Torque Wrench Calibration: Intervals, Tolerances and Records

Key takeaways

• A torque wrench is a measuring instrument that lives in a toolbox. Spring relaxation, drops and overload all shift its reading, and nothing on the joint tells you it happened.
• ISO 6789 is the reference standard: Part 1 (2017) sets the requirements and marking for hand torque tools, Part 2 defines the calibration method and its check points.
• Many industrial click wrenches are specified at +/-4 percent of the set value, some classes at +/-6 percent. The marking on the wrench and its declared class decide, not a universal number.
• The widely used recalibration rule is 12 months or about 5,000 load cycles, whichever comes first, tightened immediately by events: a drop, an overload, a repair, or joint failures that point at the tool.
• The as-found reading is the evidence. It decides whether product built since the last good check is suspect, so quick torque checker verifications between calibrations shrink the containment window from months to weeks.

Why torque wrenches drift

A click wrench stores its setting in a compressed spring acting on a release mechanism. Springs relax under sustained load, the mechanism wears with every cycle, and the pivots take a knock every time the wrench hits a concrete floor. Overload is worse: one pull past the click, or one seized fastener broken loose with the wrench, can shift the reading permanently.

None of this is visible. The wrench still clicks, the fitter still stops at the click, and every joint downstream inherits the error. The tables in our flange bolt torque guide are only as good as the wrench that applies them: the whole argument for treating torque tools as calibrated instruments rather than hand tools.

What ISO 6789 actually covers

Since the 2017 revision the standard has two parts. ISO 6789-1 covers requirements and marking (range, unit, direction) plus the manufacturer's conformance declaration. ISO 6789-2 defines the calibration method: check points, number of readings, and how measurement uncertainty is determined.

The standard splits tools into two families:

• Type I, indicating: the tool shows the applied torque on a scale, dial or display (beam, dial and electronic wrenches).
• Type II, setting: the tool is preset to a target and signals when it is reached (click wrenches, slipping and cam-over tools, preset screwdrivers).

Each family is subdivided into classes, and the class carries the accuracy requirement. Commonly, industrial click wrenches hold +/-4 percent of the set value, with +/-6 percent applying to some classes and to small capacities around 10 Nm and below. Treat those figures as typical, not universal: the class marked or declared for your wrench decides, and the calibration certificate should state it.

When to recalibrate: the interval and the triggers

The default most plants adopt, and that traces back to the guidance in ISO 6789, is 12 months or approximately 5,000 load cycles, whichever comes first. The standard leaves the interval to the user; your quality system sets the binding number.

The cycle half of the rule is the half that gets ignored. A wrench doing 250 torqued joints per shift reaches 5,000 cycles in about 20 working days. Anniversary-based calibration on that tool means it spends most of its life past due.

On top of the interval sit the event triggers, each of which means "verify before the next use":

• The wrench was dropped onto a hard surface.
• It was overloaded: pulled past the click, used above its range, or used as a breaker bar.
• It was repaired or adjusted in any way.
• Joint failures cluster on connections this wrench torques: leaking flanges, loosening bolts, stripped threads. Trending failures per asset surfaces this fast, the same way a falling MTBF exposes a repeat offender machine.

Verification between calibrations

Calibration and verification are different jobs. Calibration establishes what the tool actually reads, with a stated uncertainty, following ISO 6789-2, and ends in a certificate: laboratory work. Verification is a quick in-house check on a bench torque checker that answers one question: is the wrench still inside its tolerance today. It takes minutes and shrinks the risk between certificates.

A sound in-house check borrows the ISO 6789 check points: 20, 60 and 100 percent of the wrench's range, in the direction of use, with a few readings at each point after one or two exercise cycles to settle the mechanism. Record every reading, pass or fail. A verification that only lives in the fitter's memory is not evidence.

Verification is a defined, trainable task with a right and a wrong way (loading rate, hand position, exercising the tool first), so name who is authorised on your maintenance skills matrix rather than leaving it to whoever is near the bench.

A worked example with real numbers

A 40 to 200 Nm click wrench, marked class tolerance +/-4 percent of set value, comes to the bench for its monthly verification. Check points at 20, 60 and 100 percent of range: 40, 120 and 200 Nm, three readings each on the torque checker:

• 200 Nm: mean 197.6 Nm, deviation -1.2 percent. PASS (band 192.0 to 208.0 Nm).
• 120 Nm: mean 117.9 Nm, deviation -1.8 percent. PASS (band 115.2 to 124.8 Nm).
• 40 Nm: mean 37.6 Nm, deviation -6.0 percent. FAIL (band 38.4 to 41.6 Nm).

