Research On Fault Detection And Diagnosis Of Actuator In Control System

Actuator is an important control component in the industrial process system,and has an important influence on the control system.Among them,electric actuators have high control precision,fast action and intelligence,and are widely used in various industrial fields.However,due to the complex and compact structure of its components and poor working environment,it is difficult to detect and diagnose faults.Therefore,the research on fault detection and diagnosis methods for electric actuators has important research value.At the same time,with the development of computer technology and DCS,a large amount of operating data is collected and stored.In order to effectively utilize the fault data and improve the reliability of equipment operation,this paper conducts research on the fault detection and diagnosis method of electric actuators based on data-driven.Data-driven fault detection and diagnosis methods are widely used in actuators and are very mature at present,but still face some challenges.First of all,electric actuators have many feedback signals and a large amount of operating data,but few fault samples can be obtained,making it difficult to extract fault features effectively.Secondly,the fault diagnosis methods based on machine learning and deep learning require a large amount of data training for the diagnosis model in the multi-class identification problem,and the parameter adjustment is complicated,resulting in low model efficiency.Based on this,this paper studies the data-driven fault detection and diagnosis methods for electric actuators.The main research contents are as follows:(1)Based on the fault analysis of electric actuators in the control system,a fault detection method for electric actuators based on spiking neural network is proposed.The original signal is filtered by local mean decomposition to reduce the influence of noise.The fault features are obtained through dimensionless index calculation,and the fault features are converted into a pulse sequence that can be recognized by the spiking neurons using group coding.Considering the multi-classification problem and neuron model,the relative ordering learning algorithm is used to optimize the constructed spiking neural network,which improves the detection accuracy effectively.Comparing with support vector machine and BP neural network algorithm,the results show that this method has better detection accuracy in each fault category.(2)Aiming at the problem that the training time of the spiking neural network under big data is too long,and it cannot meet the problem of rapid diagnosis under multiple working conditions,a fault diagnosis method for electric actuators based on improved deep forest is proposed.The fault features of the collected signals are extracted by the time-frequency domain feature extraction method,and multi-level features are extracted by multi-granularity scanning.Combining D-S evidence theory with cascading forests overcomes the feature redundancy problem encountered in the process of expanding cascading forests.The parameters of the electric actuator fault diagnosis model are obtained through hyperparameter experiments,and the algorithm is compared with support vector machines and one-dimensional convolutional neural networks.The results show that the method has advantages in evaluation indicators,running time and parameter robustness.The accuracy and effectiveness of the proposed diagnostic algorithm.