Why are my belt weigher totals erratic?

This is the deep dive on unstable readings. For the wider symptom list see common belt weigher faults.

Is this really erratic, or is it drift or a weighbridge gap?

Erratic means the reading will not sit still: the rate wanders while the plant feeds steadily, repeat calibration runs scatter instead of stacking, and two shifts moving the same material do not total the same. That is not the same as the other two calls we get most. A reading wrong by a steady amount in the same direction, service after service, is bias: see zero drift. One that repeats beautifully but sits a consistent margin off the truck weights is a comparison problem: see belt weigher versus weighbridge. Erratic is scatter, not bias, and it shows up first as repeatability getting worse while the average still looks right.

What does the pattern of the noise tell me?

The shape of the noise names the family of causes before anyone lifts a spanner.

• Random scatter with no rhythm points at the electrics: load cell wiring, moisture in a junction box, a corroded terminal, earthing or interference.
• A cycle repeating once per belt revolution points at the belt: a splice, a thickness change, or the belt lifting off bed rollers through a curve. It averages out over a revolution, so a daily total looks sane while short windows read badly.
• Short sharp spikes on a smooth trend usually come from the link, not the weigher: polling or missed pulses on a counter input.
• Noise that comes and goes with conditions has a physical trigger: a wet belt slipping a speed sensor, a tired structure flexing under load, drive interference at one motor speed.

One check splits it in half: is the noise on the integrator's own face, or only in the control room? A steady integrator and a jumpy screen puts the fault downstream.

Why do we check the speed signal first?

Because speed multiplies everything, and the sensor lives in the dirt and weather as the most exposed part of the system. A spiral cage or tacho wheel riding the belt wears smooth and slips when wet, and its bearings go noisy, then seize, stopping the cage altogether. A loose bracket or slackened shaft coupling does it quietly, and a shaft-driven sensor reads steady while the belt slips on the pulley. Then the wiring: damaged cable, terminals shaken loose by vibration, moisture in the sensor junction box, and pulses dropping out at the belt speed where drive interference bites. After a like-for-like swap, the wrong sensor output type for the integrator's input gives no valid pulse at all. Time a marked length of belt against the displayed speed: more than about two percent apart warrants investigation, and because tonnage tracks speed one for one, that is two percent off your tonnes. More in our speed sensor guide.

Can a load cell or its junction box make the readings jump?

Yes, and on a wet site the junction box is the second place we look. Water in it is one of the most common causes of erratic behaviour we find: we dry and reseal, but the real fix is stopping the ingress, and a box whose seals have gone wants replacing. Corroded terminals and a nicked cable sheath do the same, and very often the cell is healthy, which is why we prove the wiring before condemning anything. On a multi-cell frame one cell can misbehave while the rest are fine, so we check-weight each in turn: one reading noticeably differently under equal load is damaged or on a degraded cable.

Earthing is the invisible one, and recalibrating will not touch it. If the structure is earthed at one point and the integrator cabinet at another, and the two sit at different potentials, current flows between them through the signal wiring and reads as noise. See belt weigher load cells and protecting belt weigher electronics.

What mechanical problems make steady flow look unsteady?

Plenty, and these are the ones a site can often spot for itself. A belt tracking off centre edge-loads the weigh idlers, and because tracking wanders in and out, the reading wanders with it. Build-up gives both symptoms at once: sitting on the carriage it reads as a steady bias, but a slab that lets go mid-shift steps the reading. Anything that stops the weigh frame floating does the same, whether a bind between frame and stringer, a seized idler in the weigh span, or tie rods stiffened by corrosion.

Vibration is the classic cause of scatter that cleaning never fixes. A noisy roller in the idler frame before the weigh rollers, or a return roller under the weigh area, shakes the frame directly, and so does an ageing structure that flexes, a cracking carriage, or an adjacent conveyor on shared steel. Material does it too: a weigh area too close to a loading point, excessive belt sag or a surging feed hand the scale a load that really is jumping, and the reading is honest. Our installation guide covers why position matters.

Can the integrator or control system settings be creating it?

They can, in both directions, so we treat settings as a suspect rather than a cure. Damping and filtering hide noise: wind the damping up and the rate goes smooth while the fault carries on wrecking your totals, and nobody notices until the month does not add up.

Settings also invent noise. On a pulse output, where a contact closes once per unit of totalised weight, a weight-per-pulse set too fine or a pulse time too short for the receiving counter means pulses go missing at high rate and the control room total quietly falls behind the integrator's own. We have fixed exactly that by lengthening the pulse time. On an analogue rate output the integrator's full scale and the receiving scaling must agree exactly, or every t/h figure is out by a fixed factor. A SCADA trend sampling a fast rate every few seconds shows real instants that are not representative, and two systems totalising over different windows never agree at shift change. See PLC and SCADA comms.

What can I check safely, and when should I call you?

Work on or near a conveyor, or inside an electrical enclosure, needs isolation under your site’s procedures and trained, authorised people. If that’s not you, stop and call us on 1300 101 666.

Quite a lot, from a safe distance with the belt running: watch how it tracks through the weigh zone, confirm the speed sensor wheel is turning on the belt, and listen for a bearing. Compare the integrator’s rate with the control room at the same moment, and trend two shifts to see if the noise is random, cyclic or spiky. Once the conveyor is isolated, trained people can look for water in the junction boxes, cable damage, and build-up on the weigh frame and sensor wheel.

Call us when the noise will not tie to anything you can see, when totals stop repeating between shifts, when repeatability worsens each service, or when a weightometer turns erratic after a storm. Our diagnosis is elimination, not adjustment. We work with repeatability rather than finished tonnes: runs that used to stack and now scatter are the clearest measure of instability there is. Then we split the chain: check weights on each load cell, junction box and cabling inspected, belt speed timed against a marked length, output proven end to end against the integrator's face. Only then do we calibrate, and we re-verify afterwards. Where the installation simply cannot be steady, we say so rather than dial in a number that will not hold. For an independent check, see the belt cut test.

Frequently asked questions