Research On Process Monitoring Of Process Industry Based On Data-Driven Technology

With the development of the big data technology and manufacturing technology,and the concept of Internet + was promoted vigorously,the integration degree of modern industrial process systems is more and more high,and the systems show high intelligence and automation,For such complex systems,because of relying on the accurate physical model,the traditional monitoring methods of faults perform not good.As the development of data technology,and the widespread use of intelligent instruments,a lot of high quality data is recorded and saved by data server.Fault diagnosis methods based on data driven technology no need to build complex system accurate physical model,only need to analyze process data,because of its advantages,the technology has been attracted widely attention.First,this paper introduced the traditional principal component analysis method,and Tennessee Eastman process was studied in detail,taking this simulation process as a verification platform.To study the simulation of the traditional principal component analysis methods in the Tennessee Eastman process.And aim at the nonlinear and dynamic characteristics problems,respectively to study the kernel principal component analysis method and dynamic principal component analysis method,compare the missing rates and the false rate of the three methods in the diagnosis of Tennessee Eastman process.Secondly,in order to apply the data-driven technology in industry field,with the Distillation Column as one of the controlled objects,the hardware-in-the-loop simulation system was developed,and the data driven methods were applied to the process industry hardware-in-the-loop simulation system.And selected a few kinds of typical faults in the process of Distillation Column to study,realized the research and application of fault diagnosis from pure simulation objects to hardware-in-the-loop simulation objects,shorten the distance of laboratory with the actual factory.Finally,aiming at nonlinear and dynamic characteristics of industrial production processes,to sudy the dynamic kernel principal component analysis(DKPCA),this algorithm considering both nonlinear and dynamic characteristics in industrial the process.Aiming at huge computation and lower efficiency of the DKPCA,a new improved DKPCA is proposed,this method will remove the irrelevant variables or low relevant variables,reduce the amount of data,and improve the diagnostic efficiency by indiscernibility and the cross-degree.For the typical faults of systems,compared with the traditional DKPCA,the simulation results show that the proposed method is more reliable,lower missing rate,and lower false rate,in addition,it can detect the small processes faults timely.