Modeling and Optimization of Half Car Suspension and System Identification with Artificial Neural Networks Using Matlab

The suspension system in a vehicle plays an important role in ensuring user comfort and safety by dampening vibrations from uneven road surfaces, maintaining vehicle stability, and ensuring contact between the wheels and the road. The half-car suspension model is a simple but effective representation for understanding and controlling the vertical and pitch dynamics of a vehicle. This model is includes the sprung and unsprung mass at the front and rear of the vehicle. In this research, a half-car suspension model is employed to analyze these dynamics, focusing on three prominent system identification methods: NLARX, Transfer Function, and Neural Networks. This model improves predictive performance compared to the linear model, based on the test results, this model achieves high model fit results with the estimated data and a very low Final Prediction Error (FPE). The Transfer Function model is used to represent the input-output relationship of the system in the frequency domain, making it useful for systems that are mostly linear. This Transfer function model gets a fairly good fit value. Neural Network model is also used to gain advantages in capturing complex nonlinear behavior. The results obtained on this Neural Network have higher accuracy in predicting system responses, especially in systems that operate outside the linearity range. Based on the tests conducted, the best validation performance achieved is 0.026939 at epoch 3, indicating good alignment between the actual and predicted displacement responses for the front and rear suspensions. This study compares the three methods based on model fit and FPE. Comparative analysis shows that combining these methods improves the predictive performance of the model, resulting in a more accurate and adaptive vehicle suspension system. This approach can improve the overall performance and optimization of the half-car suspension model.