Optimization of PID Control Parameters for Quarter-Vehicle Model Active Suspension System Using Back Propagation Neural Network and Genetic Algorithm Methods

A suspension system is designed to attenuate vibrations and enhance passengers’ comfort levels when a vehicle traverses uneven road contours. With additional power to control the actuators’ force, the active suspension system produces better ride quality than passive suspension systems. System modeling will be carried out using a quarter-vehicle model with a disturbance input form of a step input of 0.1 m. In this study, the control system used is the PID controller, with the parameters of the controller being tuned using the Back Propagation Neural Network-Genetic Algorithm (BPNN-GA) metaheuristic method employing MATLAB R2022b software. In this method, BPNN is used to produce a network that represents the correlation between controller parameters (i.e., K_p, K_i, and K_d) and system response (i.e., Integral Times Absolute Error (ITAE)). The ITAE value represents the value of Settling Time (Ts) and Peak Overshoot (PO) in a damping system. Next, the Genetic Algorithm (GA) is employed to determine the best BPNN's network with a minimum MSE (Mean Squared Error) value. The performance of BPNN-GA is then compared with Ziegler Nichols method. The best BPNN’s network is obtained with five hidden layers, ten nodes in each hidden layer, and the satlin activation function achieving an MSE training value of 1.6477 × 10^–8. The most optimum K_p, K_i, and K_d values are identified through the BPNN-GA method, namely 809193, 621978, and 243984, with a settling time (T_S) value of 1.33 s and a peak overshoot of −0.00834 m, along with a Root Mean Square (RMS) value of 0.5747 m/s². Moreover, in accordance with the ISO 2631 standard, the active suspension system model is classified as slightly uncomfortable.