基于加权模糊聚类的污水处理过程故障检测  

Fault Detection in Sewage Treatment Process Based on Weighted Fuzzy Clustering Algorithm

作　　者：慈嘉伟 罗健旭[1] CI Jia-wei;LUO Jian-xu(School of Information Science and Engineering,East China University of Science and Technology,Shanghai 200237,Chin)

机构地区：[1]华东理工大学信息科学与工程学院,上海200237

出　　处：《华东理工大学学报（自然科学版）》2018年第4期504-510,共7页Journal of East China University of Science and Technology

摘　　要：模糊C均值(FCM)聚类是一种常用的聚类方法,在工业应用时,常因数据的强噪声和非线性导致聚类效果不够理想。提出了一种密度加权、核理论和可能性模糊C均值聚类(PFCM)相结合的聚类方法。该方法采用核函数,将数据映射到线性空间进行聚类分析,消除非线性影响;通过引入点密度概念,加快算法迭代,增强可分性,提高聚类准确率。将该聚类算法用于污水处理过程的故障检测,结果表明该方法不仅能解决非线性问题,而且能有效加快收敛速度。Fuzzy C means （FCM） clustering is a conventional clustering method. As an unsupervised learning method, FCM can make full use of historical data or real-time data, and detect and diagnose faults in process by establishing fuzzy similarity relation. However, in dealing with the industrial data, the clustering performance of FCM is lower due to strong noise and nolinear data. Sewage treatment process is a complex nonlinear industrial process and it is operated difficultly in long-term and stable operation. Therefore, it is quite necessary for monitoring sewage treatment process, detecting operational failures, and dealing with faults in time. This paper presents a fault detection method in sewage treatment process by combining desity weighted, kernel theory and possibility fuzzy C means （PFCM） clustering. In the proposed algorithm, the kernel function is utilized to map data into linear space for clustering analysis and eliminates the nonlinear influence. By introducing the concept of point density, the proposed algorithm iteration can be accelerated and the clustering accuracy is improved. Meanwhile, the possibility fuzzy C mean（PFCM） clustering is also maintained for the outlier robustness and the amount of calculation is reduced by introducing the sample variance. The simulation experiments are made via Benchmark Simulink Model-1（BSM1）, in which there exist two types of fault：process fault and sensor fault and 13 effluent substances in water from the fifth biochemical reaction pool are used as raw data. The simulation runs for fourteen days and only the data in the last seven days are selected as experimental data. By using different categories of fault data clustering analysis and comparing with the fuzzy C means （FCM） clustering, the possibility fuzzy C mean （PFCM） clustering and the kernel possibility fuzzy C means （KPFCM） clustering algorithm, the proposed detection method can not only deal with the nonlinear problem, but also accelerate the convergence speed effectively.

关 键 词：密度加权 模糊聚类 故障检测 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]