Fault Diagnosis And Prognosis Of Air Brake Systems Of High-Speed Trains

The way of using the air brake system to safely stop high-speed trains is still in use today.During the service period,performance degradation or even functional faults of the components of air brake system will inevitably occur.However,there are the following aspects to be improved in the fault diagnosis and prognosis of air brake system of high-speed trains.First,the sensor layout of air brake system considers the control requirements more,and the range of state monitoring is limited.Secondly,most of the threshold logic judgment methods currently used are insufficient in real-time diagnosis of latent faults within the threshold.Thirdly,the research on the performance state prognosis of air brake system and its components is not systematic enough.Therefore,the faults of air brake system of high-speed trains were analyzed qualitatively,and the fault diagnosis and prognosis methods of air brake system from the three levels of system-subsystem-component were studied quantitatively in this thesis.The main research contents were summarized as follows:First,fault analysis and condition monitoring sensor layout of air brake system of high-speed trains.The functional principle of air brake system of high-speed trains was expounded,and the faults of air brake system were analyzed by using the Failure Mode Effect Analysis(FMEA)method.The Signed Directed Graph(SDG)model of air brake system was established,and a condition monitoring sensor layout method was proposed based on the SDG model.The case analysis results of a typical air brake system show that,for the 36 possible faults,the existing sensor set can observe 22 faults,and the monitoring coverage rate of optimized sensor set is increased by 38.9%.By adding corresponding sensors to the ground combination test bench of high-speed train brake system and collecting test data of typical working conditions,the data foundation has been laid for the research and verification of fault diagnosis and prognosis methods for air brake system of high-speed trains.Second,fault detection and diagnosis of the pneumatic units of air brake system of high-speed trains.Aiming at the problem that the non-functional faults of the pneumatic unit cannot be detected in real time and accurately diagnosed,the working pressure signal of the pneumatic unit is affected by variable working conditions,and the fault samples are limited,a univariate fault detection and diagnosis method based on Kalman Filter(KF),Sequential Probability Ratio Test(SPRT)and Support Vector Classification(SVC)was proposed.The KF state estimation function was used to identify the fault inflection point of the output signal of the pneumatic unit,the innovation sequence after the fault inflection point was reconstructed,and the SPRT method was used to detect the faults in real time.In order to further determine the fault type,based on the dynamic response and steady-state characteristics of air brake system,the characteristic parameters of the pneumatic unit were extracted to form the fault feature vector.The "one-vs-one" strategy was used to extend the SVC method to realize multi-type fault diagnosis.In view of the similar failure causes of different pneumatic units,a key pneumatic unit—relay valve was taken as an example to carry out a physical fault injection test.The test data analysis results show that the proposed method can detect the non-functional faults of the pneumatic unit in real time,and can accurately distinguish the normal,internal and external leakage faults in the case of small samples,which validates the effectiveness of the method.Third,fault detection and diagnosis of the execution unit of air brake system of high-speed trains.Aiming at the problem that the fault detection threshold is not easy to determine due to the non-stationarity of the brake cylinder pressure signal of the execution unit,a multivariate fault detection and diagnosis method based on Mutual Residual(MR),Principal Component Analysis(PCA)and Improved Reconstructionbased Contribution Plots(IRBCP)was proposed.The MRs of the four-axle brake cylinder pressure signals of the execution unit were extracted to construct the fault feature vector.The PCA monitoring model and the statistical threshold were constructed by using the normal four-axle brake cylinder pressure data of the execution unit.The comprehensive statistics was calculated and compared with the statistical threshold to detect faults in real time.In order to further determine the fault location,the reconstruction contribution rates of each variable at all fault times were calculated,and the identification of multiple variables and the location of the fault were carried out based on the IRBCP algorithm.The typical failure mode fault injection tests of the execution unit were carried out.The analysis results of the test data show that the leakage,offset and drift faults can be detected in time after the fault occurs,and the fault axle can be accurately located,which verifies the validity of the method.Fourth,performance state prognosis of air brake system of high-speed trains based on data augmentation.Aiming at the problem of the lack of prognosis methods for the performance state of the air brake system due to the complex structure of the air brake system and the time-varying nonlinearity of the working signal,a performance state prognosis method of air brake system was studied based on the black box theory and the Improved Multivariate Support Vector Regression(IMSVR)algorithm.Using the black box theory,the brake cylinder pressure time series was used as the identification signal of the performance state of the air brake system.The input and output signals of the air brake system were fused,a data augmentation prognosis model of brake cylinder pressure time series was established based on the IMSVR algorithm,and the Particle Swarm Optimization(PSO)algorithm was used to optimize the kernel parameter γ and the penalty parameter C.The performance tests of the air brake system at different levels were carried out.The analysis results of the test data show that the modeling error of the prognosis model established by using the performance test data at different levels is within ±5k Pa.The calibration errors under the two typical operating conditions of"commonly used full braking-relief and "stage braking-relief " are also within ±5k Pa,which verifies the validity and applicability of the method.Fifth,performance degradation prognosis of the components of air brake system of high-speed trains.Aiming at the problem that the performance degradation trend of the components of air brake system was not completely clear due to the signal coupling,the influence of variable working conditions,and the long performance degradation period of the components of air brake system,a performance degradation prognosis method of the components of air brake system based on Relative Characteristic(RC)and Long Short-term Memory(LSTM)network was proposed.The input and output signals of the components were isolated and fused,the condition-independent RC was extracted to construct the Health Indicator(HI),and the monotonicity,correlation and robustness indicators were used to test the validity of the HI.Considering the time memory characteristics,the trend prediction of the HI curve of the component of air brake system was carried out based on LSTM.Taking the typical brake componentsfilter elements of air compressor as an example,the performance degradation test was carried out.The analysis results of the test data show that the accuracy of the singlestep performance prognosis of the air intake filter,oil separator and oil filter is over99%,which validates the effectiveness of the proposed method.Qualitative and quantitative analysis methods were integrated in this thesis,and a systematic study on the fault diagnosis and prognosis method of air brake system of high-speed trains around the three levels of system-subsystem-component was conducted.The effectiveness of the methods was verified by the test data obtained from performance tests,fault injection tests and performance degradation tests,which enriches the fault diagnosis and prognosis methods and theories of brake systems of rail transit vehicles to a certain extent.It provides support for PHM technology of brake system and has important theoretical significance and engineering value.