TIGERFASTENERS
Practical references for the people who specify fasteners and the people who buy them.
From reading a designation to pricing a change, one question per page, with the standard and its range of validity stated.
A series in reading order, explaining why each decision sits where it does rather than only what to write. Start with the first; the rest are the steps it names.
How to read a designation, how to pick a type, and what the terms are in both languages. Everything else assumes these.
That torque on a 2⅛ inch nut develops about 500 newtons, under six per cent of an M6 proof load. The same nut needs around 400 foot-pounds to behave like a bolted joint.
An odd decimal in a specification is usually a round number in another unit. And here the fractional part is exactly the deviation a conforming wrench is permitted.
Bending stiffness goes as thickness cubed, so 35 micrometres of copper on a 1,6 mm board contributes 0,006 per cent of its stiffness. To match the board it would need 0,9 mm.
Two seatposts crushed at a torque below the stated maximum, and the top reply asked whether T25 was being read as 25 N·m. The two number series share their integers.
Fifteen degrees off perpendicular keeps 96,6 per cent of the clamping force and puts 25,9 per cent of the tension into sliding the joint. Cosine is flat near zero and sine is not.
ISO 965-1 reserves 26 micrometres of clearance at one millimetre pitch. A common layer height is 200. The standard is written in units the printer cannot address.
An M3 insert into a 4,0 mm hole displaces about 23 cubic millimetres. Sent up the bore that is 4,7 mm of climb, 83 per cent of the insert. The arithmetic lands on the fix the poster found by trial.
ISO 14579 gives each thread size one socket number and no choice. Inside a family the recess is not a free variable; across families nothing is fixed, which is what leaves the drive free to carry a message.
ISO 2320 gives a maximum going in and a minimum coming out. A new engineer added the maximum to his tightening torque and landed ten times above the supplier’s answer.
ISO 16130 says in its own scope that it cannot make an absolute statement about service loads. The test condition is found by destroying an unsecured joint first, and the requester may change it.
ISO 8734 and ISO 2338 specify the pin to m6 and never mention the hole. The words hole, bore, fit and reamer appear zero times in either preview, while the spring pin standard beside them names H12 outright.
12.9 bolts arrived where 8.8 was specified, and the drawing offered two torques for the same M12. Their ratio is the ISO 898-1 proof-load ratio, which means the joint was never in the calculation.
EN 13414-1 defines it as the maximum mass a sling is authorised to sustain, and states its coefficient of five in the scope. Chain and webbing carry different numbers, and a fastener standard gives no working load at all.
IEC 60664-1 gives the two distances separate definitions. A shoulder washer is solid insulating material in the path, so it lengthens the surface route and leaves the air route alone. And the material that insulates is the one that will not hold your preload.
A heatsink screw bottoms out on purpose. Load control moves from torque to displacement, and the scatter you inherit becomes a spring tolerance you can grade and buy instead of a friction condition nobody can see.
There is no torque to hit, only a torque plus a rule for what to do when the hole is not there. ISO 1234 names the split pin after the hole rather than the pin, and rates each leg to be bent back exactly once.
ISO 898-5 gives a hardness class, and a socket set screw can only be 45H. Its one strength test is a proof torque measured in a block harder than 50 HRC. The point is four separate standards, and it is the whole function.
ISO 2306 tabulates minor diameter limits for each class and names a drill last. Its introduction says a larger hole inside the band taps more cheaply and breaks fewer taps, and the percentage everyone quotes has a ceiling of 83%.
ISO 8752 publishes a minimum double shear load for every size: 17,54 kN at 5 mm. ISO 8734 publishes none for dowel pins at all, so the figure people quote for a dowel is supplier data rather than a standard value.
The word storage appears once in ISO 6789, as one input to choosing a calibration interval. The interval itself is 12 months or 5 000 cycles, whichever comes first, counted from first use.
ISO 898-2 says in its scope that it does not specify weldability, and it deleted class 9 in 2022. ISO 21670 gives a proof load, a carbon-equivalent ceiling and a manufacturer mark instead, and removes the projections before testing.
