trade study

Score candidates against criteria you weight yourself. Screening in the selector narrows the field; this ranks what survives.

THIS CONSUMES THE SELECTOR The candidates scored here are whatever passed your filters in the material selector, so set the limits there first. A trade study on an unfiltered database ranks materials you have already ruled out.

Weighted trade study

A single performance index answers one question. Real selection balances several, so set the weights that reflect your actual requirement and score the shortlist. Each criterion is normalised across the candidates currently passing your filters, so the score is relative to the field you are choosing from — not an absolute merit.

Criteria weights

Scoring notes

SCORED AT YOUR DESIGN CONDITIONMass efficiency uses yield derated to the design temperature and adjusted for the selected composite layup; toughness uses the selected orientation and band; fatigue uses the Marin-corrected endurance. Change the service condition in the panel above and the ranking changes with it. Materials above their service ceiling are excluded entirely.

HOW THE SCORE WORKSEach criterion is scaled 0–1 across the current candidate set — log-scaled where the property spans orders of magnitude, so one extreme material cannot swamp the result. Scores are then weighted and summed, and normalised so the best candidate reads 100.

A TRADE STUDY IS AN ARGUMENT, NOT AN ANSWERThe ranking only reflects the weights you chose. Change the weights and the winner changes — which is exactly why the weights, not the score, are the part that needs justifying in a design review. Record them.

READ THE NUMBERS PROPERLY These are typical room-temperature values for common product forms — for first-pass selection only. Stress work needs A/B-basis allowables from MMPDS or the governing AMS/DEF STAN spec, at the right thickness, temper, orientation and temperature. Composite strengths are fibre-direction only. The £/kg* column is an order-of-magnitude indication for comparison, not a quotation — get real prices from your supply chain.

SPECIFIC STRENGTH (UTS ÷ DENSITY) — TOP 18 OF CURRENT FILTER. HIGHER = MORE STRENGTH PER KG

TEMPERATURE CAPABILITY vs COST — THE REAL TRADE. UP = HOTTER, RIGHT = PRICIER

How to read the trade

Bottom-left is where you want to live: cheap and adequate. Every step up the vertical axis tends to cost roughly an order of magnitude along the horizontal one — which is why hot components are built as small inserts inside cooler, cheaper structure rather than machined from one piece of the expensive stuff.

THE INSERT PRINCIPLE A tungsten or carbon-carbon throat insert inside a steel or composite shell costs a fraction of a monolithic refractory nozzle and weighs far less. Same logic as carbide tips brazed to a steel tool body.

£/kg IS NOT £/PART Refractories and superalloys machine slowly and eat tooling, so the machining line on your quote grows with the material price. A part in Inconel 718 can cost 10-20× the same part in 17-4PH once cycle time is counted — see the cost tool.

MATERIAL	ρ g/cm³	UTS MPa	YIELD MPa	σ/ρ	MAX °C	£/kg*	CORR.	MACH.	SPECS / NOTES

ALUMINIUM TRAPS 7075-T6 is prone to stress-corrosion cracking in the short-transverse direction — use T73 where SCC matters. 2024 and 7075 are effectively non-weldable and pit badly unclad; protect (anodise / Alocrom + primer). 6082 is the UK workhorse: weldable, cheap, everywhere.

STAINLESS & TITANIUM TRAPS 17-4PH condition matters enormously — always state it (e.g. H1025); H900 is strongest but tough on stress corrosion. Ti-6Al-4V galls: never run titanium on titanium without coating, and use anti-seize on Ti threads. 316L work-hardens — rigid setups, sharp tools.

GALVANIC PAIRS Carbon fibre + aluminium, or steel fasteners in bare aluminium, corrode in service. Isolate: Alocrom/anodise, sealant, cadmium- or Zn-Ni-plated fasteners, or sacrificial washers. Check your programme's approved finishes list.

DEFAULT PICKS General bracketry: Al 6082-T6. Flight structure sheet: 2024-T3 clad. High-strength lugs/fittings: 7075-T73 or Ti-6Al-4V. Shafts/gears: EN24T. Corrosive + strong: 17-4PH. Wet systems: 316L. Wear pads/insulators: POM or PEEK.

DESIGN NOTES — DECISIONS, TRAPS AND CHECK SEQUENCES

CONCEPT DOWN-SELECT — IN ONE LINE Agree the weights before scoring, and treat must-haves as pass/fail rather than weighted points.

A down-select is only defensible if someone else could repeat it and reach the same answer.

Procedure

Kill criteria first. Must-have requirements are pass/fail. A concept that fails one is out, however well it scores elsewhere.
Weights before scores. Agree the weights with the team and record them before anyone scores a concept.
Score against evidence. Each score points to a reason — a calculation, a quote, a precedent.
Sensitivity check. Nudge each weight. If one small change flips the winner, the concepts are effectively equal.
Tie-break on risk. Between equal concepts, pick the one with the fewest unknowns.

TRAP Weights adjusted after scoring until the favoured concept wins. It's common, it's usually unconscious, and it's why the weights have to be written down first.

DESIGN RULE Record the rejected concepts and why. When requirements change later — they will — you'll want to know which rejected option the change brings back into play.

⚠ A weighted score is a way of making your priorities explicit and comparing consistently — it is not a decision. The weights are yours, and the ranking is only as defensible as they are.

ZONE 54

Machined-part cost analysis

Rough-order-of-magnitude estimator for CNC parts from billet. It won't replace a quote, but it shows you where the money goes — which is what you need at design time.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

Inputs

MATERIAL (SETS £/KG & DENSITY — EDITABLE)

SURFACE FINISH SPECIFIED

TOLERANCE GRADE

GEOMETRIC COMPLEXITY

STOCK L (mm)

STOCK W (mm)

STOCK H (mm)

£ / KG

DENSITY g/cm³

QUANTITY

CYCLE (min/part)

RATE £/HR

SETUP (hrs)

TREATMENT £/part

INSPECT £/part

MARGIN %

Result

£0 / part

Internal corner radius sets the smallest usable cutter, and the cutter sets the cycle time. Aim for R ≥ ⅓ of pocket depth, one common radius throughout, and slightly larger than a stock cutter (R5.2 beats R5 — no dwell mark). A truly sharp internal corner cannot be milled at all.

WHAT ACTUALLY DRIVES MACHINING COST Number of setups (every re-fixture ≈ new datums + time), tightest tolerance on the part, smallest internal corner radius (drives cutter size → time), depth-to-width of pockets, and surface finish below Ra 0.8. One needless ±0.01 can double a part's price.

EASY WINS Internal radii ≥ ⅓ pocket depth and one common radius everywhere. General tolerance per ISO 2768-m and only tighten what mates. Design so one face does everything (single setup). Buy stock near net size. Batch parts to amortise setup — see how the per-part price falls as you raise quantity above.

DESIGN NOTES — DECISIONS, TRAPS AND CHECK SEQUENCES

MACHINING COST — IN ONE LINE Setups dominate one-off cost; cycle time and material dominate production cost — and tight tolerances multiply both.

Design moves that cut cost most

Fewer setups — put critical features on faces reachable from one orientation.
Internal corner radii of at least a third of pocket depth, so standard cutters reach.
Pocket depth under about four times tool diameter.
Avoid walls thinner than about 1 mm in aluminium; they chatter.
Relax every tolerance that doesn't serve a function.

ZONE 23

Rocket motor components

Component-by-component material selection and how each gets defined on a drawing. Scope is structural and hardware — cases, closures, nozzles, insulation substrates, seals and interfaces. Propellant formulation, grain design and energetics processing sit with your propulsion and energetics specialists and their safety authority.

☛ CLICK ANY NUMBERED COMPONENT to jump to the tool that sizes or selects it. · SCHEMATIC ARRANGEMENT ONLY — not to scale, no design dimensions implied. Each numbered item is covered below.

The three pressures every case design is judged against.

MEOP, proof and burst

MEOP — maximum expected operating pressure. Everything is referenced to this, and it must include temperature effects and manufacturing spread, not just the nominal.
Proof — applied to every unit (or a defined sample) to demonstrate no yielding or leakage. Typically a modest factor above MEOP.
Burst — the demonstrated ultimate capability, shown by analysis and by test to destruction on qualification articles.

THE FACTORS ARE NOT YOURS TO PICK Proof and burst factors come from the governing requirement — DEF STAN 07-085, the programme design rules, or AIAA S-080/S-081 for pressure vessels — not from a textbook or from precedent on a different programme. Confirm them with your stress and safety authority before sizing anything.

Insensitive munitions (IM) & case design

IM design is about making the hardware fail gracefully when something goes wrong — not about performance. The governing policy is STANAG 4439 with AOP-39; each threat has its own test STANAG and a maximum permitted response.

THE CENTRAL IDEA An unplanned stimulus will ignite the propellant. You cannot usually prevent that. What you can design is what happens next: if the case holds pressure, you get an explosion; if the case opens early and stays open, you get a fire. Every IM feature below is a way of guaranteeing the case opens before it can burst. That is why IM work is a case, closure and venting problem as much as a formulation problem.

Threats and required responses

THREAT	STANDARD	WHAT IT SIMULATES	MAX RESPONSE	DESIGN IMPLICATION

AOP-39 response descriptors

TYPE	RESPONSE	WHAT IT MEANS

READ THE DIRECTION Type I is the worst, Type VI the best — the numbering runs opposite to how people instinctively read it. "Type V or better" means burning, deflagration is already a fail for most threats.

Mitigation features

SLOW COOK-OFF IS THE HARD CASE Fast heating tends to look after itself on a composite motor — the resin softens and the case relieves early. Slow heating brings the whole item up together, so the propellant can reach reaction temperature while the case is still cold-strong and fully intact. Most deliberate venting features exist because of slow cook-off, not fast.

Heritage → modern material lineage

What propulsion hardware used to be made from, what it is made from now, and — more usefully — why it changed. In most cases the driver was manufacturing process, not a better alloy.

THE PATTERN WORTH NOTICING Across almost every line above, the change was driven by how the part is made rather than by a stronger material. Additive manufacturing let designers put cooling channels inside a liner instead of brazing them on, and print refractories that were previously near-impossible to form. When you are selecting a material, ask what process it enables — that is usually where the real gain is.

Material selection & motor performance

Every material choice on a motor lands on one of a handful of performance factors. This maps the class of material to what it actually buys and costs you at motor level, rather than at coupon level.

THE ONE NUMBER THAT MATTERS Propellant mass fraction — propellant mass over total motor mass — dominates delivered performance. Every gram of case, insulation, nozzle and hardware is a gram of propellant you are not carrying. That is why case material choice is the single largest performance lever available to a hardware designer, and why insulation thickness is never "just a bit more to be safe".

PERIPHERAL COARSE GRAIN — A MANUFACTURING RESTRICTION WITH PERFORMANCE CONSEQUENCES Where PCG is prohibited, the minimum stock removal is not free: it sets the minimum purchased section, adds machining time, and can force a heavier starting billet or plate. Establish the requirement before sizing the part, or you will find the finished wall thickness no longer closes after the skin comes off.

DESIGN ORDER THAT WORKS Fix MEOP and the proof/burst factors first, then choose the case material class, then size the wall, then add insulation, then check what mass fraction you actually have left. Choosing the material last — after the geometry is frozen — is how motors end up overweight with no options remaining.

Grain configurations & thrust profile

Grain geometry sets the burning surface area as the web regresses, and burning surface sets the thrust-time trace. As a hardware designer you inherit this: it dictates insulation profile, stress-relief features and where the bondline is loaded.

CONFIGURATION	BURN PROFILE	GEOMETRY	DESIGN CONSEQUENCE

THE THREE PROFILES Progressive — burning surface grows, pressure and thrust rise. Neutral — surface roughly constant, flat trace. Regressive — surface shrinks, thrust falls away. Most motors aim for neutral so the case sees one steady design pressure rather than a climbing one.

Motor design factors & their tensions

Every one of these is a trade, not a target. The second column is what it buys you; the third is what it costs you — which is where the design work actually happens.

⚠ Descriptive engineering reference. Motor ballistic design, propellant selection, IM qualification and all energetics work are governed by JSP 520, DEF STAN 07-085 and your programme safety case, and remain the responsibility of the appropriate qualified specialists and safety authority.

Defining rocket motor hardware on a drawing

MATERIAL BLOCK — SAY MORE THAN THE ALLOY The alloy name alone is not a specification. State: alloy + condition/temper + product form + governing spec + melt practice where it matters. "D6AC BAR PER AMS 6431, VAR, HT TO 1520 MPa MIN UTS, DOUBLE TEMPER" is a specification; "D6AC" is a wish.

SPECIAL PROCESSES NEED THEIR OWN CALLOUTS Heat treatment (with AMS 2750 pyrometry), welding procedure, NDT method + acceptance class, coating, and cleanliness are each separate notes with a separate spec. If it can't be verified by measuring the finished part, it needs an approved process and a certificate.

FRACTURE-CRITICAL PARTS CHANGE THE RULES Once a part is classified fracture-critical, the drawing must carry the flaw-size and NDE requirements that go with it, and material comes from a controlled quality class (ultrasonic grade, melt practice). Get the classification agreed before you finalise tolerances — it drives surface finish, radii and inspection access.

TRACEABILITY IS A DRAWING REQUIREMENT Batch/heat number marking, cert retention, and the marking method (etch vs stamp — stamping is a stress raiser and is banned on many fracture-critical surfaces) all belong on the sheet. State the marking location and method explicitly.

Interfaces that bite

DISSIMILAR METALS Case steel to aluminium closure, or carbon composite to any metal, sets up a galvanic cell that will find every gap in your protective scheme. Specify isolation (sealant, primer, isolating film) and a bonding/earthing path deliberately — they pull in opposite directions.

DIFFERENTIAL THERMAL EXPANSION A refractory insert in a steel or composite housing sees large temperature swings across a joint between materials with very different expansion coefficients. Interference at room temperature and gap at temperature (or the reverse) is a design decision — state the assembly temperature on the drawing when the fit depends on it.

SEALS IN HOT / ENERGETIC ENVIRONMENTS Standard NBR and even FKM assumptions do not carry over. Confirm compatibility with the actual fluids and temperature profile, and check whether your programme mandates specific approved seal materials — many do. Where a seal is safety-significant, its groove tolerances get the same scrutiny as structure.

BONDLINES ARE STRUCTURE Insulation-to-case and liner bonds carry load and are the usual failure path. Define surface preparation, adhesive spec, bondline thickness, cure schedule and the NDT method for verifying the bond — a bond with no inspection requirement is an unverified joint.

⚠ This page is a design-office aide-mémoire for structural hardware. It is not a substitute for DEF STAN 07-085, JSP 520, your programme's design rules manual, or your safety authority. Anything touching energetics, initiation, or motor performance goes through the responsible specialists.

ZONE 24

Solid propellant families & materials compatibility

What the main solid propellant classes are, how they behave as engineering materials, and — the part that actually lands on your drawings — which materials are allowed anywhere near them.

SCOPE OF THIS PAGE This is a descriptive and compatibility reference for hardware designers. It deliberately contains no formulations, proportions, processing parameters, burn-rate tailoring or grain design — those sit with your energetics and propulsion specialists under JSP 520 and your programme safety authority, and nothing here substitutes for their approval. What it does cover is the design-office question: what is this stuff like to build hardware around, and what can touch it.

Propellant families

Materials compatibility

COMPATIBILITY IS PROPELLANT-SPECIFIC — THIS IS THE KEY POINT A material is not "compatible with propellant" in the abstract. Nitrate-ester systems and composite systems degrade by entirely different mechanisms, so the same metal can be prohibited against one and merely controlled against the other. Nine of the sixteen materials below change verdict between families. Read across the row, not down a single column.

How compatibility is actually established

Compatibility is demonstrated by test, not asserted from a datasheet. STANAG 4147 is the governing agreement for chemical compatibility of ammunition components with explosives and propellants; the methods below are its principal tools.

METHOD	WHAT IT MEASURES	APPLICABILITY

WHY THE TEST LIST ITSELF PROVES THE POINT Stabiliser depletion only means something for a nitrocellulose-based propellant — a composite has no stabiliser to deplete. The fact that the standard's methods differ by propellant family is the clearest possible evidence that the compatibility verdicts must too.

THE COPPER RULE If you remember one thing from this page: copper and copper alloys are excluded around nitrate-ester (double-base) propellant. That means no brass fittings, no bronze bushes, no beryllium copper tools, no copper-bearing solder, and no copper plating — in the hardware, the tooling, the handling equipment or the storage area. It is easy to violate accidentally through a bought-in component.

THE SILICONE RULE Silicone is banned from most propellant and bond areas — not for reactivity, but because it migrates, transfers on contact and destroys adhesive bonds. Since case-to-insulation and insulation-to-propellant bonds are structural, a silicone fingerprint can fail a motor. This extends to release agents, sealants, lubricants and hand creams.

COMPATIBILITY IS TESTED, NOT ASSUMED Formal compatibility assessment normally involves accelerated ageing of the material in contact with the propellant and measuring the effect on both. A material being "chemically inert" on a datasheet proves nothing about a specific pairing over a decades-long service life. Ask for the compatibility evidence before you specify.

WHAT THIS MEANS FOR YOUR DRAWINGS Contact materials, finishes, seals, adhesives and even temporary protective materials all need to come from the approved list. State them explicitly and never leave a finish or a seal material to shop discretion on hardware that sees propellant.

⚠ Descriptive reference only. Anything involving energetic materials — selection, processing, handling, storage, disposal or test — is governed by JSP 520, DEF STAN 07-085 and your programme safety case, and is the responsibility of the appropriate qualified specialists and safety authority.

ZONE 58

Material specification builder

Pick a material and get everything you need to write the material block: which standard governs each product form (bar, plate, tube, forging), the available tempers and heat treatments, the finishes you can call up, a copy-ready drawing callout, and the practical tips that stop it going wrong.

Stock forms

What you are actually buying. A section says more than a name — and the form decides grain direction, tolerance class and how much of it you machine away.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

THE THREE-PART RULE A complete metal callout needs three things, and most drawings only carry one or two: (1) the alloy and condition, (2) the product form standard — because properties and tolerances differ between bar, plate and extrusion — and (3) the finish with its own spec. "6082-T6" is a hope. "EN AW-6082-T6 EXTRUDED BAR TO BS EN 755-2, TOLERANCES TO BS EN 755-3" is a specification.

ZONE 16

Metal designation cross-reference

The same alloy has a different name in every system you'll meet. This maps the common name to its BS EN / ISO designation, EN material number, the legacy British spec still on heritage drawings, the UNS number, and the aerospace spec you'd actually procure to.

BS L-series aerospace specifications

What a UK aerospace drawing actually calls up — —. An alloy name is not a callout: the spec fixes form, temper and tolerance as well as chemistry.

