基于WPA与WLS-SVM方法的化工过程故障诊断  

Fault Diagnosis in Chemical Processes Based on WPA and WLS-SVM

作　　者：许贺楠[1] 添玉[1] 肖娟[1] 黄道[1] 

机构地区：[1]华东理工大学自动化研究中心,上海200237

出　　处：《控制工程》2010年第S2期193-197,共5页Control Engineering of China

基　　金：上海高校选拔培养优秀青年教师科研专项基金项目资助(SHS08012)

摘　　要：化工过程存在变量多,系统复杂,非线性等特点,这使得常规的故障诊断方法具有模型难以建立、参数难以调整、收敛速度慢、多故障无法正确识别等局限性。以标准化模型TE过程为实验平台,结合模型特点以及在故障诊断中的难点,采用小波包算法(WPA)滤除过程数据噪声,恢复原始信号,数据缩放统一数据量度,最小二乘支持向量机(WLS-SVM)为模型,K聚类方法确定权值系数,交叉验证来选择模型参数,提出了一系列具体的解决方案。通过仿真实验,验证了算法的有效性,以及在过程故障诊断中的可行性,并在最后提出了一些展望。Multi-variable,system-complex,non-linear are the characteristics of the fault diagnosis in chemical process.This makes the conventional methods difficult to establish the accurate model and adjust the parameters,make the speed of convergence slow,multiple faults can not be correctly identified.Based on the tennessee eastman process models,the model features and the difficulties in fault diagnosis are combined,the wavelet packet algorithm is adopted to filter noises to restore the original signal,and data scaling is used to unify the measurement.Weighted least squares support Vector machine(WLS-SVM) is taken as the model,the weight coefficients are determined by K-clustering method,to selected the model parameters are selected by cross-validation.A set of concrete solutions are proposed.Simulation experiments are carried out to show the validity of the algorithm and the feasibility in the fault diagnosis in the chemical process and a number of prospects are concluded.

关 键 词：故障诊断 TE过程 小波包分析 数据缩放 最小二乘加权支持向量机 K聚类 

分 类 号：TQ021.8[化学工程]