Health Monitoring And Diagnosis Of Dynamic Processes Based On System Identification

Health monitoring and fault diagnosis for industrial systems and engineering equipment are essential to ensure safe production and normal operation of the system.At present,most of the research methods in the field of fault diagnosis are data-driven methods or mechanism-based methods.In order to overcome the difficulties of the above methods in practical industrial applications,this paper proposes a method of dynamic process health monitoring and diagnosis based on system identification.The system is identified by precise experimental design,and the multi-variable system identification method is used to obtain the normal operation model and real-time model of the system.Through the comparative analysis of the fault-free model and the real-time model,the real-time detection of the health of the dynamic system can be realized;if a safety hazard occurs,the fault location and diagnosis can be further realized.The method aims to identify fault early warning and safety monitoring before system failure and safety problems through multiple identification of the system and the background knowledge of the industrial system,and around the use of system identification methods for industrial systems and engineering equipment.An in-depth study was conducted on the issues of fault diagnosis and health monitoring.The main research results include the following aspects:1.1.For system health monitoring and diagnosis,we must first propose a system identification method for fault diagnosis based on the general system identification theory.First of all,in the experimental design,it is necessary to consider different experimental objects and fault conditions.When further researching the transfer function of some models affecting the fault occurrence,the signal excitation of the fault sensitive frequency band can be strengthened;when the accuracy of the identification model is not enough to identify a certain fault,the experimental signal should be redesigned to further achieve the identification accuracy.Objective:In the step of precision test,a method of accuracy evaluation of"0.5 precision method"is proposed,that is,the damage caused by a certain fault to the system parameters is greater than twice the estimation accuracy of the model,that is,the accuracy of the estimation model is proposed.Minimum requirements.2.The fault diagnosis method is to first define a fault,then perform a system identification experiment on the system under test through different time,discriminate and analyze the model transfer function obtained at different times,and obtain the evaluation of the health status of the system through the parameter information,in conclusion.For the above proposed method,two different simulation objects are selected to perform fault detection and diagnosis based on system identification.One is the single-input single-output(SISO)double-tank series model.The simulation of the model verifies the correctness of the fault detection using the model discriminant method.The other is the multi-variable atmospheric and vacuum unit,which is simulated by this platform.The theoretical part is supplemented,and the correctness verification of the previously proposed method is given by an example.3.Based on a real robot flexible manipulator experimental platform,the theory is further verified,and the series elastic actuator(SEA)system of the joint part of the manipulator is selected for model identification and state analysis.By designing two different common faults that occur on the internal torsion spring of the system,the identification model is used to identify the experimental platform under several working conditions,and the experimental results are compared and analyzed.The correctness and effectiveness of the method is verified..