TWO SPECS CAN BE THE SAME METAL AND STILL NOT INTERCHANGEABLE L164 and L166 are both 2014A clad sheet in T4, with identical chemistry and mechanical requirements. L166 is held to tighter rolling tolerance. Stock is often dual-certified to both — but a drawing calling L166 is not satisfied by material certified only to L164.

WHICH ONE GOES ON THE DRAWING? In the UK, working to BS 8888, the BS EN designation plus the governing product standard is the correct call-out — e.g. EN AW-6082-T6 TO BS EN 755-2. Add the UNS or AMS number in brackets only where the supply chain genuinely needs it. Legacy BS specs (HE30, 817M40, 304S11) are for reading old drawings, not for writing new ones.

EQUIVALENT IS NOT IDENTICAL Cross-reference tables map nearest grades, not interchangeable ones. Composition ranges, permitted impurities, melt practice, test requirements and acceptance criteria differ between systems. Never substitute across a row on a flight or safety-significant part without engineering approval and a materials review — and be especially careful where the aerospace spec (AMS, AIR) carries melt-practice and cleanliness requirements the commercial EN grade does not.

COMMON NAME	BS EN / ISO DESIGNATION	EN No.	LEGACY BS	UNS	AERO / OTHER SPEC	NOTE

How to read the numbering systems

EN ALUMINIUM — EN 573 EN AW-6082: AW = aluminium wrought (AC = cast). The four digits match the international 6xxx system. Chemical composition is EN 573-3; you also need the product standard for properties — EN 485 sheet/plate, EN 755 extrusion, EN 754 drawn bar/tube.

EN STEEL — EN 10027 Two parallel systems. Names (34CrNiMo6) encode composition: carbon ×100, then alloying elements with factored percentages. Numbers (1.6582) are just catalogue entries — 1.0xxx–1.9xxx steels, 2.xxxx non-ferrous heavy metals, 3.xxxx light metals.

EN STAINLESS — EN 10088 X2CrNiMo17-12-2 = X (high alloy), carbon ×100, elements, then percentages. That is 316L, number 1.4404. Part 2 is sheet/plate, Part 3 is bar/rod/section.

UNS A single letter plus five digits, covering everything: A = aluminium, S = stainless, N = nickel, R = reactive/refractory (titanium), G = carbon/alloy steel, C = copper, K = misc steels. Often the cleanest way to pin down exactly which grade someone means.

ZONE 17

Density register

Every material likely to appear on a rocket programme or in its tooling — metals, alloys, polymers, glass, ceramics, composites, cores, woods and workshop materials. Primary column is kg/mm³ as requested; g/cm³ and kg/m³ sit alongside for sanity checking.

UNITS — READ THIS ONCE kg/mm³ produces very small numbers: steel is 0.00000785 kg/mm³. The conversion is simply g/cm³ ÷ 1,000,000. Values here are shown in full decimal to 9 places — count the zeros carefully. If you are entering density into FEA using the mm / N / tonne consistent unit system (Abaqus, LS-DYNA, most mm-based solvers), the value it wants is tonne/mm³ = 0.00000000785 for steel — that is your kg/mm³ figure × 10⁻³. Mixing the two is the single most common cause of a model coming out 1000× too heavy.

MATERIAL	CATEGORY	kg/mm³	g/cm³	kg/m³	TYPICAL USE

Filled polymer calculator

Reinforced grades are normally specified by weight percent of filler. Density follows the rule of mixtures on specific volume, not a straight linear blend — this calculator uses the correct form.

Inputs

BASE POLYMER

FILLER

FILLER CONTENT — 30% BY WEIGHT

WHY IT MATTERS A 30% carbon-filled nylon and a 30% glass-filled nylon are not interchangeable on mass: glass is heavier per unit volume, so the same weight percent gives a noticeably denser part.

Result

—

CALCULATED, NOT MEASURED This is the theoretical void-free density. Real mouldings contain some porosity and the compounder may add pigments, lubricants or impact modifiers, so measured density usually lands slightly below. Use the datasheet value when one exists.

ZONE 18

Plastics reference

What the acronym means, what it is sold as, what it is good for, and — more usefully — what it is bad at. Sorted from commodity through to the high-performance polymers.

Acronym quick-lookup

ACRONYM	FULL NAME	ALSO SOLD AS	MIN °C	MAX °C	Tg °C	H₂O %	µ	OUTGAS

THE COLD END KILLS MORE SEALS THAN THE HOT END Datasheets lead with maximum temperature, but a polymer below its glass transition is a different material — hard, brittle and with almost no elastic recovery. A seal that is fine at 200 °C can leak on a cold morning. Standard FKM stiffens around −20 °C, which is why low-temperature grades exist; silicone and PTFE stay useful far colder than almost anything else.

VACUUM OUTGASSING For anything flying or in a vacuum chamber, outgassing matters as much as strength — volatiles condense on optics, sensors and cold surfaces. The screening standard is ASTM E595 with limits of TML < 1.0% and CVCM < 0.10%. The ratings here are indicative: verify the specific compound and cure against the NASA outgassing database, because fillers, plasticisers and cure state change the result far more than the base polymer does.

WATER ABSORPTION MOVES DIMENSIONS Nylon absorbs up to about 1.8% and grows accordingly — which is why PA is a poor choice for precision parts and POM or PET is preferred. Absorbed moisture also plasticises the polymer, dropping stiffness and strength while raising toughness.

RADIATION PTFE and the perfluoropolymers are the outlier here: superb chemically, but among the worst materials under ionising radiation, degrading at doses other polymers shrug off. PEEK, PEI, PI and PPS are the radiation-tolerant choices.

TEMPERATURES ARE CONTINUOUS-SERVICE GUIDES Short excursions can go higher; sustained load at temperature must go lower. Any polymer under continuous stress creeps, and the creep rate climbs steeply as you approach these figures — check creep data, not just the temperature rating, before you load a plastic part in service.

ZONE 47

Thread reference — all systems

Every thread system you'll meet on UK aerospace & defence work: metric first, then Unified, Whitworth, BA and the pipe families. Dimensions, tap and clearance drills, tolerance classes, where each system belongs — and the cross-system traps that wreck hardware.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

THE GOLDEN CROSS-SYSTEM RULES 1/4-20 UNC and 1/4 BSW share diameter and TPI but not form angle (60° vs 55°) — they part-assemble then strip. 1/2 UNC is 13 TPI, 1/2 BSW is 12. NPT and BSPT look alike and leak forever when mixed. M10×1.5 and M10×1.25 differ by a quarter-millimetre of pitch. The rule: identify by gauging, never by eye, and state the full designation with pitch/TPI on every callout.

System guides

THREAD ENGAGEMENT — THE COLOURED COLUMN Engagement is computed as (major − drill) ÷ (1.299 × pitch) for 60° forms, and ÷ (1.2806 × pitch) for the 55° Whitworth form. Green 60–80% is the sensible working band. Note the traditional Whitworth tapping drills run tighter (80–85%) than modern metric practice — that is genuinely how the published BSW/BSF tables are, not an error. Open them up slightly in tough materials.

TAP DRILL LOGIC Metric rule: tap drill = nominal − pitch, giving ~77% thread engagement. Engagement above ~75% adds torque and broken taps, not strength — the joint fails by bolt fracture or thread shear long before percentage matters. In tough materials (stainless, Inconel, titanium) drop to ~65%: drill fractionally larger than D−P.

CLEARANCE HOLE LOGIC ISO 273 gives close (H12), medium (H13) and coarse (H14) series. Medium is the default; close only where the fastener locates the part (better: don't — use dowels); coarse for painted, galvanised or multi-part stacks where tolerances accumulate.

WHICH SYSTEM FOR NEW DESIGN Metric coarse, tolerance 6H/6g, full stop — unless the interface forces otherwise: US-origin equipment (Unified, class 2A/2B or 3A/3B), fluid ports (G with face seal), or legacy repair (match the original). Never introduce a new Whitworth or BA thread.

AEROSPACE EXTERNALS Fatigue-critical external threads on US-influenced programmes are UNJ form to AS8879 — enlarged, controlled root radius, rolled after heat treatment. The equivalent metric MJ form is ISO 5855. A standard cut 6g thread is not a substitute where the stress report assumed rolled UNJ/MJ.

DESIGN NOTES — DECISIONS, TRAPS AND CHECK SEQUENCES

THREADS & INSERTS — IN ONE LINE The bolt must yield before the thread strips — a yielded bolt shows itself, a stripped thread doesn't.

Minimum engagement

Tapped material	Engagement	Notes
Steel into steel, similar strength	1.0 × d	Standard nut height is sized on this
Steel into titanium	1.2–1.5 × d	Galling risk — lubricate
Steel into aluminium	1.5–2.0 × d	Use 2× if cycled
Steel into magnesium	2.5–3.0 × d	Insert strongly preferred

These are rules of thumb. Calculate stripping strength for critical joints — VDI 2230 covers it.

MJ threads for fatigue-critical joints

Aerospace MJ threads (ISO 5855) have a controlled, larger root radius and modified tolerances. The root is where fatigue cracks start, so the radius matters more than any other thread detail. Pair with rolled after heat treatment — the compressive residual stress in the root greatly improves fatigue life over cut threads. Write the phrase in full on the drawing; “rolled thread” alone doesn't guarantee the sequence.

When to use an insert

More than a handful of assembly cycles into aluminium.
Blind tapped holes in expensive parts.
Anywhere a stripped thread would mean a scrapped flight part.

Wire inserts improve load share across engaged threads. Keyed solid inserts resist vibration and torque-out better.

Galling

Titanium and austenitic stainless gall against themselves and each other. Avoid same-alloy pairs, use silver-plated nuts or dry film lubricant, and record the effect of the lubricant on torque.

TRAP Never let a thread run-out, tapped hole or cross-drilling break into a sealing surface or fillet. It creates a leak path that's usually found at proof test.

ZONE 51

O-ring design calculator

Enter the hardware you have, get the seal and groove you need — squeeze, gland fill, stretch, extrusion gap and back-up ring advice, all checked against BS ISO 3601-2 targets.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

ON THE STANDARD The O-ring standard is BS ISO 3601: Part 1 gives inside diameters, cross-sections and tolerances; Part 2 gives the housing (gland) dimensions; Part 3 quality acceptance; Part 4 anti-extrusion (back-up) rings; Part 5 material suitability. (BS 3601 on its own is the carbon steel pipe and tube standard — different thing entirely, easy to mistype.)

Inputs

APPLICATION

SIZE SERIES

CROSS-SECTION mm

BORE DIAMETER mm

SYSTEM PRESSURE bar

SERVICE TEMPERATURE °C

MEDIA

DIAMETRAL CLEARANCE mm (gap at the extrusion face)

SQUEEZE TARGET — 22%

Recommended seal & groove

—

Groove geometry

Material selector

Materials are checked live against the temperature and media you entered above. Green = suitable, amber = limited or check the specific grade, red = do not use.

CODE	MATERIAL	TEMP RANGE	FIT	NOTES & WATCH-OUTS

Media compatibility matrix

SURFACE FINISH Groove and counterface finish matters as much as the dimensions. Typical guidance: static Ra 1.6 max on sealing surfaces, dynamic Ra 0.4 max on the sliding counterface, and Ra 3.2 on non-sealing groove walls. Too rough leaks; too smooth (below about Ra 0.1) can starve a dynamic seal of lubricant.

LEAD-IN CHAMFERS Every edge the O-ring passes over on assembly needs a 15–20° lead-in chamfer, deburred and blended. More O-rings are killed during assembly than in service — a sharp bore edge or an unbroken cross-hole will shave the ring and you will never see it once assembled.

GROOVE CORNER RADII Keep bottom corners small (roughly 0.1–0.3 mm depending on cross-section) so the ring seats properly, but never leave them sharp — a sharp corner is a stress raiser in the ring and a crack initiator in the metal.

WHAT THIS TOOL IS NOT A first-pass sizing aid using the ISO 3601-2 target bands. It does not replace the standard, the seal manufacturer's catalogue, or a compatibility test with your actual fluid at your actual temperature. Confirm the final size and tolerances against ISO 3601-1 before release.

DESIGN NOTES — DECISIONS, TRAPS AND CHECK SEQUENCES

O-RING DESIGN — IN ONE LINE Check squeeze at the worst tolerance corner and the coldest temperature, with the joint gap at its widest under pressure.

Design the squeeze, then let the groove follow. Every seal works at nominal — the design is judged at the corners.

Target squeeze and fill

Duty	Squeeze	Gland fill
Static face seal	20–30%	75–85%
Static radial seal	15–25%	75–85%
Reciprocating dynamic	10–18%	70–80%

Extrusion limits pressure

The seal fails by being forced into the clearance gap. Screening values below are diametral clearance (radial gap is half). Use the supplier's chart for design.

Pressure	70 Shore A	90 Shore A
< 5 MPa	≤ 0.15 mm	≤ 0.25 mm
5–10 MPa	≤ 0.08 mm	≤ 0.15 mm
> 10 MPa	backup ring	≤ 0.08 mm or backup ring

Use the widest gap in service: maximum clearance, at maximum temperature, with the bore dilated under pressure. Checking only the cold assembly gap misses the case that fails.

The joint moves — the seal must follow

Under pressure, a joint opens slightly: flanges lift, bores dilate, mating parts rotate. An elastomer seal only keeps sealing if it springs back fast enough to follow that movement.

LESSON The 1986 Challenger loss traced to a joint whose gap opened under pressure while its O-rings were too cold to follow it. Two checks apply to any pressure joint: how far the gap opens under load, and whether the seal can track it at minimum temperature.

Cold is the dangerous end

Near its glass transition an elastomer stops behaving like a spring. Below TR-10, sealing relies almost entirely on squeeze built into the metal. Specify the compound by low-temperature capability and check squeeze with both parts contracted.

Other checks

Compression set — budget permanent loss into minimum squeeze, especially after long storage.
Thermal swell — elastomer expansion is many times the metal's. Check gland fill hot, at minimum groove volume. Over ~90% risks hydraulic lock.
Finish — Ra 0.8 static, Ra 0.4 dynamic; machining lay must not cross the seal.
Lead-in — 15–20° chamfer, deburred, or the seal is cut on assembly unseen.

ELASTOMER COMPOUNDS — IN ONE LINE Pick the compound for the coldest, hottest and chemically worst moment — in that order.

Elastomer families cover broad ranges and individual compounds vary widely within them. Use this to shortlist, then select a specific compound from a supplier with data for your fluid and temperature.

Family	Typical range °C	Good with	Bad with
NBR (nitrile)	−30 to +100	Mineral oils, many fuels	Ozone, weathering, ketones, glycol brake fluid
HNBR	−30 to +150	Oils, better heat and ozone than NBR	Ketones, strong acids
FKM (fluorocarbon)	−20 to +200	Oils, fuels, many chemicals	Ketones, esters, hot water/steam; standard grades poor when cold
Low-temp FKM	−40 to +200	As FKM, better cold	Cost, still check TR-10
EPDM	−50 to +150	Water, steam, glycols, phosphate-ester hydraulics	Any petroleum oil or grease
VMQ (silicone)	−60 to +200	Dry heat, wide temperature	Poor tear and abrasion, fuels — static only
FVMQ (fluorosilicone)	−60 to +175	Fuels and oils when cold	Weak mechanically, static preferred
FFKM	−15 to +300	Nearly everything	Cost, poor low temperature

TRAP Assembly lubricant. An EPDM seal greased with a petroleum-based lubricant swells and fails, even though the service fluid is compatible. Specify the assembly lubricant on the drawing alongside the compound.

Other checks

Low temperature — use TR-10, not the brittle point. Brittle point is far colder than where sealing performance is lost.
Shelf life — record cure date; age-control and storage conditions follow your company's spec.
Compression set — at temperature and time, not just the datasheet's standard condition.
Explosive decompression — for gas service with rapid depressurisation, use an ED-rated compound.

ZONE 42

ISO 286 limits & fits

Enter a size and a tolerance class, get the actual limits in millimetres, the resulting clearance or interference, and what kind of fit you have actually specified.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

HOLE BASIS IS THE DEFAULT Keep the hole at H and vary the shaft letter. The hole is made by a tool of fixed size — drill, reamer, broach — so it is far cheaper to hold the hole constant and grind the shaft to suit. Shaft basis (shaft at h) is for drawn bar and shafting used as-supplied, or where one shaft carries several different fits along its length.

Inputs

FIT TYPE

PREFERRED FIT

NOMINAL SIZE mm (over 0 to 500)

HOLE LETTER

HOLE GRADE

SHAFT LETTER

SHAFT GRADE

Result

—

Tolerance zones

IT grades at this size

WHAT THE GRADE MEANS IN PRACTICE IT01–IT4 gauge and instrument work · IT5–IT7 fits and bearing seats · IT8–IT9 general machining · IT10–IT11 sheet, castings and clearance holes · IT12–IT16 as-cast, as-forged and flame-cut. Every grade tighter roughly doubles the cost of the feature — do not specify H6 where H9 works.

ISO 2768 general tolerances

The tolerance that applies to every dimension you did not tolerance individually. Part 1 covers linear and angular dimensions, Part 2 covers geometry. In the UK these are adopted as BS EN 22768-1 and -2.

CHECK WHICH STANDARD YOUR PROGRAMME USES ISO 2768-2 has been superseded by ISO 22081, which replaces the fixed H/K/L geometry classes with a general geometrical specification referenced to stated datums. Many organisations still work to 2768 and plenty of legacy drawings carry it, so both are in circulation — confirm which applies before you put it in a title block.

Lookup

NOMINAL DIMENSION mm

LINEAR CLASS (ISO 2768-1)

GEOMETRICAL CLASS (ISO 2768-2)

Applicable tolerances

—

Part 1 — linear dimensions, permissible deviation ± mm

Part 1 — angular dimensions, by length of the shorter side

Part 2 — straightness & flatness

Part 2 — perpendicularity, symmetry & runout

WHAT IT DOES NOT COVER ISO 2768 says nothing about roundness or cylindricity — those are limited by the size tolerance. Parallelism is not tabulated either: it takes the larger of the size tolerance and the flatness/straightness value. Anything functionally important still needs its own explicit tolerance.

IT IS NOT A LICENCE TO STOP TOLERANCING A general tolerance covers the dimensions that genuinely do not matter. If a feature locates something, seals, or carries load, tolerance it individually — and remember that a general tolerance still has to be achievable and inspectable on every dimension it silently applies to.

THE USUAL CALL-OUT ISO 2768-mK is the common general-purpose pairing: medium linear with class K geometry. Write it in the title block, not buried in a note, and make sure the class you pick matches the process — specifying -f on a sand casting is meaningless.

BROKEN EDGES External radii and chamfer heights get their own, much looser table — typically ±0.2 mm up to 3 mm for classes f and m. Do not assume the linear tolerance applies to an edge break.

Preferred fits

HOLE BASIS	SHAFT BASIS	TYPE	DESCRIPTION & TYPICAL USE

DATA PROVENANCE IT grades and the fundamental deviations for c, d, e, f, g, h, js, k, m, n and p are the published ISO 286 table values. Letters r, s and u are now published table values on the standard's finer sub-steps. Only t and x remain formula-derived — confirm those against ISO 286-2 before releasing an interference fit.

