What Torque Actually Measures and What It Does Not

Why this matters

A torque wrench is a resistance-to-rotation meter. It reads what the fastener is fighting back with at the instant the fastener turns. That sentence contains both of the limits that get shops in trouble: it needs movement, so it cannot read a fastener that is sitting still, and it reads total resistance, so it cannot tell you which part of that resistance is doing useful work. A tech who knows the spread between torque and clamp load still trips over these two, because they look like measurement rather than like assumption.

Before putting a wrench on any installed fastener: a re-check is not a passive act, it can release the joint. On a pressurised or spring-loaded assembly, isolate and relieve to zero at the joint and restrain any loaded member first, which is the stored-energy duty at 29 CFR 1910.147 in general industry. On an electrical termination, de-energize and lock out under 29 CFR 1910.333(b)(2), and prove the conductor dead with the live-dead-live sequence in NFPA 70E-2021, 120.5, which reaches you through your employer's electrical safety program or a contract rather than as federal law on its own. Never re-torque a joint that is hot and under pressure: the burn and the release are the same motion.

The sibling article on what torque controls owns the friction-scatter problem, the nut factor, and the preload spread it produces. This one is about what the reading cannot see at all.

Deliberately not on the list

Four things a torque reading will never give you, no matter how good the wrench is.

• The tension currently in an installed fastener. There is no torque value you can apply to a stationary bolt that reports its clamp load.
• The share of your reading that becomes clamp. A locking feature, a damaged thread or a dry burr all resist rotation and none of them stretch the bolt.
• Whether anything moved. A bolt that spins its own captive bracket, a nut that turns with the stud, or a fastener bottomed in its hole all give a rising reading with the joint unchanged.
• What lever arm was actually in play. The number on the wrench is the torque at the wrench's own drive, not necessarily at the fastener.

Each of those is worth its own arithmetic, because the size of the error is what makes them matter.

The prevailing-torque case, worked

A specification calls for 50 ft-lb on a 3/8-16 Grade 5 bolt with a plain nut. The shop is out of plain nuts and uses an all-metal prevailing-torque locknut instead, the kind with a deformed top thread that grips. Say that nut runs down its free thread at 8 ft-lb before it ever touches the joint.

The tech sets the wrench to 50 and it clicks. What actually happened:

• 8 ft-lb of the reading is the locking feature deforming and dragging. It produces no clamp force at all.
• 42 ft-lb is the part doing tightening work.
• Against the intended 50 ft-lb of tightening work, the joint is at 42 divided by 50, which is 84 percent. Preload is roughly 16 percent light, and the ticket says torqued to spec.

The correct handling is in how prevailing-torque fasteners are specified. Their torque values are commonly stated as prevailing torque plus a seating value, which means the running torque of that specific nut is measured and added. So the target here is 8 plus 50, which is 58 ft-lb.

To measure the running torque, thread the nut down the free thread on the de-energized, depressurised assembly, before it reaches the bearing face, and read the wrench while it is turning. Read it during rotation, not at a stop, because the wrench only reports while the fastener moves.

What flips this. A used prevailing-torque nut has less running torque than a new one, because the locking feature is a one-way deformation that wears with each use. Many manufacturers limit reuse for exactly that reason, and the manufacturer's own reuse limit governs. If you re-use one and add the new-nut running torque, you over-tighten by the difference. Measure the running torque of the nut in your hand.

The re-check case, which is the one that fools people

A month later a different tech is asked to verify the joint. He sets a click wrench to 50 ft-lb, applies it to the same locknut, and it does not move. He records the joint as checked and holding.

He learned almost nothing, and here is why in the same terms.

What resists a stationary fastener is static friction across the thread and bearing face, plus the locking feature. The torque needed to start it moving is called breakaway torque, and it has no dependable relationship to residual clamp load. Run the two ends of the range and the point makes itself.

• A fastener with full clamp and clean threads may break away at close to its installation torque, or above it if the threads have set.
• A fastener with zero clamp can still refuse to move. This nut alone contributes 8 ft-lb of prevailing torque with the joint completely slack, and a corroded thread can hold far more than the installation torque with no clamp behind it at all.

So the same "did not move at 50" result is produced by a healthy joint and by a corroded stud with nothing left in it. The test cannot separate them, which means it is not a test.