ASME PCC-1 sorts bolting by whether it has an integral head and by whether the far side is a nut or a tapped hole. Two questions, four joints. The advantages usually credited to studs belong to the second question.
Bulk fasteners are counted by dividing net weight by an average piece weight. ISO 3269 says nothing about how short a box may be — and an AQL is not a shortage allowance.
ISO 898-1 measures its properties between 10 and 35 degrees. It does carry a −20°C impact requirement, but only under three conditions most fasteners fail — and for 12.9 the value is still under investigation.
A self-drilling screw can be rated to clamp 28 mm and to drill only 3 mm of it. The point number is not a universal scale, and the drill rating counts the gaps between layers.
The grade mark is a conformity claim carrying an accountable name, not third-party verification. Most head styles are not required to carry one, and hardness alone cannot settle a dispute.
Notched 6061-T6 showed a fatigue limit where smooth specimens showed none, and 42CrMo4 below 1400 MPa produced no failures from a million to a billion cycles. The standard behind the number excludes notched components.
The outer bolts carry more in an end-loaded joint, and the two codes handle it differently: EN reduces continuously from 15 diameters, AISC steps at 38 inches.
In a flexible bracket the bolt carries the applied load plus a prying reaction — measured equal to the load itself. Tightening harder does not remove it.
A load cell clamped in a bolted joint is a spring in parallel with the bolts, so it responds from the first pound. The error is the preload minus the load factor times the applied load.
In a lap joint, life rises with torque until the failure changes to fretting fatigue and starts falling. In a tension joint, preload never reduced the stress amplitude to begin with.
The usual advice is to measure the pitch. For #10-32 and M5×0.8 it differs by 0.006 mm per thread — so the rule fails on the pair most likely to be confused.
The +158% fatigue figure was measured at 1% preload. On coarse threads at the preloads a real joint uses, the measured benefits were +8%, −9% and 0%.
Zinc flake bolts tightened within 30 minutes of cold storage needed 11% to 30% less torque for the same clamp force. The variable is condensation, not shelf life.
Rotation was made impossible and the joint still lost up to 40% in 200 cycles. Which means the marker-pen check — the one we recommend — is blind to it.
Thin nut first, against the joint — ISO 898-2:2012 said so in words. The 2022 edition kept the pairing rule and deleted the procedure. That says more than either side of the argument.
The current standard calls them D, T and BT — one naming system with old names and new, not two in parallel. And type 17 is not in that standard at all: it is in an Australian one, defined by what it must achieve rather than by the point shape.
Grade A is a tolerance class covering dimensions and geometry. It says nothing about strength, material, coating or surface defects — and the three grades do not simply widen in order.
Two separate demands arrive at a fastener supplier, and they want different things. One needs third-party verified product-level figures; the other needs your activity data and emission factors.
Fasteners are about 15% of what they cost you. The other 85% sits in engineering, purchasing, testing, inventory, assembly and logistics — and none of it appears on the quotation.
They answer two different questions, so there are four possible arrangements rather than two. And whichever you pick, it succeeds or fails on the same thing: whether your consumption data is trustworthy.
Countersink, counterbore and spotface have symbols in ASME and a different syntax in ISO. The V you have seen quoted as the ISO countersink symbol means something else entirely.
A bolt is 8.8; the nut is 8. That single number is not the nut’s own strength — it is the highest bolt class it may be paired with, and the pairing has a stated failure mode.
Button head screws are marked 08.8, not 8.8. The leading zero is the standard telling you the head carries 80% of what the thread could.
Fastener acceptance inspection takes five pieces from a lot of half a million, and the standard admits those plans give no statistical information. Here is what they rest on instead.
The dissimilar-metal compatibility table everyone cites is attributed to MIL-STD-889. It is not in MIL-STD-889. Here is where it actually comes from.
Turning a fixed angle after snug is a different measurement from torque. VDI has the equation, and the angle is not bolt stretch.
