Research On Key Problems Of Multi-opening Process Fault Diagnosis Of Flow Control Valve

The flow control valve is a kind of rigid facility used widely in industrial production and people’s lives.Because of its working environment and structural characteristics,the flow control valve can generate massive environmental pollution and waste of resources when faults occur.The flow control valve usually can not be in the same condition during the work,which brings great trouble to the fault diagnosis.Moreover,a variety of reasons in the actual industrial production cause data acquisition problems.Based on the above problems,this paper studies the fault diagnosis of the multi-condition process of the flow control valve system and gives the solution.The primary research is as follow:Firstly,based on the operation characteristics of flow control valve,the multi-opening process is analyzed by multi-sub-model.On this basis,clustering analysis will be used to define the working interval of each sub-model in order to determine the appropriate number and scope of process division.Secondly,based on the modeling idea of data-driven model,the fault diagnosis model of sub-working interval will be modeled and analyzed.In this work,the support vector machine,the convolutional neural network and the Resnet will be used to diagnose the multi-opening system of the flow control valve.Thirdly,for the problem of missing data,imputation method will be used to process the data with missing values in order to facilitate the successful modeling of data-driven model.The generative adversarial imputation network(GAIN)based on generative adversarial neural network will be used to complete the missing values according to the distribution characteristics of observable data.At the same time,the K-Nearest Neighbor(KNN)imputation method will be used as a comparison to verify the imputation effect generated against the GAIN.Fourthly,the experimental hardware environment and related software environment will be built to meet the relevant experimental basis.The experimental platform will not only realize the real-time remote control and fault diagnosis of valves,but also have the functions of visualization of valve working status and historical data traceability.In this work,according to the experimental conditions,the relevant experiments of process division and model membership degree definition with clustering algorithm as the core,the verification of data imputation,and the actual verification of fault diagnosi s algorithm for multiopening process will be completed..By analyzing the results,the optimal parameters of clustering algorithm and the fault diagnosis algorithm suitable for multi-opening process of flow control valve will be determined.