Research On Disc Brake Uneven Wear Of High-speed Train Monitoring Based On Modified Convolution Neural Network

The braking of high-speed trains primarily relies on the friction between brake pads and brake discs to ensure complete deceleration or parking within a safe distance.In the braking process of high-speed trains,the friction block on brake pad will inevitably exhibit an uneven wear phenomenon.During the continuous mechanical braking process,uneven wear will induce a vibration reaction at the friction interface and generate the corresponding noise,which not only pollutes the environment along the railway line but also discomforts passengers.In addition,the uneven wear phenomenon will reduce the braking torque,thereby threatening the safe operation of high-speed trains.Therefore,it is crucial to monitor the uneven wear of friction blocks in the braking system of high-speed trains.The main contents are as follows:(1)Mechanism analysis and data acquisition of uneven wear of friction block of highspeed trainAiming at the problem of varying degrees of wear on the friction contact surface of the brake friction block in the braking process,considering that the wear state data was difficult to collect in the actual braking condition,an experimental scheme to obtain the braking condition data in the laboratory environment was proposed.Starting from the uneven wear mechanism and combining with the current situation that the high-speed train brake block presents the characteristics of multiple leading edges in the braking process,the braking experiment was completed through the simulation experiment platform and the reliable uneven wear data such as vibration noise,tangential vibration acceleration and friction coefficient were obtained,which provided data support for the subsequent establishment of the monitoring model.(2)Disc brake uneven wear condition monitoring of high-speed train based on 2D image and 2DCNN-Bi GRUTo ensure the effectiveness of brake uneven wear condition monitoring,1D vibration signals of uneven wear state were converted into 2D image data using gramian angular fields(GAF)to not only retain complete information of signals,but also maintain the dependence of signals upon time.Furthermore,a monitoring model based on 2-dimensional CNN(2DCNN)and bidirectional gated recurrent unit(Bi GRU)was proposed.Firstly,2D image date were taken as input data of the model,then spatial features of image data were extracted with2 DCNN,and temporal features were screened with Bi GRU.Finally,the mode recognition was finished with a classifier.The feasibility of the proposed method has been demonstrated by monitoring the friction blocks in different states of uneven wear,which was better than other commonly used algorithms.(3)Disc brake uneven wear condition monitoring of high-speed train based on MCNNSVM with small samplesAiming at the problem of small samples of brake friction data in practical engineering,CNN combined with support vector machine(SVM)was proposed to establish a monitoring model for brake uneven wear of high-speed trains under small sample data.The modified CNN(MCNN)model was established by introducing Mish activation function,batch normalization and Dropout,which was used to learn the potential rules of the uneven wear data layer by layer and migrate the feature knowledge to SVM.Finally,the SVM completed the monitoring of the uneven wear state of the brake friction block.The experimental results showed that the proposed method significantly outperforms other algorithms in both convergence speed and accuracy in monitoring experiments with small sample data sets.(4)Disc brake uneven wear of high-speed train intelligent monitoring using an ensemble model based on multi-sensor feature fusion and deep learningAiming at the problem of the complex operating conditions of the friction interface and multiple sources of vibration excitation during braking can reduce the monitoring accuracy with only a single sensor,we proposed a deep learning-based ensemble model for intelligent monitoring of the brake’s uneven wear condition.First,an ensemble model based on 1DCNN and Bi GRU was developed to learn spatial and temporal knowledge from the training set of vibration noise,vibration acceleration and friction coefficient.Second,the initial weights were assigned to each model,and the validation set of each signal was used to determine the best weights combined with a grid search algorithm.The models were integrated to achieve multisource information fusion to improve the accuracy of the final decision.Finally,the ensemble model exploited multi-source feature information from different perspectives.It established the mapping relationship between feature space and sample label space and obtained the diagnosis results through the test set.Various experimental results showed that the proposed method was highly accurate and stable as well as could effectively monitor the various uneven wear conditions of braking system.