A steel screw in an aluminium housing gains preload on heating and loses it on cooling. VDI has a formula, and it includes a term most people forget.
VDI 2230 tabulates it. A torque wrench calibrated on the joint gives ±17–23%; by operator feel it is ±43–60%. That number decides how big the bolt has to be.
Below separation the bolt sees only nΦ of an external load, and that fraction does not depend on how hard you tightened.
Three popular reasons the slot survives: one is false, two are unverifiable. And a human study found no significant torque difference against Phillips.
Run-out is unavoidable; an undercut is a feature you add. DIN warns it may miss the proof load for its property class, and gives no reduction figure.
The famous bicycle pedal explanation predicts the wrong hand. And ISO 898-1 requires a left-pointing arrow on the head, which most drawings never mention.
Unrecorded anti-seize does. One maker publishes K = 0.13 against 0.185 dry and a 30% torque cut. Another says use normal torque. Both are right.
The skin measures 30 to 150 micrometres on a 2 mm wall and is absent altogether on 10 mm. Porosity is often higher in the skin than the core, not lower.
One forum thread produced eight different rules for minimum thread engagement, all stated with confidence: 1.5 diameters, 1.25, six threads, six to eight, one diameter, two. Two of those answers are right, and they are not multipliers at all. Why the standard formula appears circular, and what to put in its place.
Published models span 17.7% to 41.9%. The famous 34/23/16 sequence is one photoelastic model, not a constant, and it stops applying the moment a thread is damaged.
ISO 10664 designates it hexalobular socket no. 20. T20 is trademark language, and the standard covers the recess only — not the driver.
The cone apex is virtual — there is nothing to put an instrument on. Every method a shop uses measures a length instead, which is why drawings call out a diameter.
Threadlocker is rated for locking, tape for sealing, putty for machinability. None is a load rating. One product class publishes one — that is the answer.
ISO 228-1 says to seal by compressing two tightening surfaces outside the threads. ISO 7-1, the one that does seal on the thread, still asks for a jointing medium.
ISO 5145 builds forty two non-interchangeable valve outlets from one Whitworth thread. The keying is two recess diameters whose sum never changes.
ISO 965 tolerances three diameters, not pitch. Assembly is checked another way: the virtual pitch diameter, which is what the GO gauge reads.
ISO 273 makes an M8 clearance hole 9 mm. Grinding the screw does not close that gap. Location is a tolerance you buy separately, on the hole as much as the part.
Type A, AB, B, C, F are not grades. They are two independent choices at once: spaced or machine-pitch thread, and forming, cutting or rolling.
Controlled geometry with a minimum and a maximum. Whether your clearance hole clears it depends on the series — for M6, two of the three can overlap.
Forming displaces metal instead of cutting it, and the hole gets about 60% of the tolerance width a tapping screw is allowed.
DIN 7975 sets the floor at one pitch of sheet thickness, and below it the screw turns without driving. The familiar two-thread minimum is in no standard.
Drop friction from 0.15 to 0.08 and the same torque delivers 1.70 times the preload. That is why wheel bolt specs say dry, and some heavy-duty specs say oil them.
Tri-Wing and tri-point both look like a Y and are not interchangeable. One has an aerospace standard behind it; the other has none at all.
The letters are a required manufacturer mark. But an unmarked head proves little — countersunk heads and anything under 5 mm are normally not marked at all.
Eurocode 5 does not let you tighten spacing when you stagger; it lets you stop discounting the row. Steel codes credit it too. The hinge standard says otherwise.
At 8 mm, H7 is +15/0. Put g6, k6 and p6 against it: 5-29 um clearance, a coin toss, and up to 24 um interference. The hole never moved.
M8 comes in 1.25, 1.0 and 0.75. The fine one has 7% more tensile stress area, and on the nut side the pitch term cancels out of the basic stripping geometry. So pitch is close to a non-decision for strength — which means it is being decided by something else.
The usual explanation has it backwards. On 316L a thicker pre-formed oxide reduced galling damage: the layer must wear away first. And carbon steel galls too.
