Fault Diagnosis and Control Performance Evaluation of a Semi-Active Suspension Vehicle using Policy Optimization based Model Predictive Control

Full-scale vehicle simulation enables virtual validation of control strategies, reducing cost and time associated with physical experiments. However, previous studies have primarily focused on passive suspension fault detection or control design without considering their effect on fault diagnostics. In an uncontrolled system, road disturbances mask fault-induced vibrations, complicating fault identification. This study proposes a semi-active suspension control strategy that integrates fault classification and adaptive control to enhance vehicle stability. The mathematical model is validated using Adams Car and is used to evaluate various control strategies, such as the Linear Quadratic Regulator, Model Predictive Control, and Policy Optimization-based Model Predictive Control. The validated model produces z-axis vibration data, which is analyzed using the Short-Time Fourier Transform to classify suspension arm failures, including joint wear, structural cracks, and bushing degradation. The proposed approach differentiates faults by analyzing variations in natural frequency and damping characteristics. Applying different control approaches based on specific fault types, the system mitigates performance degradation caused by various suspension arm failures. Results show that the proposed strategy preserves vehicle stability while dynamically adjusting control responses based on identified fault characteristics.