Multi-structure data driven wheel flat quantitative identification method and system

The invention discloses a multi-structure data driven wheel flat quantitative identification method and system, and the method comprises the steps: constructing a vehicle-track rigid-flexible coupling dynamic model, and taking a synthesized flat as wheel out-of-circle excitation to obtain axle box dynamic responses under different working conditions; respectively carrying out time domain, frequency domain and time-frequency domain processing on the vertical vibration acceleration of the axle box to prepare sample sets with different structural forms; a multi-input convolutional neural network with proper structures and configuration parameters is constructed, vehicle speed signals are used as constraint information of the network, sample sets of different structural forms and combinations thereof and the speed signals are fused and input to an MCNN model for training, and the accuracy and timeliness of quantitative recognition of the wheel flat by the MCNN model under different data structures are compared. B