The belt cut test: how do you prove a belt weigher is accurate?

Is it accurate, or is it just calibrated?

These sound like the same question, but they are not, and the difference catches people out. Calibration is an adjustment: we set the instrument's span against a known reference (test weights, a calibration chain, or a roll of known mass) so it reads correctly at that point. Verification is a check: it compares the instrument's output against an independent reference under real running conditions, with real material, and tells you whether the calibration is holding.

A belt weigher can be freshly calibrated and still read wrong under load: belt tension changes, idlers wear, material builds up on the weigh frame, a load cell drifts. Calibration says "we adjusted it." Verification says "we proved it." A belt cut is a verification.

What are the ways to check a belt weigher's accuracy?

There is a ladder here, from a quick indicator to a definitive answer, each rung costing more effort for more confidence.

• Test weights or billets: hang or roll a known mass across the weigh frame and see if the integrator picks up the expected loading. Good for a routine health check, but it only proves the instrument responds to a static reference, not that it is accurate on your material at your flow.
• Material test against a weighbridge: run a known batch across the belt weigher, then weigh the same material on a verified weighbridge (a truck load, or a divert to a weigh hopper). It checks the whole system on real product, but needs a trusted weighbridge and enough material to mean something.
• Static belt cut: stop the belt and weigh the material on measured sections. No weighbridge needed, no material diverted. This is the workhorse field check and the focus of this article.
• Formal NMI verification: for trade-approved instruments, a regulated verification to National Measurement Institute requirements. That is a regulated check, and a different thing to the maintenance calibration covered here.

What is a static belt cut test?

Safety first, and this cannot be skipped. A belt cut means people working on a stopped conveyor and physically removing material from the belt, so the conveyor has to be properly shut down and isolated before anyone approaches it. In practice that means lockout/tagout to your site procedure, the belt fully stopped and de-energised with stored energy released and a zero-energy state confirmed, a minimum of two people, and the site permits and JSA signed off first. On coal sites there are additional dust and gas controls that form part of the full procedure. What follows is a summary to help you understand the test, not a work instruction, and no site should attempt it without trained people and its own controlled procedure.

With that in place, the test is straightforward in concept:

1. Run the belt under a steady, representative load and record what the integrator (the belt weigher's electronics) is reporting: flow rate in tonnes per hour and belt speed in metres per second, or kg/m directly if it shows that.
2. Stop and isolate the belt.
3. Mark three separate one metre sections of loaded belt: one at or near the weigh frame (the most representative point), one a short distance upstream, and one a short distance downstream.
4. Scrape and brush all the material off each marked section into its own clean bucket, taking the full width of the belt and every last fine.
5. Weigh each bucket on a calibrated reference scale and subtract the empty bucket weight. That net weight is the real kilograms per metre for that section.

You now have two numbers for the same loading: what the instrument said, and what the belt actually carried. The gap between them is the accuracy of the belt weigher. We take three cuts rather than one because material is never spread evenly, and the spread across the three tells us how far to trust the average.

How are belt cut results judged?

Three things come out of the numbers.

Error percentage. We convert the integrator's reading to kilograms per metre and compare it to the three-cut average. Reading higher than the material weighed is over-reading (the customer is told there is more product than there is); reading lower is under-reading. As an illustration with assumed figures: if the integrator works out to 45.0 kg/m and the three cuts average 45.9 kg/m, it is under-reading by about 2%.

Consistency between the cuts. If the three samples are close, the loading was even and the average is trustworthy; if they are scattered, the belt was loaded unevenly and the test tells you less. As a guide, variation under about 5% is good, and over about 10% is a signal to re-run rather than trust the result. A big scatter usually means uneven feed, segregating product, or sloppy collection.

Accuracy bands. Different duties need different accuracy, and the band depends on what the number is used for:

• Around plus or minus 0.5% for trade and fiscal metering, where money changes hands on the weight.
• Around plus or minus 1% for inventory and blending control.
• Around plus or minus 2% for non-fiscal process indication, where the reading guides a process but nobody is billed on it.

These bands are service criteria we use to decide whether an instrument needs attention, not a trade certificate. For a trade-approved (NMI) instrument, a static belt cut is a diagnostic tool only: it shows how the instrument is performing, but formal legal-for-trade verification is a separate, regulated process to National Measurement Institute requirements.

What do we do with the result?

If the error is inside the band for that instrument's duty and the cuts are consistent, we record a pass, note the next check interval, and move on.

If the cuts are consistent but the error is outside the band, the instrument is reading proportionally wrong across the board, which points at the span calibration. We adjust the span in proportion to the error, then re-run the belt cut to confirm the correction landed. We never adjust and walk away: an unverified adjustment is just a hopeful guess.

If the cuts are scattered, or the error is large but the span looks right, the problem is usually elsewhere: belt speed sensor, load cell output and zero, weigh frame alignment and cleanliness, idler condition, or material sticking to the belt that never reached the bucket. Chasing span on a mechanical fault just hides the real issue until the next drift.

When is a belt cut the right tool, and when is something else better?

Reach for a static belt cut when you need a real accuracy figure on real material without a trusted weighbridge to hand, or when you cannot divert flow. Choose a material test against a weighbridge when you can run a decent batch and want to prove the whole system end to end, or when a customer wants a comparison against a scale they already trust. Use test weights or billet checks for a fast health check between full accuracy tests: they catch a gross fault quickly, but treat them as an indicator, not proof of accuracy on material. For a trade-approved instrument, none of these maintenance checks replace formal NMI verification.

Frequently asked questions