What Torque Actually Controls and What It Does Not

Why this matters

A calibrated torque wrench feels like a measurement instrument, and it is - it measures torque. The problem is that torque is not what holds a joint together. Clamp load does. Torque is a proxy for clamp load, and it is a loose one, so a joint can be torqued exactly to a published value by a competent tech with a calibrated tool and still be assembled wrong. Until you know which part of that relationship you control and which part you do not, every joint failure that follows a correct torque reading looks like bad luck.

Before loosening any fastener on a joint that contains pressure, refrigerant, fuel gas, or a stored spring load: isolate the source, relieve pressure to zero and confirm it on a gauge at the joint, and de-tension or restrain any spring or loaded member before the first fastener moves. 29 CFR 1910.147 requires stored energy to be relieved, disconnected, or restrained before servicing. Loosening a bolt to check its torque on a live joint is not a check, it is a release.

The call: a flange torqued to spec that leaked

A shop was called back to a flanged joint on a pump they had rebuilt the previous week. The joint was weeping on one side. The lead tech had done the work himself, with a wrench that had been calibrated that year, at the published torque value, in a crossing pattern, in two passes. He was, understandably, sure the torque was not the problem.

He was half right. The torque was not the problem. The clamp load was, and the two are not the same thing.

What the wrench was actually measuring

Torque is rotational effort. Preload, also called clamp load, is the tension you develop in the fastener, which is what actually squeezes the joint together and keeps it there. You want preload. You apply torque. Everything in between is friction, and friction is a property of the surfaces, not of the wrench.

The relationship used in the field is straightforward in form: torque equals a friction factor, times the fastener diameter, times the preload. The friction factor is usually called the nut factor, and it is the whole story. It is not a constant. It is a summary of the thread condition, the bearing face condition, the plating, whatever lubricant is or is not present, and how many times that fastener has been used.

Published nut-factor tables commonly show something around 0.20 for plain dry steel, dropping to roughly 0.13 with a good thread lubricant, and rising toward the 0.25 to 0.30 region for corroded or damaged threads. Those are typical published ranges, not a specification for your fastener, and the manufacturer's own value governs where they publish one.

Read what that range means before going further. Torque is fixed by the wrench. Diameter is fixed by the bolt. So preload varies inversely with the nut factor, and a nut factor that ranges from 0.13 to 0.28 across a set of fasteners produces preloads that vary by more than a factor of two at identical torque.

Where the torque actually goes

Only a small fraction of what you put into the wrench becomes clamp load. The usual accounting for a plain steel fastener puts roughly half the input torque into friction at the bearing face under the head or nut, roughly another third to two fifths into friction in the threads, and only something in the region of 10 to 15 percent into stretching the bolt.

That has two consequences worth carrying.

First, you are mostly measuring friction. The wrench is dominated by the two friction terms. A change in either one moves preload substantially while the wrench reading stays exactly where you set it.

Second, the bearing face matters as much as the threads. Techs lubricate threads and ignore the washer face, which is the larger of the two friction terms. A galled washer face, a burr, or paint under the head changes preload as much as the thread condition does.

Even with everything done carefully - clean matched fasteners, consistent technique, calibrated tool - torque control is commonly quoted as delivering preload within about plus or minus 25 to 30 percent. That scatter is inherent to the method, not a sign of sloppy work, and it is why critical joints use a different control method entirely.

The friction spread around that flange

Back to the pump. Eight studs, all reused from the original assembly, all torqued to the same published value with the same wrench.

Three studs were clean and dry, matching the condition the published dry value assumes. Take those as the 100 percent case at a nut factor near 0.20.

Two studs carried anti-seize residue from the previous service, never cleaned off. At a nut factor near 0.13, preload goes as 0.20 divided by 0.13, which is about 1.54. Those two studs are at roughly 154 percent of the intended clamp load - about 54 percent over.

Three studs had light thread corrosion from sitting in a damp room. At a nut factor near 0.28, preload goes as 0.20 divided by 0.28, which is about 0.71. Those three are at roughly 71 percent of intended - about 29 percent under.

Three plus two plus three is the eight studs on the flange, and the spread across them is 154 percent against 71 percent, a ratio of about 2.2 to 1 in clamp load, around a gasket that needs even compression to seal.

The weep was on the side carrying the corroded studs. Nothing about the wrench, the value, the pattern, or the number of passes was wrong. The assumption that eight fasteners at the same torque are at the same tension was wrong, and it was wrong by more than a factor of two.

The failure mode this produces, and the reason it is worth the arithmetic: the over-tightened studs are not harmless. At 154 percent of a preload target that is already set near a healthy fraction of the fastener's capacity, those two studs may be into yield, which means they will not return their tension after the first heat cycle. So the joint has three studs too loose to seal, two studs that will relax, and a customer who is told it was torqued to spec.