INTERFERENCE FITS NEED MORE THAN LIMITS A press fit also needs assembly force, hoop stress in the outer part, and the temperature differential if you intend to shrink it on. Check the fit is still adequate at maximum service temperature — a shrink fit that relaxes when hot has no retention at all.

DESIGN NOTES — DECISIONS, TRAPS AND CHECK SEQUENCES

FITS & CLEARANCES — IN ONE LINE Choose the fit from function across the full temperature range, then let ISO 286 supply the letters.

Hole-basis fits

Fit	Type	Use
H11/c11	Loose running	Wide tolerance, dirt, thermal movement
H9/d9	Free running	Large temperature changes, high speed
H8/f7	Close running	Accurate running at moderate speed
H7/g6	Sliding	Accurate location, parts move freely
H7/h6	Locational clearance	Snug fit, can be assembled by hand
H7/k6	Transition	Accurate location, small clearance or interference
H7/n6	Transition	More accurate, light press
H7/p6	Locational interference	Rigid alignment, no load transfer by friction
H7/s6	Medium drive	Permanent; carries some torque
H7/u6	Force	Highly stressed, shrink fit

Temperature

A fit is chosen at 20°C but has to work at the extremes. An aluminium housing on a steel shaft gains clearance hot and loses it cold. Recalculate the fit at both ends — see Differential expansion.

Tolerance stacks

Worst case — guaranteed, but often over-tight with many contributors.
Statistical (RSS) — valid only with several independent contributors from capable, centred processes. Without that evidence, RSS is optimism.

TRAP Interference fits between dissimilar metals. An aluminium housing on a steel bearing can lose its grip at the hot end and overstress at the cold end. Check both.

ZONE 48

Bolted joint & torque calculator

Preload, tightening torque, stress area and thread engagement for metric fasteners — with the friction sensitivity shown, because friction is what actually decides whether your torque produces the preload you wanted.

TORQUE IS A POOR PROXY FOR PRELOAD — READ THIS FIRST Roughly 85–90% of applied torque is consumed by friction under the head and in the threads. Only the remainder stretches the bolt. A change in friction coefficient from 0.10 to 0.16 — entirely plausible between a lubricated and an as-received fastener — changes the achieved preload by around 40% for the same torque. Published torque tables disagree with each other by 10–15% for exactly this reason. This tool follows the VDI 2230 convention — 90% of Rp0.2 reduced for tightening torsion, with D_km = 1.36·d — and reproduces published VDI values within about 10%. Treat any torque figure, including this one, as a starting point subject to your own joint analysis.

Inputs

THREAD SIZE

PROPERTY CLASS

HEAD TYPE

FRICTION CONDITION

PRELOAD — 75% OF PROOF LOAD

TAPPED MATERIAL (for engagement)

Result

—

Friction sensitivity

The same torque across the range of plausible friction conditions. The spread is the reason torque-controlled tightening has such poor preload accuracy.

CONDITION	µ	TORQUE FOR TARGET PRELOAD	PRELOAD AT FIXED TORQUE	vs TARGET

Where the torque goes

BETTER THAN TORQUE Angle-controlled (torque-to-yield) tightening, bolt elongation measurement, ultrasonic length measurement, or load-indicating washers all give far better preload accuracy than a torque wrench. If the joint matters, control something closer to preload than torque.

ENGAGEMENT AND STRIPPING The engagement figures here are the usual rules of thumb. A proper check compares the thread shear area of the internal and external threads so the bolt fails in tension before the tapped thread strips — a bolt failure is detectable, a stripped thread often is not.

HYDROGEN EMBRITTLEMENT Class 10.9 and 12.9 fasteners are susceptible after electroplating. Specify the relief bake to BS EN ISO 4042, with the time limit after plating, or use mechanically plated or non-electroplated finishes.

STAINLESS GALLS A2 and A4 fasteners seize on assembly with alarming regularity, especially in coarse threads and at speed. Specify an anti-seize or dry film lubricant, and note that doing so changes the friction and therefore the torque.

DESIGN NOTES — DECISIONS, TRAPS AND CHECK SEQUENCES

BOLTED JOINTS — IN ONE LINE Size the joint on minimum preload after losses — the question is whether the joint opens, not whether the bolt breaks.

Dividing external load by bolt area is the wrong calculation for a preloaded joint, and it's wrong in the unsafe direction for a sealed one.

Where external load actually goes

The clamped parts are usually several times stiffer than the bolt, so most applied tension unloads the clamped parts rather than adding to the bolt.

ΔF_bolt = Φ · F_ext Φ = n · K_b / (K_b + K_c) n = load introduction factor (VDI 2230), typically 0.3–0.7 metal-to-metal joints: Φ often ≈ 0.05–0.25

Preload governs, not bolt area. Upsizing the bolt in the same flange stiffens the bolt and raises Φ — it can make fatigue worse.

Tightening factor αA

αA = F_max / F_min for the method. Design separation on F_min; check bolt strength on F_max.

Method	αA	≈ scatter
Ultrasonic elongation measurement	1.1–1.2	±5–9%
Yield-controlled tightening	1.2–1.4	±9–17%
Angle-controlled tightening	1.2–1.4	±9–17%
Hydraulic tensioning	1.2–1.6	±9–23%
Torque, friction found by test on the joint	1.4–1.6	±17–23%
Torque, friction estimated	1.6–2.5	±23–43%
Impact wrench	2.5–4	never structural

TRAP About 90% of applied torque goes into friction under the head and in the threads. A torque value with no stated friction coefficient and lubricant is not an instruction. Put μ and the lubricant in the drawing note.

Losses before checking separation

Embedment — surface asperities flatten in the first hours. Use VDI 2230's values by roughness and number of interfaces; a few microns per interface is typical. Short grips lose a much higher percentage of preload.
Thermal differential — see below and Thermal growth.
Coating or gasket creep — soft platings and gaskets in the clamp stack dominate everything else.

aim for grip ratio l / d ≥ 3 — a long thin bolt is a forgiving spring

Thermal effect on preload

Aluminium flange, steel bolt, heating — flange grows more, preload rises. Check the bolt against yield hot.
Steel flange, titanium bolt, heating — preload falls. Check separation hot.

Self-loosening

Transverse slip of the clamped parts rotates the nut loose — this is the dominant loosening mechanism, not axial vibration. The real defence is enough clamp force that friction at the interface exceeds transverse load with margin, so the joint never slips. Locking features are a second line, covered in Locking & retention.

Shear joints

Decide whether the joint carries shear by friction (slip-critical — preload and interface friction carry it) or bearing (bolt shank bears on hole walls). Mixing the two in one analysis is a common error. If position must hold after slip, use dowels or a spigot.

Check sequence

Minimum preload after all losses, at the worst temperature.
Separation: remaining clamp exceeds (1 − Φ) · F_ext with margin. For a sealed joint, keep the seal's contact stress — stricter than open-gap.
Slip: clamp × friction × interfaces exceeds transverse load with margin.
Bolt stress at maximum preload plus external plus thermal.
Thread stripping and bearing under the head.
Fatigue on the alternating component, which Φ has already made small. If it isn't small, the joint is under-preloaded.

ZONE 52

Galvanic corrosion & dissimilar metals

Aerospace hardware mixes aluminium, titanium, steel, stainless and carbon composite in the same joint. Pick any two and find out whether that joint will survive, which part corrodes, and what to do about it.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

THE THREE THINGS A GALVANIC CELL NEEDS Two materials of different potential, an electrical connection between them, and an electrolyte bridging them. Remove any one and there is no corrosion. That is the whole basis of every mitigation below — you either move the materials closer in potential, break the electrical path, or keep the water out.

The couple

MATERIAL 1

MATERIAL 2

ENVIRONMENT

AREA RATIO — cathode : anode = 1 : 1

AREA RATIO IS THE MULTIPLIER EVERYONE FORGETS A large cathode driving a small anode concentrates all the attack into a tiny area. A steel fastener in an aluminium panel is survivable; an aluminium fastener in a steel or carbon panel is eaten. Get the small part on the cathodic (noble) side.

Verdict

—

Mitigation

Galvanic series

Anodic index in volts. Higher is more anodic and corrodes preferentially; lower is more noble and drives the attack. The two selected materials are highlighted.

INDEX V	MATERIAL	BEHAVIOUR	ΔV TO SELECTION 1

CARBON COMPOSITE IS NOT INERT Carbon fibre is almost as noble as gold, so a CFRP-to-aluminium joint is one of the most aggressive couples in normal engineering — roughly 0.80 V. It is a genuine and frequent cause of structural corrosion. Always isolate: a glass-fibre ply on the bond face, a sealant layer, or a titanium or stainless interface fitting.

WHY CADMIUM WAS EVERYWHERE Cadmium plating sits almost exactly alongside aluminium in the series, so a cadmium-plated steel fastener in aluminium structure is galvanically near-neutral. That is precisely why aerospace used it for decades. It is being eliminated for toxicity, and the modern replacement — zinc-nickel — is chosen to sit in a similar place.

PASSIVE FILMS DO THE WORK Stainless and titanium behave nobly only while passive. In a crevice starved of oxygen the film cannot repair and the potential shifts markedly more active, so a joint that looks fine on this chart can still corrode in the crevice under a washer or in a faying surface. Design the joint to drain and to be sealed.

THIS IS A SCREENING TOOL Anodic indices are nominal values for the MIL-STD-889 approach. Real behaviour depends on alloy, temper, surface condition, electrolyte chemistry, temperature and crevice geometry. Use it to spot problems early — then confirm against your programme's dissimilar-metals standard.

DESIGN NOTES — DECISIONS, TRAPS AND CHECK SEQUENCES

GALVANIC COMPATIBILITY — IN ONE LINE Galvanic corrosion needs dissimilar metals, an electrolyte and contact — remove any one of the three.

Anodic index (approximate)

Voltage difference between two metals indicates galvanic risk. Values are approximate — verify against MIL-STD-889 for your exact alloys.

Metal	Index V
Gold, platinum	0.00
Graphite, CFRP	≈0.25
Nickel, titanium, Monel	0.30
Copper	0.35
Stainless steel, 18% Cr	0.50
Stainless steel, 12% Cr	0.60
Wrought aluminium, 2000 series	0.75
Carbon and low-alloy steel	0.85
Wrought aluminium, other series	0.90
Cadmium	0.95
Zinc	1.25
Magnesium	1.75

Environment	Max difference
Harsh — salt spray, humidity, outdoor	0.15 V
Normal — sheltered, no condensation control	0.25 V
Controlled — temperature and humidity controlled	0.50 V

Area ratio

Small anode, large cathode is the dangerous combination. An aluminium rivet in a stainless sheet corrodes rapidly; a stainless fastener in an aluminium sheet is far more tolerable.

TRAP Carbon fibre is strongly cathodic. Aluminium fastened directly to CFRP corrodes fast. Use titanium fasteners and a glass-ply or sealant isolation layer.

Mitigation, in order of effectiveness

Choose compatible metals.
Isolate them electrically — washers, sleeves, a barrier ply.
Exclude the electrolyte — wet-install fasteners with sealant or primer.
Coat the cathode as well as the anode. Coating only the anode concentrates attack at any scratch.

ZONE 25

Motor case & pressure vessel sizing

Required wall thickness at MEOP, predicted proof and burst pressures, critical flaw size and case mass — for any material in the database, at your design temperature.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

COMPOSITE CASES USE THE LAYUP YOU SELECTED IN THE MATERIAL SELECTOR Composite strength here is the layup-adjusted value, not the unidirectional datasheet figure. A filament-wound case should use the filament wound layup; leaving it on unidirectional will size the case roughly 40% too thin. Metals are unaffected.

THIS TOOL CONSUMES THE REST OF THE SITE Yield is derated to the design temperature using the retention curves, fracture toughness comes from the min/typ/max range at the orientation you selected, and the fatigue check uses the Marin-corrected endurance limit. Change the design temperature or crack orientation in the material selector and the answers here change with it.

Inputs

MATERIAL

DESIGN BASIS

GEOMETRY

INTERNAL DIAMETER mm

MEOP — maximum expected operating pressure, bar

CYLINDRICAL LENGTH mm (for mass)

DESIGN TEMPERATURE — 20 °C (set in the material selector)

MANUFACTURING / WELD EFFICIENCY

Required wall

—

Pressure schedule

Material comparison at this duty

The same vessel in every material that can carry it — ranked by case mass, which is what actually drives propellant mass fraction.

MEOP IS NOT THE NOMINAL BALLISTIC PRESSURE MEOP must already include grain temperature sensitivity, manufacturing spread in burn rate and web, and any erosive-burning or ignition spike. Sizing a case on the nominal chamber pressure and then discovering the hot-day value is 15% higher is a classic and expensive mistake.

MEMBRANE STRESS IS THE EASY PART This tool gives you the membrane wall. It does not size the domes, the closure joints, the boss or skirt attachments, or anything with a discontinuity — and those are where cases actually fail. Bending at a dome-to-cylinder junction can double the local stress. Treat this as the starting thickness, not the answer.

THIN VS THICK WALL Below about t/r = 0.1 the thin-wall (Barlow) result is within a few percent of the Lamé solution. Above that it becomes unconservative — the tool computes both and tells you which applies.

LEAK BEFORE BURST The critical flaw size tells you whether a crack will grow through the wall and vent, or run unstably. If the critical flaw is smaller than the wall thickness, the vessel can fail catastrophically without leaking first — and your NDT must reliably find flaws well below that size.

Joint separation — the question preload was only half of

MOST OF AN EXTERNAL LOAD DOES NOT REACH THE BOLT A preloaded joint is two springs in parallel. When pressure pulls the closure, the bolt stretches a little more — but the clamped parts unload by much more. Only the fraction Φ reaches the bolt, and Φ is typically 0.15 to 0.35. Sizing a closure bolt for the full separating force overstates the load by three to six times.

Joint

BOLT DIAMETER mm

CLAMP LENGTH L_k mm

PRELOAD kN

EXTERNAL LOAD PER BOLT kN

CLAMPED MATERIAL E GPa

EMBEDDING f_Z µm

LOAD INTRODUCTION n

TIGHTENING METHOD α_A

How the load divides

—

⚠ Screening tool for early sizing. Real pressure vessel design is governed by AIAA S-080/S-081, DEF STAN 07-085, EN 13445 or ASME VIII as applicable, requires discontinuity and fatigue analysis, and must be signed off by the responsible stress and safety authority.

ZONE 37

FEA & structural test methods

Choosing the right analysis for a motor component, the traps that make a plausible-looking result wrong, and how each physical test correlates back to the model.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

THE ONLY QUESTION THAT MATTERS Not “what is the stress?” but “what is the failure mode, and does my analysis capture it?” A linear static run on a buckling-critical skirt gives a comfortable stress and a completely wrong answer. Pick the analysis from the failure mode, not from what is convenient to set up.

Analysis types

Motor component → analysis map

COMPONENT	GOVERNING FAILURE MODE	ANALYSIS REQUIRED

Three material classes, three different analyses

THE CATEGORY ERROR A motor contains a metal case at 200 GPa bonded to a propellant grain at 5 MPa — a stiffness ratio of about 40,000. They are not the same analysis with different numbers in it. They need different constitutive models, different elements, and different failure criteria. Treating a grain like a metal is the most common serious mistake in motor structural analysis.

ASPECT	STRUCTURAL HARDWARE	COMPOSITE PROPELLANT	DOUBLE-BASE PROPELLANT

Six traps specific to motor analysis

Mesh quality — the numbers that decide whether a result is real

Every solver reports these. Most engineers never look at them, and a single bad element at a fillet can dominate the peak stress you report.

METRIC	IDEAL	GOOD	MARGINAL	REJECT	WHY IT MATTERS

Element selection

ANALYSIS TASK	USE	AVOID	REASON

THE FIRST-ORDER TETRAHEDRON PROBLEM 4-node tets are artificially stiff — they under-predict displacement and therefore under-predict stress, which is the dangerous direction. They are the default in most automatic meshers because they fill any geometry. If you must use tets, use second-order (10-node) ones. A first-order tet mesh that looks beautifully dense can still be wrong by 20–30% on a bending-dominated part.

Mesh convergence

Refine, plot, and report the convergence — not just the final number.

Physical test methods & FEA correlation

Model verification checklist

Run these before believing any result. Most bad FEA fails one of them, not the physics.

THE UNITS TRAP, SPECIFICALLY In the mm–N–tonne system almost every mm-based solver uses, density must be tonne/mm³: steel is 7.85e-9, not 7850 and not 7.85e-6. Get it wrong and modal frequencies and inertia loads are out by a factor of roughly 1000, while the static stress looks perfectly fine. The density register gives the correct figure in every unit.

SINGULARITIES ARE NOT STRESSES A sharp re-entrant corner, a point load or a fixed node produces a stress that rises without limit as you refine the mesh. It never converges because the true elastic solution is infinite there. Either model the real fillet radius, or assess the region by a method that does not depend on peak stress.

CORRELATE, THEN PREDICT An uncorrelated model is a hypothesis. Correlate against strain gauges or a modal survey first, understand any disagreement, and only then use it to predict conditions you have not tested.

MARGIN IS ONLY AS GOOD AS THE ALLOWABLE State which allowable you used and its statistical basis. A-basis, B-basis and typical values can differ by 20% or more, and a margin quoted against a typical value is not a margin at all.

DESIGN NOTES — DECISIONS, TRAPS AND CHECK SEQUENCES

FEA SETUP CHECKLIST — IN ONE LINE Check boundary conditions, reactions and convergence first — read stresses last.

Checklist, in order

Units — consistent throughout. The most common silent error.
Materials — correct properties at the right temperature; linear or nonlinear as appropriate.
Elements — no linear tetrahedra in bending regions; use quadratic elements.
Boundary conditions — model the real support. Fully fixed faces are stiffer than reality and create false stress peaks at the constraint.
Contacts — bonded only where the parts really are bonded. Bolted joints usually need contact and preload.
Reactions — sum the reaction forces; they must equal the applied loads.
Deformed shape — exaggerate it and ask whether it looks physically right.
Convergence — refine at the point of interest until the result changes by less than about 5%.
Hand-calc comparison — see Hand calcs.
Only now, read stresses.

TRAP Singularities. At sharp re-entrant corners, point loads and point constraints, stress rises without limit as the mesh refines. It isn't real and it doesn't converge. Model the real fillet, or assess with linearised stresses or a hot-spot method.

Pressure hardware

For pressure-retaining parts, linearise stresses through the thickness into membrane, membrane-plus-bending and peak, and assess each against its own limit — the approach used in pressure vessel codes such as ASME VIII Division 2.

Report must state

Software and version, units, element types and mesh size at critical areas, materials and their basis, loads and boundary conditions with justification, convergence evidence, and reaction balance.

