Research And Development Of Fault Diagnosis System For Textile Hot Rolling Mill Based On Text Classification

Fault diagnosis is the key to providing after-sales service.How to quickly and accurately identify the cause of the failure,reducing downtime and making rapid maintenance become the most important point to enhancing the core competitiveness of enterprises.In this paper,the hot rolling mill for textiles is studied.The failure of textiles hot rolling mill is multiple and ancealed.It is difficult to troubleshoot,the maintenance relies on the experience of workers.In view of the above problems,this paper design and research the fault diagnosis system of textiles hot rolling mill.The main research contents are as follows:(1)The fuzzy algorithm combined with the weakest t-norm has applied to fault tree optimization.A fault tree model of hot rolling mills has established;the scores of experts and fuzzy functions have been used to decide the failure probability of basic event;the weakest t-norm algorithm has been adopted to reduce the fuzzy accumulation in the process of calculation;deblurring method has been performed to obtain the failure probability of each basic event.(2)Chinese short text feature extraction technology has been Studied.According to the optimized N-Gram model,segmentation processing and vector expression of the hot rolling mill's fault text has been performed;the feature of fault text is extracted by the Term Frequency–Inverse Document Frequency method and the dimensionality reduction processing have been performed as well;Word2vec model has also been used to extracte text feature vector and map high-dimension vector to low-dimension.(3)For the hot rolling mill model with a small number of fault samples,the support vector machine algorithm has been used to establish a fault classification model for the fault description text,finally the classification prediction accuracy is 71.17%.For the hot rolling mill models with a large number of fault samples,the convolutional neural network has been used to establish the fault text classification model,the classification prediction accuracy is 89.67%.(4)Based on Browser/Server,a hot rolling mill diagnostic system has been developed.After analyzed enterprise's needs,the goals of system,role permissions and database structures have all been designed.The system establishes the management process of the enterprise maintenance of hot rolling mill and realizes the intelligent recommendation of the fault diagnosis case.