A two-start M8 still self-locks at ordinary friction, so that is not why. The cost is a thread needing a second number, which jams in an identical-looking part.
It resists rotation. Against preload lost without rotation it does almost nothing — and that is what small screws mostly suffer. Zinc plating is also one of the substrates it struggles to cure on.
Tightening the second screw loosens the first. One mechanism — elastic interaction — explains the cross pattern, the stepped passes, the final lap round, and the re-torque.
Three questions, not one. And for a thread-forming screw the awkward one is that it may not find the thread it made — forming a second one across it, with nothing visible from outside.
Origin is a determination made by the destination's customs authority under its own rules, and WTO harmonisation is incomplete — so country of melt and cast, country of origin and the Made in mark are three different things.
The head pushes preload into the joint through a finite area. Exceed what the material under the bearing face can take and it may creep — preload drains away and the joint loosens while every part still passes inspection.
Rejection is a fork, not an ending. Rework makes parts conform; repair only makes them usable — so accepting a repaired lot means accepting parts that still do not meet your drawing.
Two similar drawings, MOQs ten times apart. It is not a sales position — the minimum is set by whichever station your part is unusual at, and knowing which one tells you what to change.
Mixing coatings on one joint is three questions with three different answers, which is why the advice you get contradicts itself. ISO 11408 does not list corrosion protection among black oxide's purposes at all.
Nobody prints a number because the corrosion rate belongs to the atmosphere, not the screw — ISO 9223 puts the spread at 250 to one. Here is the division you can do instead, and the four reasons it is an estimate.
ISO 9717 says a phosphate layer without post-treatment provides no corrosion protection at all. So the first question is not which finish — it is what was done to it afterwards, and what the name left out.
A sample proves a part. What you need proved is a process. So check what will change between the sample and the run — not the things that obviously already work.
ISO 2702 says these screws are not intended to be pretensioned by design, so torque is derived backwards from failure rather than forwards from preload — but that scope is narrower than “no nut”.
Acceptance means the evidence did not reach the rejection criterion — weaker than “the lot is good”. An AQL is the worst tolerable quality level for a continuing series, not a permitted defect rate for your batch.
Five items decide whether the quotes you get back are comparable. Leaving one out does not leave it blank — it makes it an assumption, and the cheapest quote is usually the largest one.
Substrate, thread, head, drive, finish, documentation — an order, not a checklist. Each decision removes options from the next, which is why working out of sequence produces rework rather than an answer you can adjust.
It is two callouts from two standards joined into one string — m governs sizes and angles, K governs geometry — and neither gives a number until you know the size band. Angular tolerances are banded by the length of the shorter leg, so identical angles on one drawing can carry different limits.
Size charts answer nominal diameter and pitch. What produces parts that measure right and still will not fit is the rest — tolerance class, the length datum, and whether the finish was given any room. With coarse pitches for this range.
What separates the types is where the female thread comes from — already there, cut, formed, or drilled on the way in. That decides swarf, torque window and whether the joint survives being taken apart.
A thread is a wedge and the boss carries it as hoop tension — which is why plastics screws use a narrower thread angle. Plus the joint that is tight on day one and loose in six months with nothing broken.
Most of a glossary is uncontroversial and three entries are not — “self-tapping” covers more ground in English than 自攻 does, US “flat head” means countersunk, and asking for a mould instead of a die sends the enquiry to the wrong department.
ASME B18.2.1 separates them by how they are intended to be tightened, then admits the same part can be used either way. What actually changes is whether the female thread is a certified nut or a hole in your product.
Holes, engagement, tightening, tolerance — the numbers that decide whether the joint holds.
Read from the 1864 paper that proposed it. Three objections to the British thread, none about performance, and the reason for 60 degrees is that it was more readily obtained than 55.
ISO 4753 defines thirteen end types with symbols. Eleven sit inside the nominal length and two are added to it, so a pilot-pointed M8 is four millimetres longer than ordered.
ISO 4014 subtracts twice: grip is the bolt minus the thread length, and the guaranteed shank is that minus five pitches. On an M12 by 50 the shank is 11,25 mm.
