Torque and force monitoring: what the curve tells you that a pass light does not

A final reading is weak evidence

Torque is a proxy for clamp load, and the relationship between them turns on friction, which turns on surface finish, coating, lubricant, thread condition and how often the joint has been assembled. A reading inside limits is consistent with a sound joint and with several unsound ones. That is not an argument against torque control, which remains the practical method for most joints, but it is an argument for treating one value as a single piece of evidence rather than as proof. Where clamp load genuinely matters, angle control past a snug point or direct load measurement supports a stronger claim.

The signature is the diagnostic

The shape of torque against angle, or force against displacement in a pressing operation, carries far more information than its endpoint. A missing component shows as a curve rising too late; a burr or misalignment shows as an early hump; a cross-thread shows as elevated resistance throughout the run-down; a soft joint and a hard joint have visibly different gradients. Monitoring against an envelope built from known-good assemblies catches cases a single limit waves through. Building that envelope needs real production data spanning material lots and shifts, and it needs revisiting whenever a component supplier changes.

Transducers drift and nothing announces it

The instrument judging every joint is itself an instrument, and it moves. Rotary transducers suffer shock loading when a tool is dropped, cables degrade, temperature affects output, and electrical noise appears once something else gets installed nearby. Nothing about a drifting transducer looks wrong on the screen. Verification against an independent measuring device on a defined schedule is the only way to know, with results recorded and trended rather than adjusted quietly. Decide in advance what happens to product assembled since the last good verification when a tool is found outside tolerance.

Limits from what the process does, not from the drawing alone

Limits taken from the drawing alone ignore what the process can hold. Collect the distribution from a stable run and compare it against the engineering requirement: where the spread nearly fills the tolerance, the station will reject sound assemblies regularly and operators will find ways round it. Where the spread sits far inside, the limits catch nothing at all. Set control limits from demonstrated capability and keep the engineering tolerance as the boundary of acceptability, and record who authorised each value, because these get widened quietly during a difficult launch and are never narrowed afterwards.

The record is a commercial asset

For a joint that matters, the fastening record is what you produce when a customer alleges a failure. That means tying results to a serial number or a batch, retaining them for as long as the product remains exposed, and preserving failures alongside passes, since a record showing only successes invites the question of what happened to everything else. Make sure the system cannot overwrite a failed result when a joint is re-run, and record rework explicitly. A complete, boring record is a strong position; a partial one looks like selection and is worse than none.

Frequently asked questions

Does reaching the specified torque prove the joint is tight?

It proves the tool reached a value. Whether the joint is correctly clamped depends on friction in the threads and under the head, which varies with coating, lubricant, surface condition and reassembly. A cross-threaded fastener, a missing washer or a burred face can each produce an acceptable reading. Monitoring the full curve catches most of those cases, and where clamp load is genuinely critical, angle control past a snug point or direct measurement gives better assurance than torque on its own.

How often should a torque transducer be verified?

On a schedule derived from how heavily the tool is used and how serious a wrong result would be, plus after any shock event such as a drop or a jam. Verification means comparison against an independent, traceably calibrated device, with the reading recorded rather than silently adjusted. Agree beforehand what happens to product built since the previous good verification if the tool is found outside tolerance, because settling that during an incident is considerably harder.

Should we keep every fastening record or only the failures?

Keep both, tied to the unit. Failures alone look like a selective record and invite exactly the challenge you want to avoid, while passes alone hide the rework entirely. Retention should match how long the product could generate a claim, which is a commercial and legal judgement rather than an engineering one. Ensure that re-running a joint after a failure creates an additional record instead of replacing the original, since the sequence of attempts is frequently the most informative part.