Study on Perceptive Fuzzy Petri Net-based Autoloader Fault Analysis

Abstract

To address the problems of high incidence of faults in tank autoloaders, long diagnosis cycle and low accuracy of diagnosis, this paper proposed a perceptive fuzzy Petri net-based fault diagnosis method on the basis of relevant expertise. The corresponding NFPN failure model was established according to the specific structure of the autoloader, fuzzy Petri net was used to present the process of fault propagation, the perceptron error back propagation method was adopted to learn the limited expertise, and the values of arc weights of trigger accidents in the Petri net were determined. An accurate judgment on autoloader faults was achieved by way of forwarding reasoning. At the time of backward reasoning, the minimal cut set method was also adopted to narrow the troubleshooting scope, thus improving the reasoning efficiency. By taking an autoloader with a rotary failure as an example, this paper established the corresponding PFPN fault model and made a comparison with the fault tree seasoning method and the historical statistic data. The comparison results reveal that this method can realize a quick and high-efficiency fault diagnosis of autoloaders thanks to its higher reliability and accuracy compared with the traditional fault tree diagnosis method.