A wheel lock kept shearing below spec. The 127-point answer was that they are weak. The right answer had six votes: the nut was bottoming out, and the reading climbs the same either way.
Quench cracks are banned at any depth, length or location. Forging cracks are allowed up to one thread diameter. ISO 6157-1 sorts by origin, so a ruler cannot answer the question.
ISO 898-1 defines classes 8.8 and 9.8 with a minimum tempering temperature of 425 °C and around 90 per cent martensite in the core, and gives no route back after a later thermal cycle.
An M10 nut takes 16 mm, and the ratio to the thread reverses between M5 and M6. Across corners is a separate minimum, and it is 1,13 times the flats rather than 1,1547.
About 70 per cent of the root diameter in softwoods and 90 in hardwoods, and the handbook's own gauge table gives the shank. Two errors from one sentence, failing in opposite directions.
The US Standard Gauge is a customs statute. Its thickness columns are headed approximate, its weight columns are not, and one 1893 density runs through every row.
ISO 6789-1 runs the range from the lowest marked value, makes the dial label the dead part below it, and lets one graduation be 5 per cent of maximum.
ISO 15977 gives two grades, rated shear and tensile loads, a mandrel break load that is a ceiling, and a hole tolerance of a tenth of a millimetre. The class letter is the last field of the designation.
Pitch does not scale with diameter — ×5 on diameter from M1 to M5, only ×3.2 on pitch. So the same 1×D engagement rule gives 4.0 turns at M1 against 6.25 at M5. Coarse pitch, turns and tapping holes for every size, plus why first-choice sizes quote better.
Too large and it strips before torque is reached; too small and driving torque climbs toward the breaking point. What to ask a supplier for is not a diameter — it is driving and stripping torque in your material.
The 1×D rule comes from steel into steel. Pitch does not scale with diameter, so the same 1×D is about 4.7 turns at M1.4 against 6 at M3 — same rule, a fifth less margin.
In T = K·D·F the coefficient belongs to the whole contact system, not to the screw. Change the coating, the lubricant or the clamped material and the same torque delivers something else — and below M3 there are no reference conditions yet.
Every part passes and it still will not fit. What worst case and RSS each assume — including why 0.27% is a model output, not a defect rate — and when changing the screw is the cheap fix.
Two current references give 6 mm and 6.72 mm for the same M3 head. Neither is a typo — one publishes nominal, the other theoretical, and the gap is a millimetre wide where it gets cut.
The angle is a pairing, not a screw specification. Mismatch it and the seating cone becomes one circle of contact — while screw and hole each measure conforming. Also why cabinet work pushed the angle out to 100°, 110° and 120°.
Upset ratio guidance runs around 2.25–2.3 in one blow — guidance, not a standard. Microforming reports the achievable value falling to about 1.6. Less margin and more scatter as parts shrink.
A rolled thread is displaced, not removed — the crest is pushed out of the material. That explains the fatigue difference, and a die-wear failure that leaves the parts looking perfect while the pitch diameter drifts.
Nothing is pressed on — the washer is slid onto a plain shank and thread rolling traps it there. The value is one less pick-and-place on the line, and one less way to lose a part. Plus where it is the wrong choice.
Small, normal and large series differ by more than double in outer diameter. And a washer is mostly there to make the joint repeatable, not stronger — so adding one moves the friction interface and changes the clamp force.
The first repair usually destroys the evidence needed to establish what actually failed.
Three different failures share one word and sit with different parties — the screw thread, the hole thread, and a rounded recess that is not a thread failure at all. Three bench checks separate them.
In one the screw rotated; in the other it never moved and the clamp force went anyway. Locking devices address only the first, and split lock washers barely do that. A marker pen tells you which you have.
Neither patent behind the cross recess contains the word cam. Believe the myth and you file a rounded head as normal wear instead of a controllable process fault, usually bit wear or a Phillips bit in a JIS recess.