⚠ Guidance for engineering judgement, not a substitute for a qualified stress analyst. Structural substantiation of flight or pressure hardware must be performed and signed off by the responsible stress authority under the governing programme requirements.

ZONE 43

Tolerance stack-up

Build a dimension loop, get the worst-case and statistical result, and see which contributor is actually costing you the tolerance.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

THE POINT IS NOT THE TOTAL — IT IS THE RANKING Any stack tells you whether the assembly closes. The useful output is which contributor dominates, because tightening anything else is wasted money. In an RSS stack, contribution scales with the square of the tolerance, so the largest contributor is disproportionately dominant — halving the biggest tolerance beats tightening three small ones.

Dimension loop

METHOD

PROCESS CAPABILITY C_p — 1.33

SPEC LIMIT — MIN

SPEC LIMIT — MAX

RUN MONTE CARLO (50,000 TRIALS) INCLUDE THERMAL EXPANSION

Result

—

Thermal expansion

Assign a material to any row and the stack is recomputed at temperature. CTE comes live from the material database.

GD&T in a linear stack

How each control converts to a ± contribution. Every conversion below is applied automatically when you pick a GD&T row.

THE FACTOR-OF-TWO ERROR A diametral zone is a circle, so only its radius acts along any one stack direction. Position Ø0.2 contributes ±0.1, not ±0.2. Using the full diameter doubles the apparent contribution and is the single most common GD&T mistake in stack-ups.

MMC BONUS BELONGS IN THE STACK When a position tolerance carries Ⓜ, the feature gains bonus tolerance as it departs from MMC — up to the full size tolerance band. Leaving it out makes the stack look tighter than it is and leads to tightening parts that never needed it. Only position controls accept a bonus; form and orientation controls cannot.

Monte Carlo & predicted defect rate

Worst case and RSS give you a band. Monte Carlo gives you a distribution and a number you can act on: how many assemblies actually fail.

Process capability reference

PROCESS	TYPICAL ±	NOTE

Contribution analysis

Where the variation actually comes from. Spend your tolerance budget at the top of this list.

CONTRIBUTOR	NOMINAL	TOLERANCE	WORST-CASE SHARE	RSS SHARE

WHEN RSS IS NOT VALID RSS assumes contributors are independent and normally distributed. It is wrong for one-offs and small batches, for parts from the same mould cavity or bar, where a single setup drives several dimensions, or where a process is skewed or bimodal. If in doubt on a safety-critical fit, use worst case.

GD&T IN A LINEAR STACK A position tolerance of Ø0.2 contributes ±0.1 in any one direction, not ±0.2 — the diametral zone is a circle, and only its radius acts along your stack direction. Getting this wrong doubles the apparent contribution.

THE 50% RULE OF THUMB If one contributor exceeds roughly half the RSS variance, the stack is effectively controlled by that one dimension. Improve it before touching anything else — and consider whether a datum or design change removes it from the loop entirely.

THE FOUR MISTAKES THAT WRECK A STACK 1. A minus sign flips the nominal contribution, never the tolerance — tolerances always add. 2. Mixing equal-bilateral and limit dimensions without converting first; shift the nominal to an equal ± form before stacking. 3. Using RSS on a low-volume or uncentred process, where it over-promises badly. 4. Forgetting bonus tolerance at MMC — it belongs in the stack, and ignoring it makes you tighten parts that did not need it.

THE √n RULE RSS reduces the stack by roughly √n against worst case: 4 contributors gives about 50%, 16 gives about 75%. That is why splitting a tolerance across more features can help — and why a stack dominated by one large contributor never benefits from RSS.

CLOSE THE LOOP Every stack must form a closed chain from one face of the gap, through the parts, back to the other face. If the loop does not close, the answer is meaningless no matter how careful the arithmetic. Draw it before you calculate it.

ZONE 5

Metallurgy fundamentals

Why materials behave the way the selector says they do — crystal structure, strengthening mechanisms, and the transformation diagrams behind every temper and condition on the site.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

THE THROUGH-LINE Every temper code, H-condition and heat treatment elsewhere on this site is a position on one of the diagrams below. T6 is the peak of the ageing curve; T73 is deliberately past it. A quench-and-temper condition is a path across the TTT curve. Once you can see that, the designation systems stop being arbitrary.

Crystal structure — and why it decides cryogenic behaviour

Drag any cell to rotate it. Atom positions are computed from fractional lattice coordinates, so the structures are correct by construction rather than drawn by hand.

AUTO-ROTATE SHOW NEAREST-NEIGHBOUR BONDS

THE MOST IMPORTANT CONSEQUENCE ON THIS WHOLE SITE BCC metals have a ductile-to-brittle transition; FCC metals do not. That single fact is why austenitic stainless, aluminium, copper and nickel are used at cryogenic temperature and ferritic steel is not — a carbon steel that is perfectly tough in the workshop can shatter like glass at −100 °C. It is also why the density register lists 9% Ni steel separately: the nickel suppresses the transition far enough to make a ferritic steel usable for LNG.

Iron–carbon phase diagram

The map underneath every steel on this site. Carbon content sets what phases can exist; temperature sets which of them actually do.

THE EUTECTOID POINT0.77% carbon at 727 °C is where austenite transforms directly to pearlite with no intermediate two-phase region. Steels below it are hypoeutectoid — ferrite plus pearlite — and that includes almost every structural and aerospace steel on this site.

WHY MOST STRUCTURAL STEEL SITS AT 0.3–0.5% CEnough carbon to form useful martensite on quenching, little enough to stay weldable and tough. Above roughly 0.5% the steel hardens more but cracks readily on welding and quenching — 4340 at 0.40% and 4130 at 0.30% are both deliberately in this window.

Time–temperature–transformation

What you actually get depends on how fast you cool through the nose of the curve. Same steel, three different materials.

HARDENABILITY IS NOT HARDNESSAlloying (Cr, Mo, Ni) pushes the curve right, giving more time to miss the nose. That is hardenability — the ability to harden deeper, not harder. Maximum hardness comes from carbon content alone. This is why 4340 through-hardens in thick section while plain carbon steel of the same carbon content does not.

THE THREE DESTINATIONSPearlite from slow cooling — soft, ductile, layered ferrite and cementite. Bainite from intermediate rates — a good strength/toughness balance, and the basis of austempering. Martensite from missing the nose entirely — a diffusionless shear to a body-centred tetragonal lattice, extremely hard and extremely brittle until tempered.

Precipitation hardening — the ageing curve

Where every T-temper and H-condition on this site actually sits.

WHY T73 EXISTS AND IS WEAKER ON PURPOSE7075-T6 sits at the peak. 7075-T73 is aged past it, giving up roughly 10–15% strength to buy a large gain in stress-corrosion resistance. The coarser, incoherent precipitates that cost you strength are also the ones that stop hydrogen and stress cracking the alloy. In the selector, T6 and T73 differ by exactly this trade.

OVER-AGEING IS ALSO WHAT HAPPENS IN SERVICEThe curve is a function of time and temperature. A peak-aged alloy held hot enough for long enough will walk down the right-hand side of this curve on its own. That is precisely what the temperature retention curves in the material selector are modelling.

Strengthening mechanisms

There are only six ways to make a metal stronger. Every alloy on this site uses one or more of them.

Heat treatments

TREATMENT	WHAT HAPPENS	WHY YOU WOULD SPECIFY IT

⚠ Teaching diagrams, drawn schematically to show the governing relationships. Phase boundaries, curve positions and transformation temperatures shift with composition, grain size and prior processing — take real values from the alloy's own published diagrams and specification.

ZONE 44

Surface finish & metrology

What each process can actually achieve, what finish each application needs, and why Ra on its own is not enough to specify a functional surface.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

EVERY STEP FINER COSTS REAL MONEY Going from Ra 3.2 to Ra 0.8 usually means adding a grinding operation. Going to Ra 0.1 means adding honing or lapping on top of that. Specify the roughest finish that works — over-specified surface finish is one of the most common and least noticed ways drawings become expensive, and it rarely gets questioned because it looks like diligence.

Process capability

Achievable Ra bands by process. Select a target and the chart shows which processes can reach it.

Target

REQUIRED Ra — 1.6 µm

OR PICK BY APPLICATION

Equivalents

—

Application guide

APPLICATION	TYPICAL Ra	WHY

Parameter relationships

Ra ALONE DOES NOT SPECIFY A FUNCTIONAL SURFACE Ra is an average. Two surfaces with identical Ra can have completely different peak structures — one a plateau with fine valleys, the other covered in sharp peaks. For seals, bearings and fatigue-critical faces, specify the parameter that matches the function: Rz or Rmr (bearing ratio), not just Ra.

DIRECTION MATTERS TOO Lay direction is part of the specification and is often the thing that actually fails. A dynamic seal running across a turned spiral will pump fluid along the groove and leak however good the Ra is. Specify the lay symbol, not just the number.

THE FATIGUE CONNECTION Surface finish is not cosmetic on a loaded part. It is the k_a factor in the Marin fatigue correction in the GD&T zone — a ground surface at Ra 0.4 keeps roughly 90% of endurance strength, while an as-forged surface keeps about 35%. That is a 2.5× difference in fatigue life driver from finish alone.

MEASURE WHAT YOU SPECIFY Cut-off length and filter change the number. A roughness measurement without a stated cut-off is not comparable to anything. Also confirm the feature is accessible to a stylus — specifying Ra 0.4 inside a small blind bore may be unmeasurable in practice.

ZONE 55

Welding & joining

Process selection, the carbon-equivalent preheat decision, what actually goes wrong in a weld, and which materials fight back.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

THE WELD IS NOT THE PARENT MATERIAL A fusion weld gives you three different materials in the same joint: weld metal, heat-affected zone, and unaffected parent. Design to the weakest of them. In 6082-T6 the HAZ drops to near-annealed strength — roughly half the parent yield — and it stays that way. Sizing a welded aluminium joint on parent properties is one of the most common and most expensive mistakes on this site's material list.

Carbon equivalent & preheat

Composition

PRESET

CARBON %

MANGANESE %

CHROMIUM %

MOLYBDENUM %

VANADIUM %

NICKEL %

SECTION THICKNESS mm

Preheat assessment

—

Process selection

PROCESS	THICKNESS	SPEED	QUALITY	COST	TYPICAL USE

Weldability by material

MATERIAL	WELDABILITY	WHAT TO WATCH

Defects — cause and consequence

PLANAR DEFECTS ARE THE DANGEROUS ONES Lack of fusion and cracks are crack-like and behave as pre-existing flaws in a fracture assessment. Rounded porosity of the same volume is far less severe. This is why acceptance standards treat them completely differently — and why the critical flaw size in the case-sizing zone matters so much for a welded case.

RADIOGRAPHY CAN MISS THE WORST DEFECT A lack-of-fusion plane lying parallel to the X-ray beam may produce almost no image contrast. Ultrasonic inspection detects planar defects far more reliably. Choose the NDT method for the defect you fear, not for convenience.

DESIGN THE JOINT, NOT JUST THE WELD Move welds away from stress concentrations and fatigue-critical areas. The weld toe is a built-in notch — grinding it smooth or peening it can double fatigue life at almost no cost, and is usually cheaper than upsizing the section.

DISTORTION IS THE UNDER-ESTIMATED RISK It is not an NDT defect so it does not appear in acceptance criteria, yet it scraps more parts than porosity does. Balance the weld sequence, back-step, fixture properly, and leave machining stock on anything that must hold tolerance afterwards.

DESIGN NOTES — DECISIONS, TRAPS AND CHECK SEQUENCES

WELD & BRAZE DETAIL — IN ONE LINE Place welds where stress is low and inspection is easy — never in a fillet radius or at a stress peak.

Fusion welding

Strength — use the joint efficiency from the code and the qualified procedure. Parent-metal allowables don't apply to the weld or the heat-affected zone.
Heat-affected zone in aluminium — heat-treated 6xxx and 7xxx alloys lose temper next to the weld. HAZ yield can drop to around half the -T6 value or below.
Titanium — needs full inert shielding on both faces. Weld colour is an acceptance criterion: silver to light straw acceptable, blue and grey rejected under typical aerospace criteria.
Process — electron beam and laser give narrow welds and low distortion; TIG is flexible but puts in more heat.

Design rules

Full-penetration welds in primary load paths; fillet welds are for secondary structure.
Keep welds away from geometric stress raisers and from each other.
Plan NDT access — radiography needs line of sight, ultrasonic needs a coupling surface.
Machine critical features after welding; distortion is certain.
Specify post-weld heat treatment where the alloy needs it.

Brazing

Capillary joints need controlled clearance, typically around 0.025–0.125 mm depending on filler — check the filler datasheet. Vacuum brazing avoids flux entrapment, which matters for anything that must be clean inside.

TRAP A weld symbol without a procedure, an acceptance standard and an inspection method is incomplete. The welder will use a house default.

⚠ Guidance for design decisions. Welding of pressure or flight hardware requires a qualified procedure (WPS/PQR to BS EN ISO 15614 or equivalent), qualified welders, and NDT to the governing acceptance standard such as BS EN ISO 5817.

ZONE 56

NDT method selection & probability of detection

Choosing the inspection for the defect you actually fear, and checking that what you can detect is smaller than what will kill the part.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

THE QUESTION THAT DECIDES EVERYTHING Not “what NDT shall we specify?” but “is my critical flaw size larger than what my inspection can reliably find?” If the critical flaw is smaller than the detectable flaw, you cannot inspect your way to safety — no amount of NDT makes that part safe, and the design must change. This is the single most important idea in fracture control and it is routinely missed.

Inspectability check

Your flaw

CRITICAL FLAW SIZE mm (from the tolerance-stack zone)

FLAW LOCATION

SAFETY FACTOR ON DETECTABILITY — 2.0×

WHY A FACTORDamage-tolerance practice assumes the largest flaw that could be missed is present from day one. Requiring the critical size to exceed the detectable size by a margin is what makes the inspection meaningful rather than nominal.

Verdict

—

Detection capability

Indicative a_90/95 — the flaw size found with 90% probability at 95% confidence. Your critical flaw size is marked.

Method comparison

METHOD	REACH	a_90/95	MATERIALS	STANDARD	FINDS / MISSES

POD IS A STATISTICAL STATEMENT, NOT A GUARANTEE a_90/95 means 90% of flaws that size are found, with 95% confidence in that figure. It explicitly accepts that one in ten flaws at the detection limit is missed. That is why damage tolerance assumes a missed flaw is present rather than assuming inspection is perfect.

MATCH THE METHOD TO THE DEFECT ORIENTATION Radiography can miss a lack-of-fusion plane lying across the beam; magnetic particle misses cracks parallel to the field; ultrasonic misses flaws parallel to the beam. Every method has a blind orientation. Choose for the defect you expect from the process, which is what the welding zone tells you.

SURFACE CONDITION CHANGES EVERYTHING Penetrant needs a genuinely clean, open surface — machining or shot peening can smear metal across a crack and hide it entirely. Etching before penetrant is standard practice on machined fracture-critical parts for exactly this reason.

PERSONNEL QUALIFICATION IS PART OF THE SPECIFICATION NDT results depend on the operator as much as the equipment. Aerospace work requires certification to NAS 410 or EN 4179 at the appropriate level. An uncertified inspection is not evidence.

DESIGN NOTES — DECISIONS, TRAPS AND CHECK SEQUENCES

INSPECTION & NDT — IN ONE LINE Choose the NDT method by the smallest flaw that must be found — which comes from fracture analysis, not habit.

Methods

Method	Finds	Blind to
Penetrant	Surface-breaking cracks, porosity	Subsurface; cracks smeared shut
Magnetic particle	Surface and near-surface, ferrous only	Austenitic stainless, Al, Ti
Eddy current	Surface and near-surface cracks, conductive metals; good in bolt holes	Deep defects; needs calibration standards
Ultrasonic	Planar subsurface defects, thickness	Defects parallel to the beam
Radiography	Volumetric — porosity, inclusions	Tight cracks normal to the beam
CT	Internal geometry, additive porosity	Size, density, cost

Link to fracture control

For fracture-critical parts, fracture analysis gives the largest flaw that can survive the service life. The NDT method must reliably detect flaws smaller than that — with demonstrated probability of detection. If no practical method can, the options are a proof test that screens out the critical flaw size, a redesign, or a lower stress.

Leak testing

Method	Approximate sensitivity
Bubble test	≈10⁻³–10⁻⁴ mbar·L/s
Pressure decay	≈10⁻³–10⁻⁴ mbar·L/s, volume-dependent
Helium mass spectrometer	down to ≈10⁻⁹ mbar·L/s and below

State the allowable leak rate, test gas, pressure and method. “No leaks” is not a requirement — every seal leaks at some rate.

TRAP Penetrant after machining or shot peening can miss real cracks because the metal is smeared over them. Etch first, or inspect before the smearing operation.

What a callout must carry

Method and sensitivity level
Acceptance standard and class
Coverage — surfaces and depth
Process stage — before or after machining and surface treatment

⚠ a_90/95 values are indicative of well-controlled aerospace inspection. Actual capability must be established by a programme-specific POD demonstration on representative specimens, geometry and access.

ZONE 13

Material datasheet

Everything this site knows about one material, assembled from every zone into a single page.

ONE MATERIAL, EVERY ZONE The data below is not duplicated — it is pulled live from the material selector, the property tables, the toughness ranges, the retention curves, the designation register, the density register, the spec builder, the weldability guidance and the galvanic series. If a value changes anywhere on this site, it changes here.

MATERIAL

ZONE 10

Cryogenic design

Material suitability, contraction and the two things that will actually bite you: the ductile-to-brittle transition, and titanium in oxygen.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

CRYSTAL STRUCTURE DECIDES THIS, NOT STRENGTH FCC metals have no ductile-to-brittle transition; BCC metals do. That single fact from the metallurgy zone governs almost every cryogenic material decision. Austenitic stainless, aluminium, copper and nickel stay tough to liquid helium temperature and get stronger as they cool. A carbon steel that is perfectly tough in the workshop can fracture like glass at −100 °C. Choose on structure first, strength second.

Contraction calculator

Component

MATERIAL CLASS

SERVICE FLUID

LENGTH mm

DIAMETER mm (for fit check)

Contraction

—

Material suitability by crystal structure

Liquid oxygen compatibility

In LOX the question is ignition, not strength.

MATERIAL	LOX SERVICE	NOTE

THE TITANIUM TRAP Titanium has the best specific strength in the database and is prohibited in liquid oxygen. It reacts violently with LOX and can ignite from impact or from a fresh scratch exposing bare metal. The material selector will happily rank Ti-6Al-4V first for a pressure vessel — if that vessel holds LOX, the answer is wrong. No performance index catches this.

FITS CHANGE ON COOLDOWN A clearance fit at room temperature is not a clearance fit at 20 K if the two parts are different materials. An aluminium housing on a steel shaft tightens by roughly 0.2% of diameter over that range. Size the fit at temperature, not on the bench — The fits and limits zone gives the room-temperature limits; this zone gives the movement.

