Research On Data-driven Degradation Assessment Of A Class Of Actuators And Self-healing Method Of Control System

In the process industry,under the combined influence of harsh working environment,frequent changes of working conditions,aging of materials and structures,etc.,the actuator will inevitably deteriorate continuously.This interferes with the adjustment of the working fluid,and ultimately makes the plant deviate from the expected target performance.At present,there are still many problems to be solved in the research on the development process of actuator degradation in control systems.In this paper,the related research work of the actuator degradation process is carried out from the following four aspects:abnormal state monitoring and diagnosis,degradation performance evaluation,remaining useful life prediction and control system self-healing.The goal is to realize the following functions from the normal operation of the actuator:monitor abnormal occurrence and abnormal mode,follow up and evaluate the development of degradation performance,accurately predict the failure time,and finally achieve the purpose of self-healing of the control system including the degraded actuator.The main research results are as follows:(1)A method for detecting and diagnosing abnormal state of actuators is designed respectively.On the one hand,in order to solve the problem of low anomaly detection accuracy in the full operating mode,an improved non-negative matrix factorization algorithm is proposed for both steady-state and dynamic operating conditions.Combined with the constructed actuator model,T2 and SPE statistics and their control limits are calculated to detect the occurrence of anomalies.On the other hand,for the problem that the early abnormal state is masked,the semi-non-negative matrix factorization algorithm is used to extract the eigenvectors of different abnormal data.They are combined with each other to build a binary classification network model library.Based on the static distances output by the model,the model library is updated.At the same time,the fusion of similarity measure and static distance is introduced to form an anomaly discriminating index to realize the diagnosis of anomalous patterns.The comparative analysis of the simulation experiment and the water tank experiment fully proves the high efficiency of the method.(2)A performance evaluation scheme of degraded actuators based on operating data is proposed,so as to realize the qualitative and quantitative evaluation of actuator degradation process.First,by introducing a reference signal,the traditional refractal analysis method based on statistical moment function is improved,which effectively overcomes the problem that amplitude information is ignored.Then,based on this algorithm,three performance indicators are designed from the aspects of refractality,risk and effectiveness,and the quantification and standardization of the indicators are realized by using the constructed membership function.Finally,according to the index results,an evaluation system was established,and the degradation degree of the actuator was qualitatively divided into five grades:"excellent","good","warning","medium" and "poor".At the same time,in order to measure the dynamic degradation process of the actuator,a comprehensive quantitative degradation index is obtained through fusion.The effectiveness of the method is proved based on the field data of the desuperheating water control valve of thermal power units.(3)A Wiener process-based spatial distribution prediction model of actuator remaining useful life is proposed,which aims to solve the problem that current research ignores the degradation difference of actuators under different command openings.The biggest innovation of this model is that on the basis of the traditional Wiener degradation model,the actuator’s opening degradation function is added.Therefore,it can be improved into a binary function of actuator degradation and time-space.And the correction coefficient is introduced to realize the self-adaptation of the model difference.The model parameters are obtained by the method of maximum likelihood estimation.The online update of parameters is realized by Kalman filter.Finally,the probability density distribution function of the remaining useful life of the actuator is obtained to realize the life prediction.The model is used in open-loop and closed-loop simulation experiments based on DAMADICS.Compared with the traditional algorithm,the accuracy and superiority of the method in this paper are fully proved.(4)A model free adaptive fault tolerant control scheme is proposed to realize self-healing control of control systems with degraded actuators.First,based on the compact-format dynamic linearized data model,the partial failure fault and bias fault models of the actuator are fused.Then,according to the idea of adding a compensation signal at the desired input,the control law is re-derived to realize the function that the control system can still maintain the control performance under the nominal state when the actuator is degraded.Rigorous theoretical analysis proves the stability of the designed control law.Finally,it is verified by numerical simulation experiments,model simulation experiments and semi-physical pneumatic valve experiments based on dSPACE.Many actuator degradation modes are set to test the performance of the control scheme.