US 1,908,080 of 1933 names John P. Thompson as the inventor and Phillips as the assignee, and argues about casting, broaching and cost. Read from the scans, with a correction to our own page.
ISO 8764-1 gives the width across the wings a band of 0,05 mm at every one of the five sizes, and defines conformance by an inspection gauge that checks how deep the tip sits.
The socket is above nominal and the key at or below it, so the play is built in. Compare the two across-corners columns and the socket is wider at every size.
The RCSC Specification defines snug tight by human effort with a named hand tool, and records that it tried a tighter definition and had to change back.
ISO 3266 defines the figure on the axis and scopes the product for inclined loading too. Its hazard list has one hazard and four causes, two of which are information.
Every safety-belt anchorage gets a 7/16 inch UNF hole, against 143 uses of mm. The next paragraph says when the requirement does not apply.
Three sentences of European machinery law, and the third one asks the design to make a missing fixing impossible to hide.
Two separate locking devices are required where the loss of the fastener would matter. The whole rule is 122 words and has not been amended since 1970.
The three range boundaries are 5, 0,2 and 0,001 kilograms force written out in newtons, and a Vickers number is kilograms force per square millimetre.
The lower limit is the resolution times 400, 200, 100 or 67 depending on the class. The finer the class, the more of the bottom the machine gives up.
The equally disposed angle is an implicit exact dimension. Whether the holes are one pattern has to be written down, and the previous edition of the same standard implied it instead.
Three numbered subclauses ask you to assume the functional limits are known with no uncertainty, that the tolerance limits match them, and that the part works fully inside and not at all outside.
Hole, reamer and fit appear zero times in the pin standard. The reamers come from a different committee, cover a smaller range, and one of them cites itself.
One part of the marine machinery rules gives the same feature on the pipe end opposite jobs, four paragraphs and one section apart.
Two counts are printed for every rope diameter, and a clip that was not drop forged needs one more of itself in seven of the nine rows.
Series 2 exists for keys that do not transmit the torque, because the interference fit does and the key only fixes the position.
An insulation pin is capped rather than floored, and two shear connector grades get a tensile range closed at both ends.
Not a working load and not a rating. A conforming Type 1 tie need only retain half of the value its maker declared.
Repair, reuse, enlarged and redriving appear zero times in thirty pages. The one document with the concept says replace the member, and then bans a fastener from the repair.
Divide one by the other and only the wood is left: the ratio is two over the square root of the specific gravity, widest in the lightest timber.
By our count the phrase appears zero times in 646 pages. What it uses instead is traceability, approval for the specific installation, and code markings on the head.
The type has no default, it propagates to the washer and the tension indicator, and the specification names who has to state it.
It gives a diameter, a minimum end distance and a position tolerance by product grade, and a note that the position must be calculated for each application.
It fixes the washer so it cannot rotate, then specifies hardness, roughness, flatness and a hole with no chamfer. Which end of the bolt is not the variable.
Across fifty two thousand characters of track regulation the word torque never appears, and tighten appears twice, both times about not clamping.
Seventy percent of a specified minimum, fifty where the plate bears against wood, an upper bound as well, and a torque figure that is a proof rather than a target.
The standard says indications made under former editions can only be read under the standards in force when they were issued, and gives an annex for doing it.
It names things ISO has no standard for, includes obsolete types on purpose, and its three official languages disagree about a cross reference.
The default delivery condition has been clean and lightly oiled since 1986, and the same document contains no friction, no lubricant and no coefficient.
The arbitration torque is made on the day, measured in the tightening direction, and the specification says it may reject good bolts and accept bad ones.
Its normal band ends at exactly half a diameter and one and a half for every preferred coarse size but one, and misses on the five that share a pitch with a preferred one.
The normal series hands the two obvious uses back to the user, the large series claims them, and the extra large washer is 100 HV, product grade C and rated only to property class 6.8 in timber.
A metric thread is selected by a diameter and a pitch and nothing else. The whole coarse series holds its lead angle between 2,48 and 3,60 degrees, and a second start doubles it out of that band.
