Study Of Fault Intelligent Monitoring And Diagnosis Methods For Air Supply System

As an important part of the train brake equipment,all kinds of faults may occur during the air supply system operation.However,the prevention,early warning as well as handling work for these problems are usually completed by manpower at this stage,causing that the timeliness and accuracy of fault monitoring and diagnosis cannot be guaranteed effectively,which invariably leads to varying degrees of economic loss.So the research of related methods which aims at reducing failure rates is of great theoretical significance and has important application value.Based on the facts above,this thesis has fulfiled the following jobs:1.Investigation on relevant contents of the development of fault monitoring and diagnosis for air supply system is done,and mechanism analysis according to the working principle along with the fault statistics is carried out.As a result,the common faults are classified into three categories:air leakage,high oil temperature as well as oil leakage,and several ways for fault avoidance are provided.2.Among these three faults,air leakage and high oil temperature are supposed to be the research emphases.The architectural structure of the data acquisition scheme consists of three parts:acquisition terminal,communication module and monitoring center.STM32 F107VCT6is used as the core controller of the acquisition terminal,with hardware and embedded software of associated circuits around it;adopting multiplexing I/O mode of select,GPRS/WiFi remote strategy based on TCP transfer pattern is selected for the communication module,and a protocol is customized for the application layer;in accordance with the established scheme,the air leakage sound and required temperature data are successfully collected and transmitted to the monitoring center.3.During the study of fault monitoring methods for air leakage,the sound features are analyzed in time and frequency domain.By combining time and frequency domain together with spectrogram,setting quantifiable indexes,as well as formulating principles of fault classification and systematics,intelligent monitoring and timely early warning are realized.At the same time,redundant workload is reduced,which brings about more efficient monitoring.4.During the exploration of fault diagnosis methods for air leakage,according to the treatment of underdetermined blind source separation problem,the diagnostic process is decomposed into two sub-problems:estimation of the source number and signal recovery.By using empirical mode decomposition,singular value decomposition and K-means clustering joint algorithm,the difficulty of accurately estimating the number of source signals can be solved when the sensors are inadequate;by applying fuzzy C-means clustering approach and minimizing the l₁ norm of source signals,the sparse component analysis is performed,with the numerical calculation of the mixing matrix being conducted and source signals being recovered as well.Ultimately the diagnostic task is completed.5.As for high oil temperature,the most significant point is to measure oil temperature precisely and punctually,so a soft sensing scheme based on RBF neural network is formulated.Several accessible auxiliary variables are used as input data,and their relationship with oil temperature is established by training the network.Two different modeling methods are carried out to conduct simulation experiments.The final result proves the feasibility and practicability of this solution,which also provides references for futher fault monitoring and diagnosis.