Functional Verification of Extended Kalman Filter on FPGA for Railway Wheelset Parameters Estimation

The forces generated at the wheel-rail contact patch are important and crucial to the railway vehicle's operation because they control the dynamic behavior of the entire rolling stock. The transfer of the tractive force applied by the locomotive's traction and braking system requires a certain amount of adhesion force at the wheel-rail contact area. The wheel may skid while decelerating and slip when accelerating due to the maximal adhesion force at the wheel-rail interface may be less than the applied tractive force. A major difficulty is accurately estimating adhesion conditions and implementing them in real time because the contact forces between the wheel and rail are complicated and are highly non-linear. The fast-processing speed and hardware flexibility of field programmable gate arrays (FPGAs) make them ideal for the implementation of nonlinear condition monitoring systems. The goal of our work is to estimate various railway wheelset characteristics relating to wheel-rail contact forces under variable track-contact situations by designing and implementing the extended Kalman filter (EKF) model utilizing MATLAB and FPGA. The onboard estimation of wheel-rail interaction parameters is verified using the Xilinx® System-On-Chip (SoC) Zynq and Xilinx Spartan-3 FPGA devices, built-in with the National Instruments (NI) myRIO® development board and sbRIO® single-board controller. Two FPGA platforms are used to test the dataset generated by the MATLAB railway nonlinear wheelset model under different track circumstances, which include the vehicle's acceleration and deceleration modes. The 95% of the simulation results achieved. The proposed model's hardware synthesizes on FPGA are validated, and the area-performance is analyzed. Both FPGA devices take 52 cycles per sample and hence throughput becomes 769 kilo samples per second. Comparing the Xilinx System-On-Chip Zynq FPGA device to the Spartan-3, it is found that the latter offers a worse area-performance tradeoff for the suggested calculation of railway wheelset parameters. The industry 4.0 is greatly benefit from the promising functional verification and area-performance analysis on small commercial FPGA devices to detect and identify the faults in the early stages as part of predictive maintenance that saves time, money, and reduce accidents caused by railway derailments. Hence properly estimation of wheel-rail contact conditions with real-time implementation on FPGA can meet the expectation that railway vehicles would be extremely fast, comfortable, safe, and cost-effective modes of transportation worldwide.