Tension prediction for the scraper chain through multi-sensor information fusion based on improved Dempster-Shafer evidence theory

Abstract

This paper deals with a tension prediction strategy for the scraper chain of the scraper conveyor based on multi-sensor information fusion combined with the improved D-S evidence theory (IDST). First, multiple sensors are used to obtain comprehensive information reflecting the working status of the scraper conveyor. Then, four independent artificial neural networks (ANNs) are adopted to perform the preliminary prediction of the tension variation based on the signals obtained by different sensors, and the weight coefficients of different ANNS are employed to construct the basic probability assignments (BPAs) of independent evidences. Finally, the IDST is proposed to fuse the conflicting evidences and utilize the complementary multi-sensor information for accurate tension prediction, thus providing practical guiding significance for state monitoring of the scraper chain. Experimental results indicate that the IDST has specific superiority over other conventional methods in dealing with conflicts of the evidences, as well as improving the accuracy of parameter prediction.