Neural Network Based Throttle Actuator Model for Controller 2019-26-0247

HiL is a closed loop validation setup widely used in the validation of real-time control systems. In the existing HiL setup, the ECUs to be tested are real while the remaining vehicle is modelled as plant model using Simulink. But some actuators like throttle valve, waste-gate valve, injectors, etc. are not modelled as plant model. Since these actuators exhibit hard nonlinearity, it is difficult to design accurate models of these actuators. So these actuators are connected to the HiL as real hardware components. But the major drawbacks of using real hardware components are: they need more space and they are costly. Hence, in this work, a real-time throttle actuator model for the controller is proposed. A throttle actuator contains a DC motor and a spring loaded flap. To create an accurate ODE based model of the throttle actuator, parameter identification of each component of the throttle actuator needs to be done separately by dismantling the actuator. This approach needs more effort and time. Hence, a robust non-linear learning based model is proposed. The learning based model uses neural network which is trained using input and output data across throttle actuator. To train the model, a new parameter estimation algorithm is also proposed. The proposed parameter estimation process is based on meta-heuristic simulated annealing search algorithm. The proposed model is trained by taking input and output data across the throttle actuator from HiL. The trained model is then validated using WLTP dataset and is found to be working satisfactorily. Hence, the model is planned to be tested real-time on HiL in the next step. In this work, the parameter estimation process is presented.