Subspace-Based System Identification and Fault Detection for Suspension System Based on Vibration Analysis

الملخص

System identification is a technique that can be employed to obtain a mathematical representation of how a system behaves. The objectives of this research are to determine the modal parameters, such as damping coefficient and spring stiffness, and to explore an online condition monitoring system for the suspension. This will be achieved by using a MATLAB simulation of a seven degree-of-freedom (7-DOF) model of a complete vehicle. The accuracy of the simulation will be ensured by measuring only the accelerations of the vehicle's sprung mass, using subspace identification methods. Stochastic subspace identification (SSI) approaches, which just utilize output data, are employed to ascertain the whole vehicle model that remains applicable across the entire range of operation.
Common issues related to suspension components include damaged or leaking shock absorbers and weakened springs. These deficiencies frequently lead to a decrease in the vehicle's overall performance. The simulation incorporates suspension flaws by introducing damage to the shock absorbers (dampers). The faults are induced by reducing the damper coefficient by 25%, 50%, and 80%. This has served as the foundation for evaluating the ride comfort, road handling, and stability of the car, as well as identifying any potential damping issues at an early stage.
The cars included in this analysis of system identification exhibit non-linear dynamic behavior that is primarily influenced by the stochastic nature of road-tyre excitations. Instead of utilizing tire forces as inputs, which can be challenging to measure or predict, the inputs are based on the accelerations of the sprung masses. The vehicle's bouncing, pitching, and rolling modes are identified and described.

 Stochastic Subspace Identification (SSI) algorithms provide a precise and reliable estimation of uncertain vehicle characteristics, such as the natural frequencies and damping ratio of the bounce, pitch, and roll modes for the complete vehicle model. These estimations are not affected by the initial estimates or the excitation signals used. The model results were found to closely align with the theatrical results. Furthermore, the damping estimates exhibited a significantly greater level of variability compared to the frequency estimates.

Theoretical investigation indicates that the subspace identification method, which utilizes the accelerations of sprung masses as inputs, can provide reliable estimations of model parameters such as spring stiffness and damping coefficient.