基于数据驱动的工业生产过程故障预测研究现状与方法分析

Research Status and Method Analysis of Industrial Process Fault Prediction Based on Data Driven

本文介绍了工业生产过程系统中基于数据驱动的故障预测研究现状和常见方法分类,包括基于模型的预测、基于数据驱动的预测以及基于统计学的预测。然后对基于数据驱动的几种方法进行对比分析,包括线性回归、KNN最近邻算法、决策树、GBDT、XGBoost、随机森林、SVM等,并结合应用案例比较了不同算法在故障预测的精度和时间方面的区别。

This paper introduced the research status and classification of fault prediction methods, including model-based predictions, data-driven predictions, and statistical-based predictions. Then it is based on the data-driven method, and briefly introduces the advantages and disadvantages of linear regression, KNN nearest neighbor algorithm, decision tree, GBDT, XGBoost, random forest, SVM and other algorithms. Finally, combined with the application case to analyze and compare the difference between the accuracy and time of different algorithms in fault prediction.