Failing at the low end first is the classic pattern: spring relaxation costs a roughly constant couple of newton metres, invisible at 200 Nm and fatal at 40. Two decisions follow, in order:

• Containment. Every low-setting joint this wrench torqued is suspect, back to the last passed check. Verified monthly, that window is four weeks; without interim checks it is the eleven months back to the last certificate. That difference is the business case for verification.
• Interval. The wrench goes out for calibration and repair, and drift history sets the new rhythm: two calibrations showing the low end walking toward the limit earn a shorter interval or a weekly bench check; a one-off after a known drop leaves the standard interval standing.

As-found, as-left, and the record

Every calibration and every verification produces two states. As-found is the tool as it came in, before any adjustment: the number that tells you how it behaved in service. As-left is the tool after adjustment, the number the next period starts from. A calibration report without as-found values has destroyed the evidence, the same mistake as cleaning a safety valve before the first pop in safety valve testing.

The record per wrench: tool ID, range and class, check points, individual readings as-found and as-left, pass or fail against the marked tolerance, who checked it, on which checker, and the next due date by both calendar and estimated cycles.

This is scheduling and record keeping, the same discipline as the rest of the preventive maintenance plan. In Fabrico the monthly bench checks and annual calibrations run as recurring PM tasks with checklists per tool, as-found and as-left readings are stored against the tool's history, a failed check can trigger a follow-up task, and the whole trail is there when an auditor asks how you control torque tools. To see that set up on your own tool list, book a short demo.

Storage and handling rules that protect the calibration

• Wind click wrenches down to the lowest scale setting after use, never below it. Storing the spring compressed accelerates relaxation; winding below the minimum marking can unload and misalign the mechanism.
• Never use a torque wrench as a breaker bar, in either direction. Breaking fasteners loose is a job for a bar that costs a fraction of a calibration.
• One smooth pull to the click, then stop. Re-clicking a joint adds torque; jerky pulls overshoot.
• Correct for anything that changes the effective length. An extension in line with the handle changes the lever arm, so correct the set value (set torque = target × wrench length / total length). A crow's foot at 90 degrees to the handle needs no correction.
• Case, not toolbox floor. And any wrench that hits the floor gets a bench check before its next joint, no exceptions.

Common mistakes

• Calibrating on the anniversary and never counting cycles. In high-volume assembly the 5,000 cycle limit arrives in weeks. Estimate cycles from joints per shift and let whichever limit comes first drive the schedule.
• No as-found record. A wrench adjusted before its as-found readings were captured leaves you unable to answer the only question that matters after a failure: was the product built with a good tool.
• Leaving the wrench wound up. The spring keeps relaxing the whole time it sits compressed; wind down before storage.
• Uncorrected extensions and adapters. Anything that lengthens the drive line changes the torque delivered at the fastener. The joint gets the corrected number or the fitting gets a different tool.
• Treating a torque checker verification as a calibration. The bench check confirms the tool is inside tolerance; it has no stated uncertainty and no traceability of its own. The checker itself needs periodic accredited calibration, or one uncalibrated instrument is reassuring another.
• Ignoring the failed-joint signal. Loosening bolts and leaking flanges on joints one wrench serves are a calibration trigger, not a fitter problem: pull the tool the same day.

Frequently asked questions

How often should a torque wrench be calibrated?

The widely used rule is every 12 months or about 5,000 load cycles, whichever comes first, plus immediately after a drop, an overload or a repair. Your quality system sets the binding interval, and drift history should shorten it for tools that keep walking.

What tolerance should a torque wrench hold?

Commonly +/-4 percent of the set value for industrial click wrenches, with +/-6 percent applying to some classes and small capacities. The class marked or declared for the specific wrench under ISO 6789-1 is the tolerance that applies to it.

What is the difference between calibration and verification?

Calibration follows ISO 6789-2, establishes the tool's readings with a stated uncertainty and produces a certificate, normally from a laboratory. Verification is a quick in-house check on a torque tester, typically at 20, 60 and 100 percent of range, that confirms the tool is still inside tolerance between calibrations. Verification manages risk; it does not replace the certificate.

Should a click wrench be stored at its lowest setting?

Yes: wound down to the lowest marked scale setting, never below it and never left at a working setting. Storing the spring compressed accelerates relaxation, which shows up first as under-torquing at the low end of the range.

Does ISO 6789 require accredited calibration?

The standard defines the method, not who performs it. Whether the certificate must come from a laboratory accredited to ISO/IEC 17025 is decided by your quality system and your customers; regulated industries usually demand it, and traceability to national standards is what it buys.