基于CART算法的PVC压延设备状态诊断方法研究  

Research on Condition Diagnosis Method of PVC Calendaring Equipment Based on Cart Algorithm

作　　者：王美林[1] 刘金刚 Wang Meilin;Liu Jingang(College of Information Engineering,Guangdong University of Technology,Guangzhou 510006)

机构地区：[1]广东工业大学信息工程学院,广州510006

出　　处：《现代计算机》2021年第23期53-58,共6页Modern Computer

基　　金：国家自然科学基金(U1701266);广东省科技计划(2019A050513011、2017B090901056);广州市科技计划(202002030386)。

摘　　要：PVC压延设备保持正常工作状态是整个生产过程能否顺利进行重要前提。针对现有的PVC压延设备运行状态的诊断需求,本文提出了一种基于CART算法的工业设备状态诊断方法,该方法通过计算基尼系数建立诊断模型。同时,针对单CART树容易出现过拟合现象且分类精度不高的问题,引入提升(Boosting)方法。通过构建多个CART树并进行加性组合,提高诊断模型分类精度。通过实例测试,加性组合后的诊断模型在工业设备故障识别上正确率为88.37%,AUC值为0.90,判决速度为0.15 ms/条,可以应用于PVC压延设备生产状态的实时诊断。Maintaining the normal working state of PVC rolling equipment is an important prerequisite for the smooth progress of the whole production process.In view of the existing diagnostic requirements of PVC rolling equipment running state,this paper proposes an industrial equipment state diagnosis method based on CART algorithm,which establishes a diagnosis model by calculating the Gini coefficient.At the same time,aiming at the problem that single CART tree is prone to over-fitting and the classification accuracy is not high,the Boosting method is introduced.By constructing multiple CART trees and additive combination,the classification accuracy of the diagnosis model is improved.Through the case test,the diagnostic model after additive combination has the correct rate of 88.37%and the decision speed of 0.15 ms/bar in the fault identification of industrial equipment,which can be applied to the real-time diagno⁃sis of the production status of PVC rolling equipment.

关 键 词：PVC压延设备 CART算法 提升(Boosting)方法 实时诊断 

分 类 号：TQ325.3[化学工程—合成树脂塑料工业]