ياشار اماني

Vehicle suspension systems play a crucial role in ensuring ride comfort an‎d stability, particularly on rough roads. Traditional control methods often lack efficiency in adapting to dynamic conditions, necessitating advanced techniques to enhance performance. This thesis proposes a novel framework for online identification of nonlinear suspension system parameters with seven degrees of freedom, leveraging machine learning algorithms for real-time adaptive control an‎d state estimation. The proposed configuration integrates Unscented Kalman Filters (UKF) an‎d Radial Basis Function Neural Networks (RBFNNs), overcoming the limitations of traditional control methods such as PID an‎d sliding mode control, an‎d demonstrating high adaptability to road irregularities an‎d dynamic condition changes in real driving scenarios. Simulation results indicate a 98% reduction in tracking error an‎d a 93% suppression of unwanted oscillations, with steady-state errors below 1% for parameters affecting ride comfort an‎d stability. Additionally, the framework achieves a 37.5% improvement in control convergence speed, ensuring stability within 0.5 seconds with an RMSE error of only 0.02. Comparisons with AUKF an‎d IPS-UKF methods highlight the proposed frameworkʹs superior accuracy, computational efficiency, an‎d robustness. Future research will focus on developing adaptive noise filtration an‎d advanced input estimation to enhance the performance of next-generation intelligent control systems in real-world driving conditions.

شناسايي برخط پارامترها، سيستم تعليق غيرخطي، الگوريتم‌هاي يادگيري ماشين، كنترل تطبيقي در زمان واقعي، شرايط رانندگي واقعي

Online Parameter Identification, Nonlinear Suspension System, Machine Learning Algorithms, Real-Time Adaptive Control, Real Driving Conditions

Yashar Amani

Dr. Masoud Masih Tehrani