| "Made in China 2025" proposes to accelerate the development of the manufacturing industry,continuously promote the improvement of intelligent manufacturing,and promote the development of a modern industrial system.As an important component of the industrial system,the pneumatic control valve plays an important role in the process of medium transportation.Traditional valve fault diagnosis requires technicians to disassemble the control valve and make corresponding judgments,which is very cumbersome.Currently,data-driven online methods are commonly used to diagnose control valve faults.Building an online fault diagnosis and performance evaluation system for control valves can effectively avoid the tedious maintenance process and quickly locate faults while completing the performance evaluation of the control valve.Therefore,the combination of computer-aided diagnosis system and lower computer can more directly and effectively complete the maintenance of control valves.This thesis focuses on pneumatic control valves as the research object.Using a data simulation platform to generate data-driven models,a fault diagnosis model is constructed and accurately identified for all types of faults.In addition,different types of experiments are designed based on a certain company’s pneumatic control valve to complete the detection of the control valve’s operating status and performance evaluation.This achieves an intelligent valve fault diagnosis system,providing technical personnel with auxiliary diagnostic assistance for maintaining control valves.This thesis focuses on pneumatic control valves as the research object.Using a data simulation platform to generate data as a driver,a fault diagnosis model is constructed and precise identification is carried out on all types of faults.At the same time,different types of experiments are designed based on a DN50 single-seat straight-through pneumatic valve to monitor the operational status and performance evaluation of the control valve.Finally,an intelligent valve fault diagnosis system is implemented based on this.The main research content of this thesis is as follows:Introduce the SPS-VFD algorithm for accurate identification of all types of faults in pneumatic control valves.This study focuses on the mechanism of fault diagnosis models for pneumatic valves.By using the benchmark platform DAMADICS,multidimensional sensor data of a regulating valve without faults and with 19 different faults were generated.Through analysis and research on fault diagnosis methods based on this platform,it was found that there were problems such as incomplete identification of fault types,difficulty in distinguishing fault data due to high similarity,poor feature extraction ability of fault data,and poor fault recognition performance.Therefore,this thesis proposes the SPSVFD(Sliding Window-Particle Swarm Optimization-Stacking-Valve Fault Diagnosis)algorithm to address these issues.The specific process is as follows: firstly,a time series processing method is introduced to partition the fault data on the time axis by exploring the optimal window size and sliding interval of the sliding window,which enhances the features in the continuous subsequence while learning the features of the entire time dimension.Secondly,after the data processing by sliding window,five classifiers that perform well on all fault data are selected from ten base classifiers and optimized using particle swarm optimization algorithm.Finally,the stacking fusion model is introduced to improve the model’s generalization performance,as a single model cannot fully learn the different distributions of fault data.Experimental results show that the proposed algorithm achieves an accuracy of 98% and a fault detection rate(FDR)of 99% on all types of faults,which is better than other compared algorithms.Implementing valve fault diagnosis and performance evaluation system.Using the DN50 single-seat straight-through pneumatic valve and the second-generation intelligent positioner as the experimental platform,a intelligen valve fault diagnosis system is designed and implemented on the Windows Form platform using C# and Python programming languages,based on a Client-Server(C/S)architecture.The system is divided into five subsystems: real-time monitoring,performance evaluation,fault diagnosis,communication management,and data management.Specifically,real-time monitoring completes sensor monitoring,fault alarm,and forced control.Performance evaluation assesses the regulating valve performance through four experiments: selfcheck,step response,pressure travel,and pressure supply.The SPS-VFD fault diagnosis algorithm is embedded in the fault diagnosis subsystem,which uploads continuous sensor data to give a diagnosis result.Communication management supports communication between the upper and lower computers,encapsulates and implements all HART commands,and performs abnormal communication detection.The data management subsystem assists in system management,which first realizes parameter management and settings for different components of the regulating valve.Then,it statistics the specific information about the operation of the regulating valve,manages and protects the experimental data.Finally,a complete diagnostic report is generated.In addition,the system is tested on medium and small regulating valves. |
