Enhancing the accuracy of conveyor belt weighing systems using Median and Exponential Moving Average filters in load cell signal processing

Abstract

The belt weighing system plays a vital role in industrial production lines such as mineral mining, thermal power plants, cement manufacturing, and bulk material processing. Its main function is to measure the mass of material being continuously conveyed, enabling businesses to monitor throughput, optimise production processes, and maintain the quality of the output products. This article presents a signal filtering method that combines the Median filter and the Exponential Moving Average (EMA) filter to improve the stability and accuracy of load cell signals in conveyor belt weighing systems. The system is implemented on an STM32H743ZIT microcontroller platform, featuring an integrated 24-bit ADC and the FreeRTOS real-time operating system. The signal is processed using a seven-point Median filter followed by an EMA filter with a smoothing factor α=0.3, prior to calibration and flow calculation. Experimental results from a real-world industrial conveyor line demonstrate that the proposed method reduces impulse noise by over 68%, decreases signal standard deviation by more than 62%, maintains latency at approximately 104 ms, and keeps mass measurement error below ±1.2%. These results verify the feasibility and high efficiency of the proposed approach in harsh industrial environments with significant noise interference.