Intelligent Fault Diagnosis Technology On Complex Process And Its Applications To Large Industrial Kilns

Nowadays, the existing fault diagnosis theories and technologies are facing the difficult challenge that is brought by the complexities of process control, so studying the fault technology of complex process has an important and practical meaning for ensuring the safe and stable running and improving economy of the large flow control industry. Hence, taking the large-scale industrial kilns as research subjects, this dissertation researches the application of intelligent fault diagnosis technology.Firstly, a framework of the intelligent integrated fault diagnosis modeling is proposed and constructed based on fuzzy logic, neural network and expert system: a Fuzzy Neural Network (FNN) diagnosis model is constructed for solving the automatic acquisition problem of fault symptoms for complex process, then an expert system is used to solve the fault diagnosis problems of complex process by the diagnostic results of FNN.Next, under the instruction of the intelligent integrated modeling framework, based on the artificial intelligent methods, the founding and updating of knowledge base is studied on emphases: (1) an integrated shallow-deep knowldege representation modole is proposed and constructed, with the combination of the deep knowledge based on the object mechanism knowledge and the shallow knowledge base on the heuristic rules, then the problems of the shallow knowledge imperfection and the deep knowledge searching space affecting diagnositic performance are solved. (2) knowledge acquisition of complex process is proposed to divide into complete data sets and incomplete data sets: (1) an complete data sets knowledge acquisition based on the combinatorial optimization algorithm is proposed, which can not only avoid the problems of the acquired knowledge with incompleteness and disagreement, but also conquer the problems of the single optimization algorithm with bad stability and inefficient searching. (2) an incomplete data sets knowledge acquisition based on the rough theory is proposed to solve the problem which the unknown eigenvalue of the fault state attribute can not be estimated. Two kinds of partitions are formed in the training examples of incomplete data sets: lower approximations and upper approximations, from which can simultaneously derive rules and estimate the missing values in the learning process. (3) A maintenancealgorithm based on the theory revision methods is proposed for solving the problem of acquired new knowledge as new patterns. According to the given process description, the problem of updating rules is come down to a kind of standard format, and the formalization is given. Then, this proposed method is used to fulfill updating rules of knowledge base.In the research of reasoning machine, a FNN reasoning method is used to solve the problem of collision and inefficiency in the fuzzy rules reasoning. The double direction alternate control strategy is applied to reduce the blindness of the target selected and approved for improving the reasoning efficiency. For enhancing the diagnostic capability of this modeling, the difficult faults are studied since the reasoning machine cannot be solved by the knowledge of knowledge base: (1) for the multi-fault diagnostic problem with faults and manifestations no matching, via using the degrees represent the probabilities which faults may directly cause manifestations, a new fuzzy abductive inference model is proposed based on the parsimonious set covering theory, so the problem of one criterion for describing the different faults can be solved. (2) for the novel fault diagnostic problem, a new fuzzy classifier is proposed based on the modified growing cell structure scatter partition. This algorithm is particularly efficient by unsupervised algorithm and supervised algorithm for solving to get a critical point in the implementation of the fuzzy classifier, so whose implement effect is improved. The fault isolation of the industrial coal powder boiler is successful by this method.Based on analyzing to the craftwork and faults mechanism of the...