Degradation Failure Prediction from Coil Current Signals of Electromagnetic Valves in Coal Mining Based on Neural Network

Abstract

In hydraulic system of the roof supports of longwall coal mining face, electromagnetic valve translates the electrical control signals to hydraulic signals to realize the roof support, coal mining, and the advance of longwall face. Thus, it is important to identify current as well as future conditions of electromagnetic valves to avoid unexpected failure. Most of fault diagnosis and maintenance methods mainly focus on short or open circuit failure detection and regular replacement, detection, and repair. In contrast, this paper develops real-time fault detection and degradation failure prediction methods for intrinsically safe electromagnetic valve used in underground longwall coal mining based on coil current. The analytical relationship of the force, air gap, and coil driving current are analyzed according to the mechanical kinematical character of electromagnetic actuator in the valve, and an experimental setup is developed to perform accelerated electromagnetic valve tests while the coil current information is collected from two types of electromagnetic valves that are running until failure. This information is then used for real-time detection on validity of switching activity and for degradation feature extraction. The neural network is used for degradation failure classification and prediction. The results has shown that the discrimination of 5 levels of degradation failure for the two types of tested valves can reach 92.5% and 99.3% respectively, and that the developed method can be an effective assistant for maintenance of mining used electromagnetic valve.