Research On Fault Prediction Of Electronic Products In Pneumatic Industry

As one of the components of pneumatic actuator,positioner is an important equipment to keep the stable operation of production.This paper takes the positioner as a typical object of electronic products in the pneumatic field to carry out fault prediction research.It is installed in the production site and closely related to the production technology.Once the positioner has a problem,it will cause the control circuit to oscillate and cause problems,such as product quality degradation,increased equipment loss,environmental pollution,and waste of energy consumption.However,the current maintenance of positioner mainly adopts post maintenance and regular maintenance,which will bring a series of problems: post maintenance implies a large potential safety hazard;regular maintenance often causes blind repair and increases the maintenance cost.In order to ensure that the positioner can operate reliably,stably,and efficiently,and reduce maintenance costs,it is of great significance to grasp the operation information of the positioner in real time so that potential hidden troubles can be found in time.In this paper,the fault diagnosis and state parameter prediction of positioner are studied.The contents are divided into four parts shown as follows.(1)The main structure and working principle of the positioner are explained,the common fault types of the positioner and their causes are analyzed,and an experimental platform is built to obtain the data of normal and eight common faults.(2)For the various fault data of the positioner,the missing data was interpolated,and the purity of the data was improved by wavelet noise reduction.The normalization method was used to solve the problem of inconsistent scale size among features,which speeds up the convergence of the algorithm.(3)Random forest(RF)model and Gradient Boosting DecisionTree(GBDT)model were built for the fault diagnosis research of positioner.Macro recall,macro precision,and macro F1 are selected as indicators to evaluate the performance of the models.Macro precision rate,macro recall rate and macro F1 are selected as the indexes to evaluate the performance of the model.Comparing the results of the two models,it is found that the two models have certain accuracy.The time cost of RF model is lower than gbdt model,and the difference of model accuracy is less than 0.2%.(4)Aiming at the research of fault prediction,this paper uses long short-term memory(LSTM)algorithm to establish the state parameter prediction model of positioner.Samples were constructed by rolling prediction methods,and eight types of prediction models were established.The results show that the LSTM has achieved good prediction results for the positioner state parameters.