HYDROGEN IS TWO PROBLEMS AT ONCE Liquid hydrogen brings cryogenic temperature and hydrogen embrittlement. Embrittlement of austenitic stainless is worst around 200 K rather than at the coldest point, so a system that is fine at 20 K can be damaged during chilldown and warm-up. Metastable grades like 304L are the most affected because strain-induced martensite forms on deformation.

COMPOSITES MICROCRACK BEFORE THEY BREAK Carbon/epoxy gains strength and stiffness cold, but the resin contracts far more than the fibre and the mismatch drives transverse microcracking. The cracks open permeation paths long before any structural failure — this is what ended the X-33 composite LH2 tank programme, and why liners and toughened resins exist.

CLEANLINESS IS A STRUCTURAL REQUIREMENT IN LOX Hydrocarbons in liquid oxygen are an explosion hazard. LOX hardware is cleaned to a specified particulate and non-volatile-residue level, assembled with oxygen-compatible lubricants only, and kept sealed. A perfectly designed part contaminated with cutting fluid is a hazard.

⚠ Contraction values are integrated figures for material classes, adequate for sizing and screening. Take alloy-specific data from NIST cryogenic property databases or the material specification for final design, and confirm LOX compatibility by the governing standard such as ASTM G93 and G72.

ZONE 36

FEA material cards

Elastic constants, a Ramberg-Osgood plasticity table in Abaqus format, and an estimated S-N curve — at your design temperature.

READ THIS BEFORE YOU PASTE ANYTHING INTO A SOLVER None of the curves below are test data. They are generated from the published monotonic properties this site holds — E, R_p0.2 and R_m — using standard models. Three tiers, marked on every value:

EXACT — the engineering-to-true conversion. Pure mathematics, no assumption.
DERIVED — the Ramberg-Osgood curve, computed from properties held in the database.
ESTIMATED — the hardening exponent n and the S-N curve. Class-typical values, not measurements.

Use these to get a model running and to explore sensitivity. Replace n and the curve with MMPDS, CMH-17 or your own coupon data before any analysis anyone relies on. Where a material cannot be honestly represented by these models, this page says so and gives you nothing rather than guessing.

Material & condition

MATERIAL

DESIGN TEMPERATURE °C — 20

RAMBERG-OSGOOD n — 16 ESTIMATE — OVERRIDE ME

TABLE POINTS

Elastic constants

—

Stress–strain curve

Abaqus material card

Copy straight into an input deck. Units are N, mm, MPa, tonne — density is in tonne/mm³, the trap the FEA zone warns about.

—

Estimated S-N curve

Basquin fit between a 0.9×R_m anchor at 10³ cycles and the Marin-corrected endurance limit. An estimate, not test data.

WHERE RAMBERG-OSGOOD IS A POOR MODEL Mild steel with a Lüders plateau yields at constant stress — RO cannot represent that at all. Annealed austenitic stainless hardens so heavily that a single exponent fits badly across the whole range. Magnesium has strong tension-compression asymmetry that no isotropic model captures. In all three cases use a measured curve.

THE CURVE STOPS AT ULTIMATE FOR A REASON Beyond the ultimate point the specimen necks and the engineering stress-strain relationship stops being a material property — it becomes a structural response of that specimen. Abaqus needs true stress against true plastic strain up to necking; past that you need a corrected curve from a Bridgman analysis or inverse FE fitting.

TEMPERATURE SCALING Strength is scaled by the retention curves from the material selector — the same validated model the selector and case sizing use. Modulus is scaled by a smaller factor, since E falls more slowly than yield for most metals. Both are estimates: at temperature, get real data.

FOR A TIME-DEPENDENT MATERIAL, THIS IS THE WRONG CARD Everything on this page is rate-independent plasticity. Polymers, elastomers, propellant grains and bondlines are viscoelastic — their stiffness depends on how fast you load them and how long you wait. A plasticity table cannot express that at any number of points. The viscoelasticity zone generates the *VISCOELASTIC, TIME=PRONY card you need instead, with the WLF shift for temperature.

WHAT IS GENUINELY EXACT HERE The engineering→true conversion is σ_true = σ(1+ε) and ε_true = ln(1+ε), with plastic strain ε_pl = ε_true − σ_true/E. No assumptions, and it is the step most often got wrong when hand-building a card.

⚠ Screening and model-setup aid only. Design allowables and material cards for released analysis must come from MMPDS, CMH-17, the governing specification, or your own qualified test programme.

ZONE 7

Viscoelasticity

Why some materials answer “solid or liquid?” with “how long are you watching?” — the models, the experiments, and the card that goes into a solver.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

THE ONE SENTENCE VERSION In an elastic material stress depends on strain. In a viscoelastic material stress depends on the whole history of strain — on how fast you got there and how long you have been waiting. Every complication in this zone follows from that one change.

WHERE THIS LEADS The molecular sandbox shows the mechanism behind these curves, and the propellant grain zone applies all of it to real hardware — grain margin, bondlines and storage life.

Ten ways of seeing it

Work through these in order. Each builds on the one before.

The models, and what each one gets wrong

MODEL	RELAXATION	CREEP	GOOD FOR / FAILS AT

The two experiments

Everything is measured one of two ways, and they are mirror images of each other.

Controls

TEMPERATURE °C — 20

PRONY ARMS SHOWN — 8

At this temperature

—

Watch it happen — creep and relaxation

The two experiments, run on the same material. Pick a model and see why the classical ones are not enough.

Experiment

MODEL

TEST

HOLD FOR — 60 s, then release

What the model says

—

The master curve — heat as a time machine

Short tests at several temperatures, slid along the time axis until they overlap into one curve spanning far longer than any test.

Prony series — straight into a solver

The generalised Maxwell coefficients in Abaqus form, with the WLF constants supplied as a thermorheologically simple shift.

—

gᵢ ARE NORMALISED, AND MUST SUM TO LESS THAN 1 Abaqus wants dimensionless relaxation moduli: gᵢ = Eᵢ / E₀, where E₀ is the instantaneous modulus. The remainder 1 − Σgᵢ is the long-term fraction. If Σgᵢ reaches 1 the material relaxes completely and the solver will reject it — a common error when a fit is scaled wrongly.

THE ELASTIC CARD MUST MATCH Give *ELASTIC the instantaneous modulus, not the room-temperature datasheet value, or state MODULI=LONG TERM and give the relaxed one. Mixing the two is the other common error, and it silently scales every stress in the model.

Dynamic mechanical analysis

Push it sinusoidally. What comes back is E′, what turns to heat is E″, and the ratio finds the transitions.

⚠ Teaching material. The Prony spectrum and WLF constants used in the charts are illustrative of the behaviour, not measurements of any real propellant. Grain structural analysis requires characterisation of the actual material to the governing programme requirements.

ZONE 8

Molecular sandbox

A live simulation of what is actually moving, a glossary of every particle, and a controlled side-by-side comparison. This is the mechanism behind the curves in the viscoelasticity zone.

Why nitration makes a molecule fast

SCOPE — WHAT THIS IS AND IS NOT This explains the physical chemistry of why nitrated molecules release energy quickly: oxygen balance, and what it means to carry your own oxidiser. It is standard energetics-course material, and it is what you need to reason about propellant stability and ageing.

It contains no synthesis route, no conditions, no quantities, and nothing about sensitivity, initiation or handling. The rest of this site treats propellant the same way — described structurally, never as something to produce.

Nitration level

–OH REPLACED BY –ONO₂ — 0 of 3

Oxygen balance

—

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

Molecular sandbox — a live simulation

Real Langevin dynamics: beads joined by springs, with excluded volume so chains genuinely cannot pass through one another. Move the cursor over the box to push the molecules around. Drag to rotate the view.

WHAT THIS IS, AND WHAT IT IS NOT This is a qualitative sandbox, not a quantitative model. It honestly shows thermal motion scaling with temperature, plasticiser moving faster than the chains it lubricates, a star polymer diffusing slower than a linear one, chains blocking one another, and binder dewetting from filler particles. It does not reproduce elastic recovery: the bonds here are enthalpic springs, while real rubber elasticity is entropic — chains recoil because a coiled state has more entropy, and nothing in this model represents that. I tested for recovery across damping and crosslink density and it does not appear, so the page does not claim it.

POLYMER

TEMPERATURE — 0.45

BOND STIFFNESS — 60

DAMPING — 0.10

CURSOR FORCE — 110

CURSOR RADIUS — 0.45

RUN LOCK VIEW — DRAG TO GRAB CHAINS SHOW BONDS MOTION TRAILS

Cursor pushes · drag rotates · arrow keys rotate when focused

What each particle is

Every species in the simulation, what it physically is, and what it does mechanically.

Side by side — the same material, two temperatures

The sandbox above is free-running. This is a controlled comparison: one identical chain network, drawn at a cold reference and at your chosen temperature, with the excursion volume showing how far each segment can move. Drag either panel to rotate.

ARCHITECTURE

TEMPERATURE °C — 20

OBSERVATION TIME — 1 s

MOTION SPEED — 1.0×

ANIMATE THERMAL MOTION SHOW CHAIN BONDS SHOW ENTANGLEMENTS SHOW EXCURSION VOLUME

⚠ Qualitative teaching model. The parameters are chosen to make the behaviour legible, not to reproduce any specific material.

ZONE 26

Propellant grain

Viscoelasticity applied to hardware: what a double-base grain is structurally, whether it survives cold soak, whether the bondline holds, and what twenty years of storage does to both.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

SCOPE This zone covers the mechanical and structural behaviour of the grain — the same discipline as NASA SP-8073, and the reason a grain does or does not survive storage. Formulation, energetics, burn rate and internal ballistics are outside the scope of this site.

Double-base propellant

SCOPE This section covers the mechanical and structural behaviour of the material — the same discipline as NASA SP-8073, and the reason a grain does or does not survive storage. Formulation, energetics, burn rate and internal ballistics are outside the scope of this site.

A double-base propellant is, structurally, a plasticised polymer: a nitrocellulose backbone with nitroglycerine acting as plasticiser. The NC is the structural material; everything above applies to it directly, and rather strongly — heavily plasticised cellulosics are among the most temperature-sensitive engineering polymers there are.

Three relaxations, not one

DMA on double-base propellant finds three distinct transitions. Which one you call T_g changes every shift factor you compute.

Three states

Bondline stress

Interfacial shear from thermal mismatch. It is not uniform — it concentrates at free edges, which is why debonds start at terminations.

Stack

CASE

PROPELLANT

LINER / INSULATION

CURE TEMP °C

SERVICE TEMP °C

AGE — 0 years (bond strength falls with age)

Bond margin

—

Ageing & storage life

Arrhenius kinetics on a single dominant mechanism — what an accelerated test actually proves, and how the properties drift meanwhile.

Programme

MATERIAL

ACTIVATION ENERGY kJ/mol — 110

STORAGE TEMP °C

TEST TEMP °C

TARGET LIFE years

TEST DURATION days

What the test proves

—

Grain structural margin

Bore strain from cure cool-down and from pressurisation, against capability at temperature.

Motor

PROPELLANT

CASE

BORE RADIUS mm

GRAIN OD RADIUS mm

CURE TEMP °C

SERVICE TEMP °C

MEOP bar

GRAIN RESTRAINT

TAKE VISCOELASTIC RELAXATION CREDIT ON COOL-DOWN

Margin

—

Grain structural load cases

LOAD CASE	WHAT IT IS	WHY IT MATTERS

WHY CRACKING IS A SAFETY PROBLEM, NOT A PERFORMANCE ONE A crack in a grain is new burning surface. More surface means more gas generated per second, which means higher chamber pressure, which can mean the case fails. That is why grain structural integrity is treated as a safety discipline and why cold-soak cracking is taken so seriously — the structural failure and the pressure excursion are the same event.

AN ELASTIC ANALYSIS GIVES THE WRONG ANSWER TWICE Use slow-test data on a fast event like ignition and you under-predict stiffness and stress. Use fast data on a slow event like cure cool-down and you over-predict stress, because months of relaxation genuinely bleed it away. Viscoelastic analysis is not a refinement here — the elastic answer is wrong in a different direction for each load case.

AGEING MOVES THE CURVES Over decades of storage, plasticiser migrates and evaporates and stabiliser is consumed. The material stiffens, tan δ falls and the transitions shift. A propellant characterised when new is not the material you have at twenty years, which is why storage-life prediction is done by tracking modulus and transition temperature rather than by chemistry alone.

PRONY SERIES GO STRAIGHT INTO A SOLVER The generalised Maxwell coefficients are what Abaqus wants for *VISCOELASTIC, TIME=PRONY, with the WLF constants supplied separately for temperature dependence. A fit is only valid over the time and temperature range it was fitted to — extrapolate it and you are inventing data.

⚠ Screening only. The Prony spectrum, WLF constants and propellant properties used here are illustrative of the behaviour, not measurements of any real material. Grain structural analysis requires characterisation of the actual material to the governing programme requirements.

ZONE 9

Worked example

One case-bonded motor taken through every phase, using the site's own tools. Change the inputs and the whole chain recomputes.

EVERY NUMBER BELOW IS COMPUTED, NOT TYPED Each step calls the same function the zone itself uses, so this page cannot drift out of step with the tools. If a model changes, this example changes with it — and the regression suite asserts the chain end to end.

CASE MATERIAL

MEOP bar

CASE INNER RADIUS mm

PROPELLANT

CURE TEMP °C

COLD SOAK °C

WHAT A WORKED EXAMPLE CANNOT DO This is the arithmetic chained together correctly — it is not a design. It uses screening values, illustrative propellant properties and generic factors. A real programme substitutes MMPDS or specification allowables, measured propellant characterisation, and the factors its own requirements impose. The value here is seeing how the steps connect and where one answer becomes the next question.

ZONE 27

Ablative liners

How thick a nozzle liner or case insulator has to be: what erodes away, and what has to stay cool behind it.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

AN ABLATIVE DOES TWO JOBS AT ONCE It sacrifices its surface — pyrolysing, charring and receding, carrying heat away with the mass it loses — and it insulates what is behind it. Size for only the first and the structure cooks; size for only the second and the liner burns through. The thickness you need is both, plus margin.

HEAT FLUX IS AN INPUT, NOT AN OUTPUT This site does not compute heat flux — that comes from your own ballistic and gas-dynamic analysis. Enter what your analysis gives you. Everything below is the thermal and materials consequence of that number.

Duty

LINER MATERIAL

STRUCTURE BEHIND

HEAT FLUX MW/m²

BURN DURATION s

DESIGN FACTOR — 1.5

RECESSION RATE mm/s ESTIMATE — OVERRIDE

Required thickness

—

Why the two jobs pull in opposite directions

LOW RECESSION DOES NOT MEAN GOOD INSULATION Carbon-carbon and graphite barely erode, and conduct heat a hundred times better than carbon phenolic. They protect the surface superbly and the structure hardly at all. A graphite throat insert almost always needs a separate insulator behind it.

PENETRATION GROWS AS √t Doubling the burn time increases thermal penetration by √2, not 2. A long-burn motor is less punishing per second than the arithmetic suggests — but recession is linear in time, so the balance between the two shifts as duration grows.

WHAT IS EXACT HERE Thermal penetration is x = 3.64√(αt), the depth at which the temperature rise falls to 1% of the surface step in a semi-infinite solid, with α = k/ρc_p. No fitting, no assumption beyond the idealisation.

WHAT IS NOT Recession rate is an estimate. It depends on flux, gas composition, particle loading, shear and the specific material lay-up — none of which this site knows. The default scales with flux from a class-typical datum. Replace it with a measured rate from your own firings or supplier data before sizing anything real.

⚠ Preliminary sizing aid. A real nozzle liner is sized with a transient thermal model using measured material response, and confirmed by test. Char, pyrolysis-gas blowing and in-depth reactions are not modelled here.

ZONE 32

Shell buckling

Thin shells fail far below their theoretical strength. How far below, and when it governs at all.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

THIS IS WHY EIGENVALUE BUCKLING IS ALWAYS UNCONSERVATIVE A linear eigenvalue analysis solves the perfect shell. Real shells have out-of-roundness, weld shrinkage, thickness variation and handling dents, and a cylinder in axial compression is extraordinarily sensitive to them. A slender case skirt can carry less than a quarter of the classical value. The knockdown below is the industry's answer, and it is empirical: a lower bound fitted to test data, not a physical law.

Shell

MATERIAL

RADIUS mm

WALL mm

LOAD CASE

DESIGN FACTOR — 1.5

Allowable

—

THE KNOCKDOWN IS A LOWER BOUND, NOT A PREDICTION SP-8007's curve was drawn beneath a scatter of test results from the 1960s, much of it on shells built to standards a modern filament-wound or friction-stir case would beat comfortably. It is safe and often very conservative. Programmes that can afford the testing measure their own imperfection sensitivity instead.

AXIAL AND EXTERNAL PRESSURE ARE DIFFERENT PROBLEMS Axial compression scales with t/r and is imperfection-dominated. External pressure scales with (t/r)³ — far more sensitive to wall thickness, and much less sensitive to imperfections. A shell comfortable in one can be marginal in the other.

CHECK WHETHER IT GOVERNS AT ALL For a thick enough wall the buckling allowable exceeds yield, and the shell simply yields first. The tool says which mode governs — running a buckling analysis on a part that yields first is effort spent on the wrong failure.

WHAT TO DO WHEN IT GOVERNS Thicken the wall, shorten the unsupported length, or add rings or stringers. Stiffening is usually far lighter than thickening, because the classical stress rises only linearly with t while mass rises linearly too — a ring buys stability for very little mass.

⚠ Preliminary sizing. Real buckling assessment uses measured imperfection data, non-linear analysis with realistic geometry, and test. Boundary conditions, unsupported length and combined loading all matter and are not modelled here.

ZONE 33

Leak before break

When a crack finally goes through the wall, does the vessel hiss — or does it come apart?

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

THE SAFEST FAILURE IS ONE YOU CAN HEAR A crack growing in service eventually reaches the back face. If it penetrates before it goes critical, the vessel leaks: pressure drops, the leak is detectable, and the failure is benign. If it goes critical while still buried, the crack runs and the vessel bursts with no warning at all. Which of the two you get is a design decision, whether or not anyone made it deliberately.

Vessel

MATERIAL

WALL mm

HOOP STRESS MPa

GEOMETRY FACTOR Y

MEAN RADIUS mm (0 = flat plate)

MARGIN ON WALL — 1.0×

Verdict

—

Every material at this duty

A THICKER WALL CAN MAKE THINGS WORSE Thickening a wall lowers stress, which helps — but it also means a crack must grow further before it breaks through. Past a certain thickness the flaw reaches critical size while still buried, and a vessel that would have leaked now bursts instead. Leak-before-break is a property of the wall relative to the toughness, not of the material alone.