Worse, the reading is destructive in the other direction. If the fastener does move at 50, the tech has not measured anything either. He has just tightened it further, into a joint that has since settled, and the new preload is unknown.

What to do instead, in rising order of what it costs you:

• Mark the fastener at assembly. A paint or marker line across the nut, washer and member turns any later rotation into a visible fact. This is the cheapest honest re-check there is and it answers the question the wrench cannot.
• Loosen and retighten to value where the joint tolerates being opened, having relieved stored energy first. You have replaced an unknown with a known.
• Use a method that measures the thing itself on a joint that matters: angle control past a defined snug point, bolt stretch measurement, or load-indicating hardware. The sibling article covers where each belongs.

The lever-arm case

Torque is force times distance, so the distance the wrench thinks it has is part of the reading. Put an adapter on the end of the wrench and that distance changes.

Take a wrench with an 18 inch effective length from the drive square to the centre of the handle grip, and a crowfoot adapter that adds 2 inches in line with the wrench axis. The torque delivered at the fastener is the wrench setting times the new length over the old:

• Set 50, deliver 50 x 20 divided by 18, which is about 55.6 ft-lb. That is 11 percent over.
• To deliver 50, set 50 x 18 divided by 20, which is 45 ft-lb.

State the geometry with the correction, because it only applies that way: this is the in-line case. Mount the same adapter at 90 degrees to the wrench axis and the effective length is essentially unchanged, so no correction applies. Mount it at some angle in between and the correction sits between the two. A shop that applies the in-line correction to a crowfoot turned sideways under-tightens by 10 percent while believing it is being careful.

What the reading is genuinely good for

Naming the limits is not an argument for abandoning the tool. Torque is the right method for the large majority of field joints, and it does three things well.

• It is repeatable. The same tech, the same fasteners, the same lubrication state, the same value gives a consistent result across a set of bolts, which is what a multi-bolt gasketed joint needs most.
• It is a ceiling. It stops the strong tech from yielding the fastener and stops the careful tech from leaving it slack. Both failures are common and both are worse than the scatter torque leaves behind.
• It is recordable and auditable. A value with its lubrication state, pattern and pass count written down is a reproducible assembly. That matters more than the last few percent of accuracy.

When to stop and get a different method

• The wrench climbs and the fastener has not turned. Something is bottomed, seized, or the whole assembly is rotating. Stop and look rather than adding value; that is how heads snap.
• The reading is smooth and rising with no seating step. On a joint that should have a distinct hardening as it pulls up, a soft continuous climb usually means thread damage or a fastener stripping in the hole.
• Different fasteners on the same joint reach value at wildly different rotations. One nut arriving at value half a turn before its neighbours means their friction states differ, and the assumption behind torque control has already broken.
• The joint is critical and cannot be opened later. Route it to angle control or stretch measurement and get the specified snug value and angle from the manufacturer rather than inventing them.

Checking you got this right

• For every prevailing-torque fastener, confirm whether the published value already includes prevailing torque or expects it added, and measure the running torque of the nut in your hand rather than a table value.
• Confirm the adapter geometry before applying any correction, and skip the correction entirely for an adapter at right angles to the wrench.
• Mark every fastener you tighten on anything you expect to revisit. A month later that mark is the only piece of evidence in the joint that a wrench cannot fake.
• If you were asked to verify a joint and your answer is "it did not move," say what that does and does not prove in the note, or the record will read as a passed test.
• Never record a torque value without its lubrication state and its fastener type. Fifty ft-lb on a plain nut and fifty on a locknut are two different assemblies.

References

• 29 CFR 1910.147, control of hazardous energy, for stored energy released by opening or slackening a joint
• 29 CFR 1910.333(b)(2), for de-energizing and locking out circuits before working on an electrical termination
• NFPA 70E-2021, 120.5, for the live-dead-live verification sequence, binding through an employer's electrical safety program or a contract rather than on its own
• Fastener manufacturer documentation for prevailing-torque values, reuse limits, and whether a published torque includes or excludes prevailing torque
• See related: What Torque Actually Controls and What It Does Not; How to Tighten a Bolted Joint and Know It Is Right; Why a Bolted Joint Loosens