基于k近邻变量贡献与重构理论的工业过程故障诊断

Fault diagnosis of industrial processes based on k nearest neighbor variable contribution and data reconstruction method

针对工业过程中发生故障时异常变量的精确识别以及如何准确建立"故障–征兆"表的问题,本文提出了一种基于k近邻(k–NN)变量贡献分析和数据重构的异常变量精确识别方法.首先,将k–NN算法中各个采样时刻的统计距离指标细化,分解为每个变量的贡献并对其进行详细分析,分别从单变量和多变量异常角度进行方法的可行性验证,确定过程故障时异常变量具有较大的贡献值;其次,建立正常数据中每个变量的贡献模型用于对故障样本中的异常变量进行"一次"识别;随后提出基于k–NN理论的数据重构算法,并从重构原理方面进行分析,验证该方法具有一定的有效性.对于故障样本,根据变量贡献分析方法求取每个变量对距离指标的贡献,"一次"识别出故障发生时所对应的异常变量或征兆;进而通过数据重构理论对故障样本中异常变量值进行重构、检测和"二次"识别,直至辨识出过程中发生异常的所有变量,并得到故障与变量之间的关系,即"故障–征兆"表.

In view of the problems of abnormal variables precise identification and how to establish "fault-sign"table when there are faults in industrial processes, a novel abnormal variable identification method based on k nearest neighbor(k–NN) contribution analysis and data reconstruction is proposed. Firstly, this method gives a detailed analysis of distance control indexes for each sample, assigns them to each variable, and establishes the contribution model of the normal variables. To ensure a greater contribution for abnormal variable, the feasibility of the method is verified from the angle of univariate and multivariate anomaly respectively;Secondly, for the normal data, the contribution of each variable is set up used to "first time" identify abnormal variables. And then data reconstruction method based on k–NN is proposed. The principle is analyzed, and the advantage of data reconstruction method is validated. For fault samples, the contribution of each variable in the distance control index is calculated according to the contribution analysis method firstly, which is used to identify the abnormal variables "first time". And then abnormal variables can be reconstructed, a new detection and"second time"identification processes should be implemented until all the abnormal variables are found. However, this method can ensure the accuracy of anomaly variable identification, and get the relationship between fault and variables,namely"fault-sign"table.