IT IS AN ARGUMENT, NOT A CALCULATION A real leak-before-break case has to show that the leak is detectable in time to act — leak rate, detection method, crew or system response. A vessel that leaks into a sealed compartment nobody monitors has gained very little. This tool answers only the fracture-mechanics half.

IT PAIRS WITH INSPECTION, NOT INSTEAD OF IT If leak-before-break holds, a missed flaw leaks rather than bursts — which is exactly the defence you want when NDT cannot reliably find the critical size. Check both together: the case sizing zone gives the critical flaw, the NDT zone gives what you can find, and this zone gives what happens if you miss it.

HIGH-STRENGTH ALLOYS FAIL THIS TEST Toughness generally falls as strength rises. A maraging or 7075 case sized to a high working stress can have a critical flaw smaller than its own wall — there is no leak path, only a burst. That is the real cost of chasing strength, and it does not appear anywhere in a static stress margin.

⚠ Linear-elastic screening only. A released leak-before-break argument uses J-integral or R-curve analysis, measured toughness at temperature and orientation, realistic flaw shapes, and a demonstrated leak-detection route. ASME and NASA-STD-5019 set the real requirements.

ZONE 35

Proof testing

A proof test does not demonstrate that a vessel is sound. It sorts the ones that are from the ones that are not — and tells you how long the survivors have.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

PROOF IS A SORTING OPERATION Any vessel carrying a flaw large enough to fail at proof pressure fails during the test — safely, on the ground, with nobody near it. Every survivor therefore carries a flaw smaller than the critical size at proof pressure. That is not a statement about workmanship; it is a guarantee produced by the test itself, and it holds even for flaws no inspection could find.

Test & duty

MATERIAL

OPERATING STRESS MPa

PROOF RATIO — 1.25×

STRESS RANGE PER CYCLE — 80% of operating

GEOMETRY FACTOR Y

GROWTH LAW

What the test buys

—

THE PROOF TEST ITSELF GROWS CRACKS Pressurising to proof applies the single largest stress the vessel will ever see. It can extend a sub-critical flaw, and in a susceptible material it can leave the vessel worse than it found it. This is why proof pressures are kept as low as the argument allows, and why repeated proofing is not a substitute for inspection.

A HIGHER PROOF RATIO BUYS LIFE AS R² Because critical flaw size goes as 1/σ², raising the proof ratio shrinks the guaranteed flaw as 1/R² — and the smaller starting flaw then takes longer to grow. The gain is real but it is bought against the risk above, and against losing good hardware in the test.

IT COVERS WHAT INSPECTION CANNOT NDT finds flaws down to its detection limit and no further. A proof test screens by consequence rather than by detection — it does not care whether the flaw was findable. Where the critical size sits below what NDT can reliably see, proof is the argument that remains.

THE GROWTH LAW IS THE WEAK LINK Everything after the screened flaw size rests on Paris constants that vary with environment, frequency, temperature, R-ratio and orientation. They are class-typical estimates here. Hydrogen, salt water or elevated temperature can move the life by an order of magnitude.

⚠ Screening only. A released proof-test argument uses measured toughness and growth data at the service condition, realistic flaw shapes, and the governing standard's factors — NASA-STD-5019 and the AIAA S-080/S-081 series set the requirements for flight pressure vessels.

ZONE 46

Residual stress

Stress the part carries with no load applied at all — and which decides how long it lasts.

IT IS NOT A LOAD, SO IT NEVER APPEARS IN A LOAD CASE Residual stress is locked in by manufacturing: machining, grinding, welding, forming, heat treatment. It does not show up in a static margin, it is rarely measured, and it routinely decides whether a part reaches its design life. Compressive residual buys fatigue life; tensile residual spends it — and welding leaves tensile residual approaching yield.

Surface condition

PROCESS

RESIDUAL STRESS MPa — 0

MATERIAL

APPLIED MEAN MPa

FLAW DEPTH mm

CREDIT COMPRESSIVE MEAN BELOW ZERO

What it costs or buys

—

Where it comes from

A COMPRESSIVE LAYER IS ONLY AS DEEP AS IT IS Peening protects the first few tenths of a millimetre. A crack that grows past the layer sees the full applied stress again, and the benefit ends abruptly. Peening a component whose critical flaw is deeper than the layer buys initiation life and nothing else.

IT RELAXES Residual stress is not permanent. Thermal exposure, overload and cyclic plasticity all shed it — peening benefit measured on a new coupon may not survive service. Anything relying on residual stress needs to show it is still there at end of life.

MACHINING IT AWAY REMOVES THE BALANCE, NOT JUST THE STRESS A residual field is self-equilibrating. Remove material and the remainder re-equilibrates — which is why thin machined parts distort after the last cut, and why rough-machine, stress-relieve, finish-machine exists as a sequence.

THE NUMBERS HERE ARE RANGES, NOT VALUES Residual stress depends on the specific process parameters, geometry and material condition. The ranges below are indicative. If a design depends on residual stress, it must be measured — X-ray diffraction, hole drilling or neutron diffraction — not assumed.

⚠ Screening only. Residual stress fields are three-dimensional, vary through thickness, and redistribute under load. This page treats a single surface value against a mean-stress criterion, which is a first approximation and no more.

ZONE 6

Fatigue

Why parts fail at loads they survived the first thousand times — and how the pieces scattered across this site fit together.

A STATIC MARGIN CANNOT SEE THIS COMING A part can pass every static check with room to spare and still fail in service, because fatigue is damage from repetition, not from magnitude. The load that breaks it is often a fraction of the one it survived on day one. Everything below exists because that first sentence is true and deeply unintuitive.

Eight ways of seeing it

In order — each one sets up the next.

From a laboratory number to your part

Published endurance data comes from a polished coupon. Your part is none of those things. Drag the sliders to see what survives the journey.

Knockdowns

LABORATORY ENDURANCE MPa

Surface finish k_a — 0.75

Roughness is a stress raiser and a crack starter. Ground is kind, as-forged is brutal.

Size k_b — 0.85

A bigger section holds more surface, more inclusions and more chances to start.

Load type k_c — 0.85

Axial loading has no stress gradient to arrest an early crack, so it is worse than bending.

Temperature k_d — 0.95

Above roughly a third of the melting point, creep and oxidation join in.

Reliability k_e — 0.75

Moving from a 50% mean life to a 99.9% survival allowable.

What reaches the part

—

The S-N curve, and why aluminium is different

MATERIAL

SERVICE STRESS AMPLITUDE MPa — 200

Mean stress — the same amplitude, very different life

MEAN STRESS MPa — 0

CRITERION

Scatter — a life is a distribution, not a number

WHERE THE REST OF THIS LIVES The material selector applies the Marin factors and shows the corrected endurance for every alloy. FEA cards generate an S-N curve for a solver. Residual stress shows the mean-stress shift peening buys. Surface finish gives the roughness that sets k_a. This zone is what connects them.

INITIATION AND PROPAGATION ARE DIFFERENT PROBLEMS Everything on this page is initiation — stress-life, for a part with no known flaw. Once a crack exists, stress-life stops applying and you switch to fracture mechanics: the proof testing zone integrates Paris growth from a known flaw. Using the wrong one is a common and expensive mistake.

THE CHEAPEST FATIGUE FIX IS USUALLY THE SURFACE Because cracks start at the surface, improving it is disproportionately effective. Polishing, peening or rolling a fatigue-critical feature often buys more life than a material change, at a fraction of the cost and with no requalification of the alloy.

THESE ARE ESTIMATES, AND FATIGUE PUNISHES OPTIMISM The Marin factors are correlations, the criteria are curve fits, and scatter is a factor of ten. A fatigue-critical part is qualified by test, not by calculation. Use this to size and to understand; use coupons and full-scale testing to substantiate.

⚠ Teaching and screening. Real fatigue substantiation uses material-specific S-N or ε-N data at the correct R-ratio, environment and temperature, a validated damage model, and test evidence at component level.

ZONE 34

Random vibration

A motor is carried long before it fires. What the ride does to it, and whether it survives qualification.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

THE RIDE IS OFTEN WORSE THAN THE FIRING A motor spends minutes firing and possibly hours strapped to a vehicle. Random vibration is broadband and statistical — there is no single load to check, only a spectrum and a duration. It routinely sizes brackets, attachments and internal hardware that the pressure case never troubles.

Environment & item

SPECTRUM

PSD g²/Hz

DURATION s

NATURAL FREQUENCY Hz

DAMPING ζ % — 5.0

ITEM MASS kg

SUPPORT AREA mm²

Response

—

Three-sigma damage — where the life actually goes

MATERIAL

S-N SLOPE b — 6.0

MILES ASSUMES ONE MODE AND A FLAT SPECTRUM It is a single-degree-of-freedom answer. Real hardware has many modes, spectra have breakpoints and notches, and the PSD at your resonance may be nothing like the plateau value. Use Miles to size and to sanity-check; use a modal or random-response FEA, and then a shaker, to substantiate.

STIFFER IS NOT AUTOMATICALLY BETTER Response goes as √f_n, so raising the natural frequency increases g-response even as it reduces deflection. The reason to push f_n up is to clear the input spectrum or to avoid coupling with the vehicle — not to reduce the number of g.

DAMPING IS THE STRONGEST LEVER YOU HAVE Response goes as √Q, and Q is the one term you can change by an order of magnitude. A bolted joint with some friction may reach ζ = 5%; a machined monolithic bracket might be 1%, which is 2.2× the response for the same input.

RANDOM IS NOT SINUSOIDAL There is no single peak. The response is Gaussian, so the item spends 4.3% of its time above 2σ and briefly exceeds 3σ. Designing to the RMS value alone understates the peaks by a factor of three, which is why qualification is written around 3σ and why the damage sum below is banded rather than single-valued.

⚠ Screening. Real qualification follows the governing standard — MIL-STD-810, NASA-STD-7001 or the programme's own tailored spectrum — with measured modal data and a shaker test. Miles' equation is a first estimate, not a substantiation.

ZONE 28

Transient thermal

How fast something heats or cools — and when treating it as one temperature stops being honest.

THE STRUCTURE IS OFTEN HOTTEST AFTER THE MOTOR STOPS During firing the liner takes the heat. When firing ends, the gas stops moving but the insulation keeps releasing what it stored — inward, because that is the only direction left. Soak-back regularly peaks minutes after burnout, and it is what cooks bond lines, seals and electronics that survived the burn itself.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

Part & environment

ENVIRONMENT

MATERIAL

CHARACTERISTIC LENGTH mm

FILM COEFFICIENT W/m²K

START TEMP °C

ENVIRONMENT TEMP °C

LIMIT TEMPERATURE °C — 200

Response

—

ONE NUMBER: THE BIOT NUMBER Bi = hL_c/k compares how fast heat arrives at the surface with how fast it spreads inside. Below 0.1 the interior keeps up and a single temperature is a fair description. Above it the part has a real gradient, and every figure on this page becomes a fiction — which is exactly when a quench cracks things.

τ IS A PROPERTY OF THE PART AND ITS SURROUNDINGS τ = ρc_pL_c/h. Thicker, denser or higher-capacity means slower; a more aggressive environment means faster. Conductivity does not appear — in the lumped regime the inside is already keeping up, so the bottleneck is entirely at the surface.

THREE TIME CONSTANTS IS 95% The approach is exponential, so it never quite arrives: 63% of the gap closes in one τ, 95% in three, 99.3% in five. Quoting a “time to reach” the environment temperature is meaningless — only time to reach a stated limit has an answer.

L_c IS THE WEAKEST INPUT Characteristic length is volume over surface area, which is unambiguous for a slab and a judgement call for anything real. A fin, a boss or a joint changes it locally. Get L_c wrong by two and you get τ wrong by two.

⚠ Screening. Real transient analysis uses a finite-element or finite-difference model with temperature-dependent properties, radiation where it matters, contact resistance at joints, and measured film coefficients. Lumped capacitance is a first estimate that tells you whether a detailed model is even needed.

ZONE 19

Silicone & polyurethane

Two elastomers with opposite chemistry, both everywhere in a motor — and a mutual incompatibility that ruins hardware quietly.

THEY POISON EACH OTHER, AND IT LOOKS LIKE BAD MATERIAL Polyurethane cures with amine and tin catalysts. Addition-cure silicone cures on platinum, which those catalysts poison at parts-per-million. The silicone simply never cures where they meet — and the usual conclusion is a bad batch of silicone rather than contamination. Meanwhile silicone's very low surface energy means it migrates everywhere and stops anything else bonding at all.

The two backbones, in 3D

Drag either model to rotate. Every property below follows from the shape of the chain.

Platinum poisoning, simulated

Drag the contamination level. Every green dot is a platinum site that can still catalyse the silicone cure; every amber dot is an amine or tin molecule migrating out of the urethane.

Contamination

RELATIVE CONTAMINATION — 15

RUN

WHY PARTS PER MILLION IS ENOUGH The platinum catalyst is itself present at parts per million. A poison only has to match the catalyst concentration, not the bulk — which is why a trace of urethane residue on a glove or a mould kills a whole batch.

Where they fight

Each with the actual mechanism, not just the warning.

ONE BOND EXPLAINS MOST OF SILICONE Si–O is 452 kJ/mol against carbon–carbon's 348, and the Si–O–Si angle opens to 143° rather than 109°. Stronger bond, floppier chain: high temperature capability and flexibility to −100 °C. The same open coiling is why gases walk straight through it.

ONE BOND EXPLAINS MOST OF POLYURETHANE The urethane linkage carries an N–H that hydrogen-bonds to its neighbour's carbonyl, pulling hard segments into stiff domains inside a soft matrix. That two-phase structure gives the abrasion and tear resistance — and the same bond is hydrolysable, which is the material's main weakness.

HOW TO KEEP THEM APART Separate storage and separate tooling. Cure polyurethane fully before silicone goes anywhere near it. Use a neutral-cure silicone, never acetoxy, next to metal or urethane. Where a joint between them is unavoidable, use a dedicated primer and expect mechanical interlock to do the work.

TEST THE COMBINATION, NOT THE MATERIALS Both may be perfectly qualified alone and still fail together. A small cure-compatibility coupon — silicone cast against the actual cured urethane, from the actual batch — costs an afternoon and catches platinum poisoning before it reaches hardware.

⚠ General elastomer chemistry for design awareness. Specific grades vary enormously — addition-cure and condensation-cure silicones behave differently, as do polyester and polyether urethanes. Confirm compatibility with the actual products and their datasheets, and by test.

ZONE 61

Design guides & drawing callouts

Open only what you need. Every guide ends with a copy-ready callout in a dashed box, exactly as it should appear on the drawing.

ZONE 40

GD&T — geometrical tolerancing

ISO 1101 / ASME Y14.5. Start with the frame anatomy, then each symbol: the shape of its tolerance zone, whether it needs a datum, and how the inspector will actually measure it.

Mental models

The parts of this that are most often got wrong, pictured before they are calculated.

Feature control frame builder

Assemble a frame and it checks the combination as you go — datum requirements, when a diameter symbol is needed, and which modifiers are legal for the characteristic you picked.

Build the frame

CHARACTERISTIC

TOLERANCE ZONE SHAPE

TOLERANCE VALUE mm

MATERIAL CONDITION MODIFIER

PRIMARY DATUM

PRIMARY MODIFIER

SECONDARY DATUM

SECONDARY MODIFIER

TERTIARY DATUM

TERTIARY MODIFIER

Frame

The 14 characteristics

ZONE 2

Requirements & interface control

Turn every “shall” into a numbered requirement with a verification method before any geometry exists.

A requirement you can't verify is an opinion. The first job on any new part is to convert the spec into a compliance matrix: one row per requirement, each with how it will be proven.

Verification methods

Method	Use for	Evidence
Test	Anything with a pass/fail measurable under load	Test report against a released procedure
Analysis	Strength, stiffness, thermal, where test is impractical	Stress report with margins
Inspection	Dimensions, markings, finish, material	Inspection record, certificate
Demonstration	Assembly, access, operability	Witnessed record

Words that make a requirement unverifiable

Wording	Problem	Rewrite as
“suitable”, “adequate”	No threshold	The number it must meet
“minimise”, “maximise”	No end point	A limit, plus a goal if wanted
“as required”, “where necessary”	Delegates the decision	The specific condition
“and/or”	Two requirements in one	Split into two rows
“should”	Non-binding in most specs	“shall”, or move to guidance

DESIGN RULE Track TBD and TBC items in the same matrix with an owner and a date. A TBD that reaches drawing release becomes whatever the designer assumed on the day.

Derived requirements

Most of what governs a detail part is derived, not stated: a seal requirement becomes a surface finish, a flatness and a squeeze range. Write the derived ones down and trace them to their parent. When the parent changes, you can then find every feature it drives.

Interface control

IN ONE LINE Both sides design to the interface control document — never to the other team's current CAD.

Interfaces fail at the boundary between teams, not at the boundary between parts. The ICD is the only thing both teams can be held to.

What an ICD has to define

Field	Content
Geometry	Mating features, datum reference frame shared by both sides
Tolerance split	How the allowable mismatch is divided between the two parts
Loads	Forces and moments across the interface, both directions, all cases
Thermal	Temperature each side, heat flow, allowable differential growth
Sealing	Which side carries the groove, which the face, finish and flatness
Fasteners	Who supplies them, size, grade, torque and lubricant
Bonding / earthing	Electrical continuity requirement and surface prep
Freeze date	When it stops moving, and the change route after that

TRAP Designing to the other team's live model. It feels efficient until they move a hole for their own reasons and nobody tells you. If it isn't in the ICD, it isn't agreed.

DESIGN RULE The mating datum schemes on both drawings must be the same features in the same order. If one side datums on the spigot and the other on the bolt circle, the tolerance stack across the interface is undefined.

⚠ Screening guidance from general practice. Ranges are typical, not allowables — confirm against the governing standard, your programme specification and company procedures before anything reaches a released drawing.

CHECK ALONGSIDE

ZONE 3

Duty cycle & environment

Count the real pressure and thermal cycles — single-use hardware still sees proof, leak and acceptance tests before it ever flies.

Most structural failures traced back in review started here, with a load case nobody wrote down. This zone exists to stop detail design opening before the environment is agreed.

What must be on the sheet before detail design opens

Input	Why it governs	Common miss
MEOP (max expected operating pressure)	Sets wall and joint sizing	Quoted nominal, not worst-case hot
Pressure rise rate	Dynamic load factor on joints and seals	Treated as static
Temperature: soak and transient	Knockdowns, seal state, differential growth	Only steady-state given
Duration at temperature	Short- vs long-time allowables, creep	Never asked for
Vibration and shock spectra	Preload retention, fatigue, fretting	Deferred to qualification
Handling and transport	Frequently exceeds operating loads	Ignored
Assembly cycles	Thread wear, inserts, seal reuse	Assumed one
Storage life and conditions	Elastomer ageing, corrosion, compression set	Treated as packaging

Count the cycles honestly

“Single use” describes the mission, not the hardware's life. A pressure part typically sees development pressure checks, a proof test, one or more leak checks, acceptance testing, and sometimes rework and re-proof — before its one operational cycle. Each is a full stress cycle on every joint and seal.

