Quantitative Association Rules Mining For Industrial Process Parameters And Faults Correlation Analysis

In recent years,there has been a trend towards automation,intelligentization and upsizing in industrial production systems,while the production process has become increasingly complex.In order to ensure that industrial processes operate in a safe and reliable way,it has been a necessity to utilize relevant technologies to find out and locate faults as well as the nature of faults,figure out the cause of faults and predict the development tendency of faults.For the purpose of ascertaining the key parameters characterizing the faults,it is necessary to analyze the correlation between the industrial process parameters and faults,find the correspondence between the parameters and faults,ascertain the key parameters characterizing the errors and describe the correlation in a form of fault rules,for deeply mining a large number of fault data in the historical monitoring data set.These fault rules can provide assistance for decision-makers,guide fault detection and location,and predict potential faults.In order to analyze the correlation between industrial process parameters and faults,we have done the following work:In allusion to the large amount of industrial process data with low utilization rate,and the situation that correlation analysis is mostly used on single equipment or specific objects,we put forward an algorithm of applying association rule mining method to the process of correlation analysis of industrial process parameters and faults.Firstly,the correlation analysis model is constructed.The characteristics of industrial process parameters and faults correlation analysis and the applicability problems in the implementation process are discussed,which is summarized as the problem of quantitative association rules mining and the corresponding mining model is proposed.Secondly,in order to solve the problem that only the single parameter is clustered separately and the detailed information of the partition interval cannot be obtained in advance,the ISODATA clustering is used to discretize the slowly varying parameters.In allusion to the original algorithm,which is sensitive to the initial cluster center and noise data,a density-based method is introduced,avoiding the contingency of obtaining the result of dividing the interval and transform the original fault data set into a data set that is easy to be mined.Finally,aiming at the applicability of the fault rules form in industrial process parameters and faults correlation analysis,a fault rule mining algorithm based on constrained optimization matrix is proposed.The matrix structure and mining process are adjusted and the constraint conditions are added to generate appropriate fault rules by obtaining the maximum frequent item sets with constraint conditions.These fault rules can better describe the correlation between industrial process parameters and faults.In general,we propose a quantitative association rule mining algorithm based on density ISODATA clustering and constrained optimization matrix(NISO-CIMC)to find the correlation between industrial process parameters and faults.The algorithm mines a large number of fault data in the historical data set of industrial process slowly varying parameters,and describes the correlation between industrial process parameters and faults as unified,intuitionistic,understandable,inheritable fault rules and verifies the validity and accuracy of the NISO-CIMC algorithm by simulation experiments based on TE process fault data set.