Overload switches and load weighing that drift

A car rated at 1000 kg has to detect an overload somewhere around 110 percent, raise an alarm, and refuse to close the doors. That is a safety function, and like every safety function it is only as good as its calibration. The problem is that load weighing drifts, quietly, in both directions. There are four common ways to weigh a car. Load cells under the car platform, usually four, wired into a summing junction. Sensors that measure rope or belt tension. A pressure transducer on a hydraulic system. And on some modern gear, a system that infers load from the motor current or from the car's position relative to a reference. Each has its own drift mechanism. Load cells drift with temperature and with age, and they hate water. Rope tension sensors drift when the ropes are re-tensioned or replaced. Pressure transducers drift with oil temperature. Motor-current inference drifts when the machine is re-roped or the drive is retuned. Then there is the building. A car that was calibrated before the lobby renovation now has 80 kg of marble on the floor. New ceiling panels, a mirror, a handrail, a different floor covering, and the empty-car zero is wrong by the weight of the renovation. The system does not know; it still thinks the car is empty when it is 80 kg heavy. The symptoms are easy to miss individually. A car that alarms at 90 percent of rated load, so the last two passengers get off and take the next one. A car that never alarms, even with the doors held open and six people standing at the back. A car that levels badly only when loaded. A car that starts with a jolt when full and smoothly when empty. Calibration is not complicated, but it needs weights and patience. Empty the car completely, including the maintenance tools and the mop bucket that lives in the corner. Zero the system. Then load in known steps, 25 percent at a time, using test weights, and record the reading at each step. The curve should be straight. A curve that bends at the top is usually a mechanical problem, not an electronics problem: a platform that is binding, a load cell that is bottoming out, or a rope tension sensor on a rope that is carrying more than its share. The 110 percent check comes last. Load to 110 percent, confirm the alarm sounds, confirm the doors will not close, and confirm the car will not start. Then take the weights off and confirm the alarm clears and the car runs normally. If the alarm is a buzzer in the car and a lamp in the controller, test both, because passengers ignore a lamp and maintenance ignores a buzzer. For the functional check, certified test weights are the right tool, but most buildings do not have 1100 kg of them. Water drums and sandbags are acceptable if you weigh them, and a car full of adults is not: people move, lean, and get off, and nobody knows what they weigh. If you use people for a quick sanity check, say so in the report, and do the real calibration with weights. There is a temptation, when a car alarms with a normal load, to raise the threshold. Do not. The threshold is set by the rated load and the standard, and raising it hides a sensor fault behind a number that now lies at both ends. If the system alarms early, find out why the zero moved. The other temptation is to disconnect the overload switch because it is "causing complaints". This is the same category of decision as bridging a door lock contact, and it should be treated the same way. A car that runs with the overload disabled is a car with a known defect, and the person who disabled it owns the consequences. A few practical notes from the field. Water ingress into a load cell junction box is the most common single fault I see, usually after a wash-down or a leak from above; the box should be above the car floor line and sealed. Loose load cell mounting bolts produce a reading that changes when the car is rocked by hand, which is a quick check worth doing on every visit. The wiring is where cheap faults live. The summing box on a four-cell platform takes four signals and adds them, and it is usually mounted under the car where it gets kicked, washed, and dripped on. Open it once a year, look for corrosion on the terminals, and check that the cable glands are tight and the cable entry is at the bottom, not the top. A cell with a broken wire reads zero, and the system simply ignores that corner; the car then weighs itself as lighter than it is, on one side, and the overload point moves. On rope or belt tension systems, the equivalent fault is a load sensor on a rope that is no longer sharing load. After a re-rope, a re-tension, or a sheave change the distribution is different, the reading is wrong, and nothing has failed. Recalibration is not an optional extra after rope work; it is part of the rope job. Keep the calibration sheet in the machine room, with the date, the weights used, and the readings at each step. When the car starts alarming at 95 percent two years later, that sheet tells you whether the sensor drifted or the building changed. Both are fixable, but only one of them is fixed with a screwdriver.