What torque does not control at all

Beyond the friction problem, there is a set of things a torque value has no authority over. Applying more of it does not reach any of them.

• Even compression around a gasket. That is controlled by the tightening sequence and the number of passes, not by the final value. A perfect value applied in a circle instead of a crossing pattern leaves the last-tightened side high and the first side low.
• The condition of the mating faces. A scratch across a sealing face, a raised burr, or a face that is not flat is a leak path, and clamp load cannot close a radial scratch. Trying is how a joint gets crushed.
• The gasket's actual state. A gasket that has taken a permanent set does not spring back, so it cannot follow the joint as it moves. Torque set correctly on a set gasket seals until the first thermal cycle.
• Thread engagement. A fastener that is only engaged a few threads can hit its torque value and still be holding on almost nothing. The wrench cannot tell you how much thread is in the hole.
• Whether the fastener is the right one. A wrong-grade or wrong-material fastener reaches the same torque number. See the sibling article on fastener material for what happens next.
• What happens after you leave. Embedment of surface roughness, gasket creep, and thermal stress relaxation all bleed preload off a joint that no one has touched. Torque is a snapshot at assembly.

The methods that control preload more directly

Where the joint matters, move up the ladder. Each of these narrows the gap between what you set and what you get.

• Turn of the nut, or angle control. Snug the fastener to a defined light torque, then turn it a specified additional angle. Past snug, rotation relates directly to stretch, so this bypasses most of the friction problem. It requires the manufacturer's snug value and angle, and it is not something to invent.
• Bolt stretch measurement. Measure the fastener's length before and after and tighten until it has stretched by a specified amount. This measures the thing you actually want. It needs access to both ends and a specified stretch figure.
• Load-indicating hardware. Washers or fasteners that give a visible or mechanical signal when a target load is reached. The signal, not the wrench reading, is the acceptance criterion.
• Controlled, matched conditions with torque. If torque is the only method available, you narrow the scatter by controlling the variables: identical fasteners from the same lot, clean threads and clean bearing faces, the same lubricant on all of them or none of them, and a torque value that matches that lubrication state.

That last point is the one most often broken. Published torque values state whether they are dry or lubricated. Applying a dry value to lubricated threads over-tightens the fastener by roughly the ratio of the two nut factors, which on the figures above is about 54 percent, and that is enough to yield a fastener that was correctly specified.

What the fix was, and the part that was not a torque value

The flange came apart, with the line isolated and relieved first. The old studs went in the scrap bin rather than back in the holes, because a reused stud carries an unknown friction history and a used gasket carries a set. New studs and nuts from one box, threads and bearing faces clean and dry, one gasket, the published dry torque value, a crossing pattern in rising passes.

Two things then changed in how the shop worked, and neither one is a number.

The rebuild procedure got a line stating that studs and nuts on gasketed joints are replaced, not reused, with the reason written next to it so the next person does not optimize it away. And the ticket got a field for the lubrication state used, because a torque value recorded without its lubrication state is not a record of anything - the same number means two different clamp loads.

Confirmation was a re-check after the first full heat cycle rather than a pressure test on the day. The first test tells you the assembly holds cold; the second tells you it survived the relaxation that follows the first heat cycle, which is when a marginal joint actually declares itself.

Checking you got this right

• Confirm the published value you are using states a lubrication condition, and that the fastener matches it. If the value does not state one, treat it as dry and do not lubricate.
• Run a thumb over every bearing face and thread before assembly. You are checking for burrs, paint, plating damage, and leftover compound - the friction terms that the wrench cannot see.
• Use fasteners from one box on one joint. Mixed sources means mixed plating means mixed friction.
• Record the value, the pattern, the number of passes, and the lubrication state. A recorded value alone cannot be reproduced.
• On any joint that runs hot or carries a gasket, schedule the re-check after a full heat cycle and note the temperature at which it was checked. A joint checked hot and a joint checked cold give different readings and neither is the wrong one.

References

• Manufacturer torque specifications for the specific joint, including the stated lubrication condition and any specified re-torque interval
• SAE and ASTM fastener standards for proof load and tensile stress area, used to relate a torque value to a fraction of fastener capacity
• Published nut factor and torque-tension data from fastener manufacturers for the specific plating and lubricant
• 29 CFR 1910.147, energy control, for relieving or restraining stored energy before servicing a pressurized or spring-loaded joint
• See related: The Torque Sequence That Matters; The Fastener That Was the Wrong Material; Why Threaded Connections Leak