Experimental research for semi-active suspension system with piecewise learning fuzzy neural network-based adaptive control

The semi-active suspension system, which adjusts the damping characteristics in real-time based on road conditions, effectively enhances ride comfort and stability while reducing energy consumption. Although fuzzy control has been applied to vehicle suspension systems, the formulation of fuzzy rule functions involves a certain degree of subjectivity, limiting control accuracy. To address the limitations of traditional fuzzy control, this paper proposed a semi-active suspension control strategy based on a fuzzy neural network (FNN). The learning capability of neural network was used to optimize the fuzzy rule base of traditional fuzzy controllers. Moreover, the piecewise learning approach is employed to optimize the initial parameters of the neural network. In first stage, a clustering algorithm is employed to optimize the mean value c jm and the standard deviation σ jm within the neural network. Based on the obtained optimal initial values, the second stage applies a gradient descent method to further optimize the connection weight coefficient w ij . A segmented and improved fuzzy neural network system has been built. Then the different weight ratios of ride comfort and stability are discussed to the control strategy. Furthermore, a semi-active suspension dynamic performance test rig was built to analyze the proposed control strategy. Compared to the traditional fuzzy control, the FNN control strategy further reduces the a rms and φ rms . Under B-class random road condition at vehicle speeds of 10 km/h, 20 km/h, and 30 km/h, the a rms decreased by 9.19%, 13.60%, and 15.44%, respectively. Meanwhile, φ rms decreased by 10.49%, 8.96%, and 7.82%, when the vehicle speeds increased. Under 10% eccentric load conditions, for A, B, and C-class random roads, the a rms decreased by 16.78%, 13.18%, and 9.59%, while φ rms decreased by 11.63%, 15.03%, and 17.09%, respectively. These experimental results demonstrate that the designed FNN control strategy can effectively improve the ride comfort and stability of vehicle.