Research On Clustering Algorithms And Its Application Of The Fault Diagnosis Of Wastewater Treatment Process

Nowadays,the main intelligent technology used for fault diagnosis in wasterwater treatment plant'in our country is knowledge-based expert system,one of the difficulties inherent to the development of a knowledge-based system is to obtain the knowledge base. The establishment of the knowledge base is mainly due to the few operators'or experts'experience, so it causes knowledge acquisition bottleneck problem. Clustering analysis technology is one kind of non-surveillance classification technology which good at analyzing the data which have few prior knowledge, therefore we proposed using clustering analysis technology to analyze the wasterwater treatment's history data, through analyzing the clustering results, to generate a set of inference rules for supplementing the knowledge base of fault diagnosis system of wasterwater treatment process.The main contributions of this dissertation are summarized as fellow:①After analyzing some elementary knowledge of clustering technology and genetic algorithm, we discussed the coding mode and operation in the genetic algorithm and also analyzed the choice of control parameters.②This paper is engaged in the hybrid algorithm of K-means algorithm and genetic algorithm. We presented an improved cluster algorithm based on genetic algorithm. It can enhance convergence speed and solve clustering problems.③In order to implement clustering under the condition that the number of clusters is not known a priori, a novel clustering algorithm which is based on the nearest neighbor clustering and genetic algorithm is proposed in this paper. We take some data sets to test this algorithm and the clustering result is analyzed also. Experiments show that this algorithm is effective for clustering when the number of clusters is not known a priori. A two-stage clustering algorithm based on the nearest neighbor clustering is presented in this paper on basis of the prior research which could be used for partitioning clustering or hierarchical clustering.④We used a novel abnormal detection algorithm based on the nearest neighbor clustering and genetic algorithm to analyze the wastewater treatment history dataset, then used outlier measure of selection the top-n abnormal item from dataset based on distance sum. After that analyzed the abnormal sample with expert's interpretation and discussed the establishment of fault diagnosis rule of fault diagnosis system. The finding shows that the algorithms based on genetic algorithm can be effective for clustering . when used the abnormal detection algorithm to the wastewater treatment plant's history data, it can be found the abnormal samples successfully, and then used the results to built the fault diagnosis rule which it is meaningful to the establishment of the knowledge base of fault diagnosis system of wasterwater treatment process.