Fault Prediction of Process Industry Based on Fuzzy Clustering

EasyChair Preprint 7663

Abstract

In the process of process industry production, once the failure will often bring heavy losses to the enterprise and the country in terms of human resources and financial resources, so it is very important to give early warning of the failure and take corresponding solutions. In this paper, a real-time fault prediction algorithm based on incremental model update is proposed for the continuous growth of industrial data flow, real-time update of objects, complex and variable attributes, and value attenuation with time. The algorithm consists of several steps, including data preprocessing, data flow clustering based on sliding window, abnormal data judgment, fault prediction and model prediction update. In each stage of the algorithm, Spark framework is used for parallel acceleration, so as to improve the efficiency of fault prediction.

Keyphrases: Data stream clustering, abnormal data judgment, data preprocessing, fault prediction, model prediction update

@booklet{EasyChair:7663,
  author    = {Tao Mi and Luoyifan Zhong and Xin He and Yi Sun and Chenyang Yan and Hao Yue},
  title     = {Fault Prediction of Process Industry Based on Fuzzy Clustering},
  howpublished = {EasyChair Preprint 7663},
  year      = {EasyChair, 2022}}