Engineering stress, tensile strength taken at maximum force rather than at fracture, and a yield that is usually a line drawn at two parts in a thousand of permanent stretch.
The performance standard measures scatter rather than torque. One test joint reaches full torque in a twelfth of a turn, the other takes two turns, and the same setting has to cover both.
In January 1951 it permitted a like number of bolts of the same diameter in place of hot-driven rivets. The torque table it came with was withdrawn three years later.
Two forum arguments about a torque figure on a lug. The federal electrical standard has no figure, no table and not the word, and the check on an old connection is a thermometer.
The wall between an outer and an inner thread is set by two roots, not two nominal sizes. Only the outer pitch moves it, and with coarse pitch outside it goes negative at M12.
A steel erection standard counts the fasteners that must be in place before the hoisting line is released. It never says torque, and the section on changing an anchor rod does not cover metal buildings.
The regulation never approves a method. It asks whether the repaired condition is at least equal to the original, in four named qualities, one of which is resistance to vibration.
Safetying prevents disengagement rather than holding torque. On emergency mechanisms the strong wires are forbidden and brass or soft copper shear wire is used instead.
One free document answers the reuse question with a table of minimum prevailing torques by thread size. And it says one thread must show past the nut, three times in three chapters.
A centre distance is not a size. ISO 14405-2 says a plus and minus on it is ambiguous on a real workpiece, and does not recommend it.
A gypsum board fastening table allows screws at 1,5 to 2 times the nail spacing. The one row where they are equal is a garage ceiling under a habitable room.
Slip resistance is proportional to pretension, so with none there is none. The joint reverts to the bolt shank bearing on the hole, a few millimetres away.
ISO 28741 fixes the thickness for an unused plug, tightened once, at the specified torque, on threads that are clean, smooth and dry. Four conditions in one sentence.
One publicly released bonding standard names four things that shall not be used. Star washers are the third. The bond lives in the faying surfaces, not the hardware.
Bolt length comes off a fixed ladder of 54 values and the threaded part is arithmetic. In the first edition of ISO 888, 88 of the 89 cells obey the formula exactly.
ISO 2768-2 is withdrawn and ISO 22081 replaced it. The new general specification is one surface profile tolerance to a datum system, and Rule D excepts screw threads.
The 2005 table of ISO 10664 has sixteen sizes and starts at 6. The 2014 table has twenty four, and the sixteen that were already there did not move.
The 1918 act had a six-month sunset clause. The 1921 report has a wrench fit class with no numbers in it, tolerances sized to the taps already for sale, and a delegation that agreed nothing.
Whitworth opens his 1841 paper by saying no rule for screw threads can be deduced from mechanical principles, then describes collecting bolts from workshops across England and averaging them.
Eight months of Franklin Institute minutes, to the committee report that fixed the flat crest, the 60 degree angle and the across-flats rule. Its opening sentences closely parallel Whitworth of 1841.
ISO 15330 rules itself out of goods inward in its own scope, and the reason is a 24 hour clock. It also says baking cannot make a susceptible steel unsusceptible.
Whitworth defined a good fit in 1857 and put an allowance inside it, chosen from how the customer treats the machine. Three gauges, a ten-thousandth of an inch, and a drop of oil that reverses which is tight.
ISO 4527 writes the whole finishing sequence as one string, with a double slash for anything omitted, and defines the thickness as applying anywhere a 20 mm ball can touch.
The 1928 revision fills in the wrench fit class published empty in 1921: an appendix, still tentative, four permitted re-uses, and a lute of 40 per cent zinc dust.
ISO 2064 is two pages and defines the words every coating standard borrows. Local thickness is a mean inside a square centimetre, and the definition was chosen so different instruments would agree.
The wrong material does not fail on the line. It fails during the warranty period.
ISO 16048 puts the film at about 0,002 µm and a 2019 measurement on 316L agrees at ~2 nm. The mechanism is dissolving iron, and the standard says there is no referee test.
ISO 10684 forbids retapping and puts a letter after the property class. Z or X on the nut with U on the bolt has, in the standard’s words, a high probability of thread stripping.
