Research On The Fault Diagnosis Method Based On Data Driven For Industrial Process

As the scale of modern complex industrial process system is growing,the equipment,system structure and complexity of industrial production process are also increasing,and problems such as external disturbance,nonlinearity and system uncertainty will also arise.If such complex system fails,the lighter will cause economic loss and the more serious will cause casualties.Therefore,certain fault diagnosis shall be taken for industrial system It is necessary to ensure the reliability and safety of industrial system.In the early stage of industrial process fault diagnosis,on the premise that the process mechanism of the system is known,the fault diagnosis of the system itself is realized by establishing the system model,and the effect is good.However,in the relatively complex industrial process system,it is particularly difficult to establish a more accurate mechanism model.Because the complex industrial process system contains huge process data,which are online data collection and offline data storage,the operation characteristics and laws of the industrial process system will be presented by these process data,so using appropriate methods to analyze the data is one of the feasible directions of fault diagnosis.The data-driven process fault diagnosis method,is mainly used in this paper,which can detect the fault accurately and quickly especially in the complex industrial process system.It can analyze and identify the fault,then classify,isolate and repair it to ensure the stable and safe operation of the system.Some methods of fault diagnosis are mainly focused in this paper,including the following aspects:(1)The fault model of complex industrial process system is diverse,dynamic and uncertain.In this paper,an improved dynamic fault diagnosis algorithm based on evidence updating is proposed and applied to the fault diagnosis system of butadiene production process with artificial intelligence.Through the analysis of the experimental diagnosis results,and compared with the original evidence theory and fuzzy reasoning method,the test results show that the improved evidence updating rules have better application advantages in the field of dynamic fault diagnosis of industrial production and play an important role.(2)Considering the large amount of data generated in industrial process system and its non-linear characteristics,an improved fast independent component analysis matching string fault diagnosis method(CICA)is proposed.Based on the improved fast ICA algorithm,the identified fault types are represented by strings,so that the problem of fault type identification can be transformed into a string matching problem.The string matching method is only driven by data and does not need to build any model.It is very convenient to apply.Finally,through experiments,the feasibility and validity of CICA method are proved.(3)In view of the non-linear,multi loop,strong coupling and complex operation of complex industrial process system,which lead to the difficulty of fault diagnosis of complex industrial process system,a new system fault diagnosis method(CICA-WSVM)based on independent component analysis and improved weighted support vector machine is proposed.CICA-WSVM integration algorithm can greatly improve the accuracy of fault diagnosis.Through the experiment and the analysis and comparison of the experimental results,it can be concluded that the integrated fault diagnosis method of CICA-WSVM is feasible and effective,and provides a new tool for industrial process fault diagnosis.