Event	Typical count	Note
Proof test	1–2	Highest stress the part will ever see
Leak checks	2–5	Each disassembly adds one
Acceptance / functional	1–3	Programme-dependent
Thermal cycles in storage	many	Diurnal swings on seals and preload

DESIGN RULE If an input is unknown, write the assumption on the sheet with a name against it. An assumption with an owner gets challenged; a blank gets filled in silently by whoever is drawing at the time.

Limit, ultimate, proof

Limit — highest load expected in service. No factor applied.
Ultimate — limit × the factor from your programme spec. Must not rupture.
Proof — acceptance test level on delivered hardware, from the contract or code.

Factors come from the customer specification or governing code — never from a handbook default.

CHECK ALONGSIDE

ZONE 4

Criticality & DFMEA

Classify criticality first — it sets the allowable basis, the inspection level and who signs the drawing.

Criticality is decided by the consequence of failure, not by how hard the part is to make. Assign it at concept stage; retrofitting it after release triggers re-analysis and re-inspection of everything.

What criticality drives

Class	Allowable basis	Inspection	Approval
Safety-critical / fracture-critical	A-basis, damage tolerance	100% NDT, FAI, serialised	Independent check + design authority
Mission-critical	A- or B-basis by load path	NDT on critical features, FAI	Design authority
Standard	B-basis or spec minimum	Sample inspection	Normal route

Your company's classes and their consequences will differ in detail — this is the shape, not the rule.

DFMEA, done usefully

List functions, not parts. “Contain pressure”, “maintain clamp”, “locate within 0.1”.
For each function, list how it fails and what happens next.
Score severity independently of likelihood — a rare catastrophic mode is still catastrophic.
For every high-severity line, name the design feature or test that controls it.

TRAP Single-point failures hidden by an assumed redundancy. Two seals in series with no way of testing the inner one is one seal plus a placebo. See Sealing strategy.

CHECK ALONGSIDE

ZONE 14

Allowables & statistical basis

An allowable is a statistical statement with a basis, a temperature and a direction — a bare number is untraceable.

Quoting a typical value into a margin calculation is the quiet way to lose a design review.

Basis	Means	Use when
A-basis	99% of population exceeds, 95% confidence	Single load path
B-basis	90% exceeds, 95% confidence	Redundant load path, where permitted
S-basis	Specification minimum, no statistical backing	Usually treated as A-basis — check your spec
Typical	Mean of test data	Never in a margin

Knockdowns before the allowable is yours

Grain direction — short-transverse can be 10–20% below longitudinal in thick wrought product, with much lower elongation.
Section thickness — quench rate falls with thickness; heavy sections don't reach the thin-section allowable.
Welds — joint efficiency from the code and procedure qualification, not the parent metal.
Temperature and time — short-time and long-time allowables are different curves.
Process — additive and casting allowables are tied to the qualified process and supplier.

DESIGN RULE Keep basis, temperature, direction and product form in the same cell as the number. “σ = 880 MPa” on its own will be queried at every review until someone finds where it came from.

CHECK ALONGSIDE

ZONE 20

Coatings & surface treatment

Every coating changes dimension, friction, fatigue or conductivity — state which you're accepting.

Treatment	For	Costs you
Sulphuric anodise	Corrosion, paint base on Al	Dimension growth, fatigue debit, insulates
Hard anodise	Wear on Al	Large fatigue debit, growth about half outward, brittle on edges
Chromic anodise	Corrosion with smaller fatigue debit	Hexavalent chromium — REACH restricted
Conversion coating	Corrosion, conductive, paint base	Thin protection; hexavalent being replaced by trivalent
Zn-Ni plating	Steel corrosion, cadmium replacement	Hydrogen embrittlement risk on high-strength steel
Electroless nickel	Uniform wear and corrosion layer	Fatigue debit; hardness varies with heat treatment
Passivation	Stainless steel corrosion	Nearly free — but must be specified
Dry film lubricant	Anti-galling on threads	Friction changes your torque-preload relationship
Shot peening	Fatigue life	Surface roughness; needs intensity and coverage control

TRAP Plating high-strength steel — roughly above 1000 MPa tensile — introduces hydrogen that causes delayed brittle fracture, sometimes days after assembly. A de-embrittlement bake within a short time window after plating is mandatory. Specify it on the drawing with its governing standard, such as ASTM B850.

Drawing notes this zone requires

Whether dimensions apply before or after treatment.
Masking for threads, sealing faces, bonding points and electrical contact areas.
Any change in torque for coated or lubricated fasteners.

CHECK ALONGSIDE

ZONE 21

Load paths & joint architecture

Trace each force from where it enters to where it leaves before any CAD; if you can't, the structure can't either.

Every structural problem in review is a load path problem in disguise. Sketch it by hand, per load case, with arrows.

Rules the load path obeys whether you like it or not

Load goes to stiffness. In parallel paths, the stiffer one carries more, regardless of which you intended to carry it.
Offsets create bending. Any eccentricity between where load arrives and where it's reacted becomes a moment. Moments make thin parts heavy.
Joints are discontinuities. Every joint is a change in stiffness and a stress raiser. Fewer joints in the primary path is almost always better.
Thermal loads are self-reacting. They never show up in the external free-body, which is exactly why they get missed.

Prying in tension flanges

When the bolt line sits outboard of the wall, the flange levers about its outer edge and the bolt sees more than its share of the applied load. The extra is the prying force, and on a flexible flange it can add tens of percent to the bolt load.

DESIGN RULE Keep the bolt circle as close to the wall as tool access allows, and make the flange stiff enough that it doesn't lever. Thickening the flange usually beats upsizing the bolts.

TRAP Lap joints carry secondary bending from the offset between the two sheets' mid-planes. Single-lap shear joints fail earlier than a pure shear calculation predicts for exactly this reason.

Joint architecture

IN ONE LINE Choose the joint type by what must come apart, seal and be inspected — mass comes after.

Joint choice locks in more mass, cost and risk than any detail decision that follows. Decide it explicitly against the requirements.

Joint	Separable	Seals well	Strengths	Watch
Bolted flange	Yes	Yes	Robust, inspectable, standard methods	Mass, prying, preload loss
Radial pin / bolt	Yes	Separate seal needed	Low mass on thin cylinders	Bearing, net section, hole out-of-round
Threaded closure	Yes	Separate seal needed	Compact, few parts	Galling, thread root stress, torque control
Retaining ring	Yes	Separate seal needed	Very compact	Groove capacity, ring dishing
Welded	No	Yes	Lightest, no leak path	HAZ knockdown, NDT access, distortion
Bonded	No	Sometimes	Spreads load, joins dissimilar materials	Peel, process control, inspection is hard

TRAP Welding a joint that must be opened for inspection, refurbishment or failure investigation. Ask who needs to get inside, and when, before choosing a permanent joint.

DESIGN RULE Separate the locating function from the clamping function. A spigot or dowels locate; bolts clamp. Bolts in clearance holes are not a location feature.

CHECK ALONGSIDE

ZONE 22

Envelope, access & assembly

Reserve space for tools, hands and inspection probes, not just for parts.

The envelope a part needs in service is smaller than the envelope it needs to be assembled, torqued, inspected and taken apart. Design to the second.

Check every fastener for

Socket clearance — the socket's outside diameter, not the bolt head, sets the minimum pitch from walls. Check against the actual tool you'll use.
Torque wrench path — the wrench length and the swing it needs. A fastener you can reach but not torque to spec is an uncontrolled preload.
Insertion path — the bolt has to get into the hole from somewhere. Count its full length.
Visibility — the assembler must be able to see the fastener seat and any locking feature.

Beyond fasteners

Assembly order: can every part get to its position with the others already in place?
Lifting points and centre of gravity for anything you can't lift by hand.
Inspection access for the CMM probe, borescope or NDT transducer.
Thermal clearance at the extremes — parts that fit at 20°C can touch at the operating limit.
Harness and pipe routes with their bend radii, not as thin lines.

DESIGN RULE Model the critical tools as solids in the assembly. It takes ten minutes and catches problems that otherwise turn up on the shop floor.

CHECK ALONGSIDE

ZONE 29

Hand-calc library

Hand calcs bound the answer before FEA — if the two disagree by more than about 20%, find out why before trusting either.

Pressure-loaded shells

thin cylinder (r/t > ~10): σ_hoop = p·r / t σ_axial = p·r / (2t) thin sphere: σ = p·r / (2t) thick cylinder, bore (Lamé): σ_θ = p·(r_o² + r_i²) / (r_o² − r_i²)

Flat plates — uniform pressure, radius a

clamped edge, max at edge: σ = 3·p·a² / (4·t²) simply supported, max at centre: σ = 3·(3+ν)·p·a² / (8·t²)

Real flange-to-plate joints lie between these two bounds. Use both.

Beams and columns

bending: σ = M·c / I cantilever tip load: δ = F·L³ / (3·E·I) Euler buckling: P_cr = π²·E·I / (K·L)²

Thin cylinders in axial compression

classical: σ_cr ≈ 0.605 · E · t / r (ν = 0.3)

TRAP Classical shell buckling is badly unconservative. Real cylinders buckle well below it because of imperfections. NASA SP-8007 knockdown factors are substantially below 1 and fall as r/t rises. Never use the classical value unfactored.

Fasteners and lugs

metric tensile stress area: A_s = (π/4) · (d − 0.9382·p)² bearing: σ_br = P / (d·t) shear-out: τ ≈ P / (2·e·t) e = edge distance to hole centre

Thermal

fully constrained: σ = E · α · ΔT

Steel fully constrained through 100 K reaches roughly 240 MPa — before any mechanical load.

Stress concentration

Net-section stress × K_t from Peterson's charts. K_t matters for fatigue and brittle materials; ductile static failure is usually governed by net section.

CHECK ALONGSIDE

ZONE 30

Differential expansion

Absolute growth rarely breaks anything. Differential growth between parts that are bolted, pinned, fitted or sealed together does.

DECIDE WHETHER EACH JOINT IS CONSTRAINED OR FREE — AND DRAW IT THAT WAY Joints meant to slide but bolted solid are a classic source of cracking at the outermost fastener. The calculator below gives the three numbers that matter: the mismatch, what it does to a fit, and what it does to bolt preload.

Pair and temperature

OUTER / CLAMPED PART

INNER PART / BOLT

ASSEMBLY TEMPERATURE °C

SERVICE TEMPERATURE °C

LENGTH / GRIP mm

FIT DIAMETER mm

CLEARANCE AT ASSEMBLY mm (− = interference)

BOLT STRESS AREA mm²

JOINT / BOLT STIFFNESS RATIO

Result

—

Expansion coefficients used

Material	α ×10⁻⁶/K	E GPa	Pairs badly with

differential at a joint: Δ = (α₁ − α₂) · L · ΔT fit change: Δc = D · (α_outer − α_inner) · ΔT fully constrained stress: σ = E · α · ΔT preload change: ΔF = −Δ / (δ_bolt + δ_joint), δ_bolt = L / (E_b · A_s), δ_joint = δ_bolt / ratio

TRANSIENTS BEAT STEADY STATE During heating, thin sections lead thick ones. A flange and its bolts at different temperatures for a few seconds can see a larger mismatch than either steady state. Check the transient case with the Transient thermal zone, not just the end points.

THE FIT HAS TO WORK AT BOTH ENDS An aluminium housing on a steel bearing gains clearance hot and loses it cold. Size the interference at the hot end, then check assembly stress at the cold end.

LONG BOLTS ARE FORGIVING For the same mismatch, a longer, more compliant bolt changes preload less. Raising the grip ratio is often cheaper than changing material.

ELASTOMERS GROW TEN TIMES MORE Seal compounds expand roughly ten times as much as the metal around them. Gland fill has to be checked at the hottest point, at minimum groove volume.

⚠ Linear α at a single representative value. α varies with temperature, strongly so below about −100°C — use the Cryogenic design zone for cold cases. Stiffness ratio is an input, not a calculation; VDI 2230 gives the joint compliance properly.

Where it bites

Bolt preload — short grip, dissimilar materials, large ΔT. Rising preload can yield the bolt; falling preload opens the joint.
Interference fits — an aluminium housing loses interference hot. Size at the hot end, check stress at the cold end.
Large flanges — radial differential forces bolts into shear and fretting. Use slots, a spigot or a deliberate sliding joint.
Transients — during heating, thin sections lead thick ones. Transient gradients can exceed the steady-state case.

DESIGN RULE Decide early whether each joint is constrained or free, and draw it that way. Joints meant to slide but bolted solid are a classic source of cracking at the outermost fastener.

CHECK ALONGSIDE

ZONE 31

Margin of safety policy

A margin is only meaningful if load, factor and allowable are a consistent set — yield with yield, ultimate with ultimate.

MS = allowable / (FoS × limit load) − 1 pass when MS ≥ 0

Errors behind most review findings

Mixed pairs — yield allowable against ultimate load. Run yield and ultimate as separate lines.
Wrong basis — typical or room-temperature values in a hot, single-load-path check.
Missing factors — see below.
Stacked conservatism — pessimistic load × allowable × model gives a heavy part and a margin nobody believes. Record where conservatism sits so it can be removed deliberately.

Additional factors to look for in your spec

Factor	Applies to
Fitting factor	Joints and fittings, where stress state is uncertain
Bearing factor	Bearing at fastener holes with clearance or vibration
Casting factor	Cast parts, linked to inspection level
Weld factor	Welded joints, linked to procedure and inspection
Model uncertainty factor	Unvalidated analysis methods

These multiply onto the basic factor. Missing a required one makes every margin downstream wrong.

Combined loads

Use interaction equations for combined loading — for example fastener tension with shear. The form varies by spec (linear, quadratic or mixed exponents). Use the one your spec mandates; NASA-STD-5020 covers fasteners for spaceflight.

Reporting

One row per failure mode per load case: load case, failure mode, allowable with basis and temperature, factors with source, MS, analysis reference.

DESIGN RULE A margin above roughly 0.5 on a mass-driven part is a finding, not a success. Either the part is heavy or the load case is wrong.

Factor values are not held here. Yield, ultimate, proof and special factors come from your programme specification or governing code.

CHECK ALONGSIDE

ZONE 38

Test correlation

Write the predictions and pass criteria before the test, and instrument where the model is least certain.

Before the test

Record predicted strains, deflections or frequencies at every instrument location.
Set correlation criteria now. Common targets: strains within about 10–15% at significant locations; primary mode frequencies within about 5% with MAC ≥ 0.9. Confirm against your verification spec.
Place gauges where the model is least certain, not just where it's easy.

Instrumentation

Rosettes where principal directions are unknown; uniaxial gauges only where they're certain.
Avoid steep gradients unless deliberately measuring a peak — a gauge averages over its length.
Temperature-compensate gauges and record temperature alongside.
Calibrated, traceable load cells and transducers.
Zero readings before load; check return to zero after.

After the test

If results fall outside criteria, find the physical reason — a stiffer joint, a different boundary condition, a material property. Update the model on that basis, then re-predict.

TRAP Tuning parameters until the model matches, with no physical justification. A model forced to match one test predicts the next one no better than before.

CHECK ALONGSIDE

ZONE 39

Storage, ageing & surveillance

Hardware that sits on a shelf for years is still changing — seals set, adhesives oxidise, preload relaxes, coatings degrade. Plan for it, and test for it honestly.

ACCELERATED AGEING ASSUMES ONE MECHANISM An Arrhenius extrapolation is only valid if the same degradation mechanism governs at the test temperature and in service. Push the test temperature too high and you age the part by a different chemistry — then extrapolate the wrong answer with great precision.

Arrhenius equivalence

ACTIVATION ENERGY kJ/mol

SERVICE TEMPERATURE °C

TEST TEMPERATURE °C

TEST DURATION days

Mean kinetic temperature

STORAGE TEMPERATURE LOG °C — one reading per line, equal intervals

Result

—

acceleration factor: AF = exp[ (Ea / R) · (1/T_service − 1/T_test) ] T in kelvin, R = 8.314 J/mol·K equivalent service: t_service = AF · t_test mean kinetic temp: T_MKT = (Ea/R) / −ln[ (1/n) · Σ exp(−Ea / (R·T_i)) ]

What ages, and what to watch

Item	Mechanism	Design response
Elastomer seals	Compression set, oxidation, loss of low-temperature resilience	Budget set into minimum squeeze; record cure date; surveillance samples
Adhesive and bonded joints	Moisture uptake, plasticisation, interface degradation	Humidity-aged allowables; protect the bond line edge
Bolted joints	Preload relaxation, embedment, creep of coatings and gaskets	Long grips, no soft layers in the clamp, retorque policy if allowed
Coatings and platings	Breakdown at edges and damage, galvanic attack	Coat the cathode too; inspect edges and fastener heads
Lubricants and greases	Drying, separation, migration onto seals	Compatible with the elastomer; specify quantity, not "lightly"

THE AVERAGE TEMPERATURE IS THE WRONG AVERAGE Ageing rate rises exponentially with temperature, so hot days count for far more than cold nights. A store that cycles between 10 and 40°C ages hardware faster than a constant 25°C store. Use mean kinetic temperature, which weights the hot readings properly.

THICK SECTIONS AGE FROM THE OUTSIDE IN At high test temperatures, oxygen is consumed near the surface faster than it can diffuse inward. The skin ages and the core doesn't — a pattern that never happens in service. Thick elastomer and polymer parts need modest test temperatures or a check for this effect.

Ea IS A PROPERTY OF THE MECHANISM, NOT THE MATERIAL NAME Typical values for polymer oxidation fall broadly in the 50–120 kJ/mol range, and the answer is very sensitive to it. Measure it from at least three test temperatures — the slope of ln(life) against 1/T — rather than assuming it.

SURVEILLANCE BEATS PREDICTION Hold back representative samples stored with the hardware, and test them on a schedule. A prediction tells you what should have happened; surveillance tells you what did.

BOND AND LINER AGEING HAS ITS OWN TOOL Case-bond and liner property drift with age is modelled in the Propellant grain zone.

⚠ Single-mechanism Arrhenius model. ISO 11346 covers lifetime estimation for rubbers properly. This zone covers seals, adhesives, coatings and joints — not energetic materials, whose ageing sits with your energetics specialists.

ZONE 41

Datums, position & form in practice

Datums should mirror how the part is located in its assembly — primary on the largest, most stable functional contact.

Degrees of freedom removed

Datum	Feature	Removes
Primary	Plane (3 points of contact)	1 translation, 2 rotations
Secondary	Plane (2 points) or axis	Further translation and rotation
Tertiary	Plane (1 point) or slot	Remaining clocking

Choosing them

Primary: the face that seats the part in its assembly — usually the largest mating face.
Secondary: the feature that locates it — spigot, bore or dowel hole.
Tertiary: the feature that clocks it — second dowel, slot or keyway.

Datums must be the same features, in the same order, as the mating part and the ICD.

