| With the development of information technology,the industrial production process is becoming more and more large-scale,intelligent and integrated,and fault diagnosis is playing a more and more important role.Data-driven fault diagnosis technology based on historical fault data to achieve fault diagnosis has become an important research direction.However,there are some problems in fault diagnosis,such as high dimensional nonlinear data and multiple types of faults.Therefore,based on the data-driven method,this paper studies the process multi-type fault diagnosis algorithm,and the specific work is as follows:(1)A SVM fault diagnosis algorithm based on EEMD sample entropy reconstruction is proposed to solve the problem that it is difficult to mine fault features and affect the accuracy of fault diagnosis.Firstly,an improved gray Wolf optimization algorithm was proposed,which initialized the population by chaotic sequence to increase sample diversity.The local search strategy was added to the population update position to select the optimal position,which effectively reduced the possibility of the model falling into the local optimal position.Nine test functions were compared with the original gray Wolf optimization algorithm,and the results show that the proposed algorithm is effective.Then,a fault diagnosis method based on SVM was proposed to reconstruct the data features based on sample entropy.The integrated empirical mode decomposition was used to decompose the process data,and a series of eigenmode components were obtained.The modal components are screened by sample entropy theory to form characteristic data which is helpful to distinguish faults.Finally,support vector machine classifier is used for fault diagnosis.Experiments with five kinds of fault data in TE process show the effectiveness of the proposed method.(2)In view of the difference between different types of fault data,a fault diagnosis method based on the combination of boundary discriminant projection and support vector machine is proposed to solve the problem of high dimension of fault data and multiple working conditions.Firstly,an improved sparrow optimization algorithm is proposed.The population was initialized by chaotic sequence,and the firefly search strategy was added in the position update of sparrows to ensure the accuracy of searching and improve the speed of searching.The proposed algorithm is compared with particle swarm optimization,whale optimization and Sparrow optimization.The results show that the improved algorithm has better performance in global search and local search.Then a fault diagnosis method based on boundary discriminant projection combined with support vector machine is proposed.By labeling the historical fault data,the discriminant boundary projection method is used to increase the distance between different types of faults and reduce the distance between similar faults,so as to obtain the discriminant boundary which is favorable for classification.Finally,the support vector machine is used to classify faults,so as to realize fault diagnosis.The TE process experiments show that the method presented in this paper can effectively discriminate multi-type faults with high dimensional nonlinear. |
