Machine Learning Approach to RFID Enabled Health Monitoring of Coal Mine Conveyor Belt

Abstract

The mining conveyor belt is an imperative component of the mine industry which serves the crucial role of transporting materials. Predicament such as conveyor belt cracks has to be avoided, because any potential cause of conveyor belt failure constitutes risk of fatalities and significant economic implications. Hence, a new crack detection method is developed which can duly alert prior to any inception of failure. The research incorporates an interdigital capacitor (IC) based Ultra High Frequency (UHF) radio frequency identification (RFID) tag as sensor and Machine Learning (ML) to monitor the formation of cracks. This is done by collecting data from the UHF sensor when it is placed on a conveyor belt. Two states of the conveyor belt are investigated -stationary and a travelling belt moving at speeds between 1-5 m/s. For the case of motion, it is successfully replicated with the help of an electrical treadmill. By using ML integration, the outcomes will be more effective in tracking crack features. The model is tested with different types of inputs, followed by classification based on experimental data. This method results in a very accurate determination of cracks, crack orientation and width. The experimental model is 97.2% accurate in identifying the presence of cracks and 93.9% accurate in detecting their orientation; it can also determine half a millimetre-wide crack width with 97% of accuracy. This detection system is extremely accurate, even when there is movement. The achieved detection rate of 99.4% depicts the effectiveness of the model performance by successfully combating the interferences caused by motion. This research is intended to be a stepping stone towards an efficient remote health monitoring (HM) system suitable for the smart future.