ISO 17668 lists seven national names for one process. Zinc dust, a slowly turning drum, and a temperature under 419 degrees, giving a coating that is iron-zinc alloy all the way through.
ISO 12683 defines it as compacting zinc powder in a rotating barrel with impact media, without electric current or applied heat, and then admits there is no satisfactory adhesion test.
ISO 11408 has no shade, no gloss value and no grade. It asks the buyer to supply appearance as a marked sample, and bounds only what must be absent.
Stainless is not rust-proof — it self-repairs, and that needs oxygen. Surface contamination, pitting and crevice corrosion sit with different parties, and three bench checks tell them apart.
An insert is not reinforcement — it moves the load off a small area of thread onto a large interface, and brings pull-out and spin-out, which a tapped hole does not have and a strip torque test does not find.
The European assessment document requires every basic test in the actual brick, at the worst setting position, and then takes the mean down by 3,40 standard deviations.
A recess is a torque interface, and its depth is what is left after head height, head strength and the forming process have taken their share. Which is why the decision order puts the drive after the head.
A threaded joint has one property the others lack — it comes apart and goes back together. The hole, the preload that leaks, the controlled tightening: all of it is the price of that. If the joint never opens, nothing was bought.
A screw is a spring and grip length is the spring. Embedding is an absolute distance, so it costs a short screw far more of its preload — and in a blind hole, too long means torque with no clamp force.
The two digits of 8.8 are an equation: 800 MPa nominal tensile and a yield ratio of 0.8. A2-70 encodes tensile only — which is why the swap costs 12% of tensile and 30% of yield.
A4-80 encodes two independent decisions: corrosion grade and strength. Buyers routinely change one while assuming they changed both.
Zinc is a consumable that runs out and warns you as it goes; stainless repairs itself but only where oxygen reaches. And at the same size the stainless screw is the weaker one — a strength change disguised as a corrosion change.
Galvanic corrosion needs three conditions at once, and removing any one stops it. Severity comes from the area ratio — and a fastener is the smallest piece of metal present, so whether it is the anode decides the outcome.
A single-side coating of t adds 4t to pitch diameter. At 6h the upper deviation is zero, so the allowance has to be put on the drawing — and it surfaces on the first production lot, not at sample stage.
The number is how much variation is allowed; the letter is where the band sits — which is really the coating budget. And a higher number means looser in ISO and tighter in the inch system.
Hot-dip galvanizing puts 50–100 µm on the surface, and an M2 thread has tens of microns of total tolerance. The best line in the corrosion table is not expensive here — it is geometrically impossible.
ASTM B117 says in its own scope that results do not reliably predict service life. And a requirement that does not name white rust or red rust has specified nothing at all.
The zinc is corroding, not the steel — and freshly plated parts are the most vulnerable, not the least. Which makes plating line straight into a sea container the highest-risk route there is.
It passes every incoming test and splits on the line a day later. The delay is not incidental — it is why the control has to sit in the supplier's process record rather than your goods-in procedure.
Once production has started, the cost of a change is whatever it obliges you to re-verify.
The screw is often the last adjustable part in an assembly — but adjustable is not free. Length, head, material and finish each pull a different list back into verification, and the ones that touch preload and embrittlement are the long ones.
Comparing two quotations misses schedule, verification and risk — and the failure costs differ by an order of magnitude. With a real case: a tapping fault that snapped standard screws during assembly, solved by taking the major diameter down.
ISO 16047 starts at M3. Charpy needs a 16 mm shank. The decarburisation hardness method needs a 1.25 mm pitch. Four unrelated standards, one pattern — and what to ask once they run out.
Under EN 10204 there are four different inspection documents and they are not interchangeable. What each type gives you, which fields to check, and what to put in the purchase order.
Lead time, stocking models and compliance — what starts after the specification is fixed.
Delay rarely happens in the making — it happens in the waiting. Big-order priority, setup batching and outsourced process steps, plus the seven questions that surface each of them before the order is placed.