Datum targets

For castings, forgings and large thin parts, use datum targets — specific points, lines or areas — so the part is fixtured identically every time, before its faces are machined.

TRAP Datuming on a face that's convenient to the machinist instead of functional in the assembly. The part passes inspection and then doesn't fit, because it was measured in a reference frame it never uses.

Position & MMC

IN ONE LINE Use MMC where the feature only has to assemble; use RFS where it locates or seals.

Position defines a tolerance zone — usually cylindrical — for a feature's axis, relative to datums. Applied at MMC, it gains bonus tolerance as the feature departs from its maximum material size.

Fastener formulas

floating fastener (clearance holes in both parts): T = H − F fixed fastener (one part tapped or pinned): T = (H − F) / 2 each part, equal split H = MMC hole size F = MMC fastener size T = position tolerance at MMC

Example — M8 bolt, ⌀9.0 clearance holes: floating gives ⌀1.0 at each hole; if one part is tapped, ⌀0.5 at each.

Projected tolerance zone

For tapped holes and press-fit pins, the position error that matters is where the mating fastener sits, above the surface — not inside the hole. A slight tilt inside the hole becomes a large offset at the mating part. Specify a projected zone (Ⓟ) with a height equal to the mating part's thickness.

TRAP Fixed-fastener patterns without a projected zone. Parts pass inspection individually and won't assemble.

Patterns

Composite position separates the pattern's location to the datums (loose) from the feature-to-feature spacing within the pattern (tight). Use it where the bolt pattern must match a mating part closely but can float as a group.

Flatness

IN ONE LINE For a sealing face, flatness comes from the squeeze budget — and the measurement method can move the number.

Flatness controls one surface against itself: two parallel planes that float to best fit. It constrains form only.

Choose it from the function

Application	Flatness	Ra	Driven by
Face-seal gland face	0.05	0.8	Minimum squeeze across the seal
Metal-to-metal bolted flange	0.1–0.2	3.2	Rocking, preload uniformity
Datum feature, precision	0.02	0.4	Repeatability of measurement
General machined face	0.3	3.2	Process capability

Typical starting points — derive the real value from the seal squeeze or joint stiffness it serves.

TRAP Free state versus restrained. Flatness is inspected unclamped by default. A thin flange that pulls flat under bolt load will be rejected even though it seals perfectly. Either add a free-state note with a realistic value, or define the restrained condition — don't just tighten the number.

Measurement method moves the result

The tolerance is defined by the minimum-zone condition. A least-squares fit, the default on many CMM programs, always reports flatness equal to or larger than minimum zone — so it can reject conforming parts, never the reverse. Agree the fitting method and the point density with inspection before release. A handful of points on an annular face proves very little.

Profile of a surface

IN ONE LINE Profile with datums controls size, form, orientation and location in one frame — use it on any surface that isn't a simple feature of size.

Profile of a surface — a three-dimensional zone around the true profile.
Profile of a line — the same, applied cross-section by cross-section.

Datum referencing changes what it controls

Applied	Controls
No datum	Form and size only — the zone floats
Orientation datum	Form, size and orientation
Full datum reference frame	Form, size, orientation and location

Zone placement

Default is bilateral and equal about the true profile. For unequal zones, ASME uses the U modifier; ISO uses UZ with an offset. These are not interchangeable notations — use the one your governing standard defines.

DESIGN RULE On complex surfaces — blends, aero shapes, freeform — profile replaces coordinate ± tolerances entirely. Coordinate tolerances on curved surfaces create square zones that don't match the function.

Runout & concentricity

IN ONE LINE Use runout for rotating features; avoid concentricity, which is hard to measure and was removed from ASME Y14.5-2018.

Control	What it checks	Use
Circular runout	Each cross-section as the part rotates	Wobble at individual sections
Total runout	The whole surface as the part rotates	Overall rotating fit, seal surfaces on shafts
Position (RFS) of an axis	Axis location to a datum axis	Coaxial features without rotation measurement
Concentricity	Median points of a feature	Removed in ASME 2018 — avoid

ISO retains coaxiality, which is effectively position of an axis. ASME removed concentricity and symmetry in the 2018 edition because they were rarely measured as defined.

Datum axis

For shafts supported in two bearings, establish the axis from both journals together: datum A-B. A single short datum feature gives a poorly defined axis.

TRAP Runout checked in V-blocks on a lobed datum feature. Out-of-round datums produce false readings. Use centres or a roundness machine for tight runout.

Runout includes form error — it's a composite control. A part can fail runout from roundness alone.

CHECK ALONGSIDE

ZONE 45

Measurement & gauging

Agree how each critical characteristic will be measured before release — an unmeasurable tolerance becomes a dispute at goods-in.

Measurement capability

Uncertainty — aim for ≤10% of tolerance; 25% is a common floor.
Decision rules — ISO 14253-1 reduces the acceptance zone by the measurement uncertainty. Tight tolerances lose a large share of themselves to the gauge.
Gauge R&R — <10% of tolerance acceptable, 10–30% marginal, >30% unacceptable.

Method by characteristic

Characteristic	Usual method
Position, profile	CMM; functional gauge for position at MMC
Flatness	CMM minimum-zone, or surface plate and indicator
Bore size	Bore gauge, air gauge
Roundness, runout	Roundness machine, or centres and indicator
Surface finish	Stylus profilometer — state cutoff
Threads	Go/no-go thread gauges

Temperature

Dimensions are defined at 20°C. An aluminium part 200 mm long measured at 25°C reads about 0.023 mm long — more than a tight tolerance. Soak parts in the measurement room, or compensate.

TRAP The CMM program defines the measurement. Two labs measuring the same feature with different fits, filters or point densities get different answers, and both are “right”. Specify the method for critical features.

CHECK ALONGSIDE

ZONE 50

Sealing strategy

Decide where the leak path is and how many barriers cross it — then make every barrier testable on its own.

Sealing strategy is decided at layout, not at detail. Once the joint architecture is frozen, you only get to pick the O-ring size.

Face or radial

Arrangement	Tolerates	Sensitive to
Face (axial) seal	Radial movement, diametral mismatch	Joint gapping, flatness, preload loss
Radial (bore) seal	Axial movement, joint gapping	Bore dilation, eccentricity, extrusion gap

If the joint is likely to open slightly under load, a radial seal keeps sealing while a face seal unloads. That single fact decides a lot of layouts.

Seal families by duty

Family	Temperature	Reuse	Use when
Elastomer O-ring	Compound-limited	Usually replace	Default — cheap, forgiving
O-ring + backup ring	Compound-limited	Replace	Higher pressure, larger gaps
Spring-energised PTFE	Wide	Sometimes	Dynamic, chemically aggressive, low friction
Metallic C- or E-ring	High	No	Beyond elastomer limits; needs fine finish and high load
Flat gasket	Material-limited	No	Avoid in structural joints — creep kills preload

Redundant seals

TRAP Two seals in series with nothing between them. You can't tell whether the primary is working, and the trapped volume between them can pressurise during a slow leak or a temperature rise. Put a test port between the seals, or don't claim redundancy.

DESIGN RULE Every claimed barrier needs its own leak check. A redundant seal you can't test separately only proves the pair works on the day — not that you still have a spare.

CHECK ALONGSIDE

ZONE 49

Locking & retention

Preload is the primary lock — locking features only stop a joint coming apart once it has already loosened.

Retaining rings

Capacity is usually limited by the groove, not the ring. The groove edge yields in the softer housing material, or the ring dishes if the abutting part has a chamfer or radius.

Keep the groove edge sharp and the abutting face square. A chamfer on the abutting part can cut capacity drastically.
Check groove edge yield in aluminium housings — it governs well before ring shear.
For high thrust, use a shoulder, a thrust collar or a threaded retainer instead.

Locking methods ranked against transverse vibration

Method	Effect	Notes
Adequate preload, long grip	Best	Prevents the slip that causes loosening
Lockwire / safety cable	Prevents loss	Allows some rotation; inspectable
Wedge-locking washers	Good	Check compatibility with surface coatings
Prevailing-torque nuts	Moderate	Reuse limits; check running torque on reuse
Anaerobic thread-locker	Good when cured	Temperature limit, cure on passive metals needs primer
Split spring washers	Negligible	Flatten at preload — don't rely on them

TRAP Thread-locking compound changes friction and therefore preload for a given torque. If you add it, the torque value must change too.

CHECK ALONGSIDE

ZONE 53

Process selection & additive

Choose the process from quantity, material and the tightest feature — then design everything else to that process's natural capability.

Process	Typical tolerance	Ra µm	Fits
CNC milling	±0.05, ±0.01 precision	0.8–3.2	Prismatic parts, low–medium volume
CNC turning	±0.025	0.4–3.2	Rotational parts
Grinding	±0.005	0.1–0.8	Bearing and seal surfaces
Wire EDM	±0.005	0.4–1.6	Hard materials, sharp internal profiles
Investment casting	≈±0.13 per 25 mm	1.6–3.2	Complex shapes, medium volume
Sand casting	±0.5–2	6.3–25	Large, low volume, needs machining
Closed-die forging	±0.5, machined after	—	Strength, grain flow, volume
Sheet metal	±0.1–0.25	mill finish	Thin structure, brackets
Metal additive (LPBF)	±0.1–0.2	6–15 as-built	Internal geometry, consolidation

Plate, forging or bar

Machining from plate is fast to obtain but cuts across grain flow. Forgings put the grain where the load is and carry better short-transverse properties, but lead times are long and tooling is costly. Your allowables must match the form actually procured.

DESIGN RULE Only the features that need precision should get it. One tight tolerance on a part doesn't justify tight tolerances on the rest.

Additive suitability

IN ONE LINE Use additive where it removes assemblies or enables internal geometry — not to make a machinable part more expensively.

Good and poor candidates

Good	Poor
Internal channels and passages	Simple prismatic parts
Consolidating several parts and joints into one	Large flat sealing faces as-built
Lattices, topology-optimised brackets	High-volume parts
Low-volume complex parts	Parts with no process qualification route

Laser powder bed fusion rules of thumb

Overhangs steeper than about 45° from the build plate self-support; shallower needs support.
Horizontal holes up to roughly 6–8 mm can self-support; larger need teardrop shapes or support.
Minimum wall roughly 0.3–0.5 mm, machine-dependent.
Internal channels need powder exit routes and inspection access.
Build orientation sets anisotropy — properties in the build direction usually differ.

Post-processing is not optional

Stress relieve before removal from the plate.
HIP to close internal porosity for fatigue-loaded parts.
Machine every sealing face, thread and datum.
CT or other volumetric inspection for internal features.

TRAP Allowables belong to the qualified machine, powder and parameter set — not to the alloy. Change the supplier or the parameters and the data no longer applies.

CHECK ALONGSIDE

ZONE 57

Leak testing

Every seal leaks at some rate. The requirement is a number, a method that can see it, and a test that isn't fooled by temperature.

“NO LEAKS” IS NOT A REQUIREMENT State the allowable leak rate, the test gas, the pressure, the method and its sensitivity. Without all five, a pass means only that the test couldn't see anything.

Pressure decay test

TEST VOLUME litres

TEST PRESSURE bar absolute

HOLD TIME s

MEASURED PRESSURE DROP Pa

GAS TEMPERATURE FALL K (negative = rise)

GAUGE RESOLUTION Pa

ALLOWABLE mbar·L/s

Result

—

Unit converter

VALUE

UNIT

Method by sensitivity

Method	Approximate sensitivity mbar·L/s	Good for	Watch
Bubble immersion / spray	≈10⁻³ – 10⁻⁴	Locating gross leaks	Operator dependent; contaminates
Pressure decay	≈10⁻³ – 10⁻⁴	Integral check of a whole assembly	Temperature, volume — see calculator
Helium sniffer	≈10⁻⁶ – 10⁻⁷	Locating leaks on pressurised parts	Background helium, sniffer distance
Helium accumulation (hood)	≈10⁻⁷ – 10⁻⁸	Integral rate of a pressurised part	Hood volume and soak time
Helium hard vacuum	≈10⁻⁹ and below	Highest sensitivity integral test	Wrong pressure direction for some seals

TEMPERATURE SWAMPS PRESSURE DECAY Gas pressure follows absolute temperature. At 11 bar absolute, a 0.1 K drift moves pressure by roughly 375 Pa — often more than the leak you are trying to see. Soak to equilibrium, log temperature, and correct for it or the test is measuring the weather.

TEST THE SEAL IN THE DIRECTION IT WORKS A vacuum test loads a seal the opposite way to service pressure. Some seals — lip seals, pressure-energised designs — leak backwards and seal forwards. Make sure the test pressure direction matches the service direction, or justify the difference.

TRACER GAS LEAK RATES DO NOT EQUAL SERVICE LEAK RATES Helium is small and fast. Converting a helium rate to a service fluid rate depends on the flow regime through the leak path. Specify the requirement as a helium rate at stated conditions, and keep the conversion argument separate.

EVERY CLAIMED BARRIER NEEDS ITS OWN TEST With two seals in series, a port between them is the only way to prove each is working. Without one, the second seal is untestable — see Sealing strategy.

⚠ Pressure decay uses Q = V·ΔP/Δt with an ideal-gas temperature correction. Sensitivities are typical orders of magnitude; the governing figures come from your test procedure and calibrated standard leaks.

ZONE 59

Certificates & traceability

A datasheet typical value is marketing; a certificate against a specification is evidence.

Material certificates — EN 10204

Type	Content	Validated by
2.1	Declaration of compliance, no results	Manufacturer
2.2	Non-specific test results	Manufacturer
3.1	Specific results for the delivered batch	Manufacturer's independent inspector
3.2	As 3.1, also verified	Purchaser's representative or third party

Aerospace structural parts normally need 3.1 at minimum. Check your spec.

Check every certificate for

Specification, issue and condition match the drawing.
Heat or cast number traceable to the material received.
Test specimen orientation and location relevant to your load direction.
Results from the delivered batch, not historic data.

TRAP Counterfeit and suspect parts, fasteners especially. Check head markings against the certificate and buy from approved sources. A certificate that doesn't match the markings is a non-conformance.

CHECK ALONGSIDE

ZONE 62

Drawing frame & note library

The title block declares the rules the whole drawing is read by — governing standard, projection, units and general tolerances.

Title block must state

Field	Why
Governing standard and issue	Sets how every symbol is interpreted
Projection symbol	First angle in UK and Europe, third angle in the US — never assume
Units	Millimetres unless stated; state it anyway
General tolerances	Apply to every untoleranced dimension
Default surface finish	Applies to every surface not otherwise marked
Material, specification and condition	Including heat treatment
Criticality classification	Drives inspection and approval
Export control marking	Legal requirement where applicable
Mass, scale, sheet n of m, revision	Housekeeping that gets audited

ISO and ASME read size differently

Under ISO 8015 the default is the independency principle: a size tolerance doesn't control form. A shaft can be in size and still be bent. Apply the envelope requirement Ⓔ where form must be inside size.

Under ASME Y14.5 the default is Rule #1: size controls form via the envelope. The same drawing can be read differently depending on which applies.

TRAP General tolerances from ISO 2768 on a functional feature. They suit non-critical geometry but don't express any function. Tolerance every functional feature explicitly.

Note library

IN ONE LINE Every note is a requirement — specific, verifiable and stated once.

Notes worth standardising

Fill the brackets from your specs. These show the shape of a good note, not your company's wording.

Interpret drawing per [governing standard and issue].
Remove burrs. Break sharp edges [range] unless otherwise specified.
Dimensions apply [before / after] surface treatment.
Mask threads and surfaces marked [symbol] from treatment.
Tighten to [torque] using [lubricant], friction coefficient assumed [range].
Bake for hydrogen embrittlement relief per [spec] within [time] of plating.
Mark part and serial number per [spec] at location shown, by [method]. No impression stamping on [class] parts.
Clean to [cleanliness level] per [spec] before assembly.
Inspect [features] by [method] to [acceptance standard and class].

Notes that don't work

Note	Problem
“Break all sharp edges”	No size — unenforceable
“Clean before assembly”	No level or method
“Finish to suit”	Delegates a design decision to the shop
“Torque to standard values”	Which standard, which lubricant?

DESIGN RULE State each requirement once. A torque value in a note and a different one on the view creates a non-conformance whichever is followed.

CHECK ALONGSIDE

ZONE 63

Revision control & sign-off

If the change keeps parts interchangeable it's a revision; if not, it's a new part number.

Revision or new part number

Change	Interchangeable?	Action
Clarify a note, fix a typo	Yes	Revision
Relax a tolerance, parts still fit	Yes	Revision
Change material grade	Usually not	New part number
Change a mating feature	No	New part number
Add a feature old parts lack	Usually not	New part number

Practice

Every revision references its change record and lists the zones affected.
Avoid revision letters that read like numbers — I, O, Q, S, X and Z are commonly excluded.
Decide whether the drawing or the model is the master, and say so on the drawing. Model-based definition is legitimate, but only if one source is authoritative.
Mark superseded copies clearly, or remove them from use.

TRAP Updating the model without re-releasing the drawing, or the reverse. The shop builds whichever they have.

Sign-off route

IN ONE LINE Each signatory checks something different — if two check the same thing, a gap exists somewhere else.

Role	Checks
Drawn	Complete and matches the model
Checked	Drawing practice, GD&T, every feature defined once and unambiguously
Stress	Margins exist and reference released analysis
Manufacturing	Producible with the intended process and tooling
Quality	Every requirement can be measured or verified
Design authority	Meets the requirements; approves release

Higher criticality adds independent checks. See Criticality & DFMEA.

Checker's short list

Is every feature fully defined, with nothing double-dimensioned?
Does each datum scheme match its mating part and ICD?
Is every functional feature explicitly toleranced?
Are material, condition, treatments and notes complete?
Can each characteristic be measured?

TRAP Checking in the live CAD model rather than the released drawing. The checker must sign what will actually be issued.

CHECK ALONGSIDE

ZONE 64

Archive & export control

Release in a neutral, frozen format carrying its export control marking — the native CAD file is not the record.

Archive formats

Content	Format	Why
Drawings	PDF/A (ISO 19005)	Self-contained, designed for long-term readability
3D models with PMI	STEP AP242 (ISO 10303-242)	Neutral, carries tolerances and annotation
Native CAD	Keep, but not as the record	Becomes unreadable as software versions move on

Retain records for the period your contract requires — often the life of the product plus a defined number of years. Store checksums so integrity can be shown.

Export control

Classify the technical data against the UK Strategic Export Control Lists — military and dual-use — before it leaves your organisation.
Mark drawings and file metadata with the classification.
Sending technology by email, cloud share or hand-carried drive counts as an export.
US-origin content can bring ITAR or EAR obligations on top of UK controls.
Exports need the right licence — SIEL, OIEL or an applicable OGEL. Your export compliance team decides; the designer marks.

TRAP Sending a drawing to a supplier or overseas colleague without checking its classification. The liability sits with the sender, not the recipient.

CHECK ALONGSIDE