Automatic Hysteresis Feature Recognition of Vehicle Dampers Using Duhem Model and Clustering
Hong He, Yonghong Tan, Wei Yang, Feihu Peng, Wuxiong Zhang · 2018
Dampers of vehicle suspension systems usually show unsymmetrical nonlinear hysteresis characteristics, which affects the control accuracy of suspension system. Automatic recognition hysteresis features of unknown vehicle dampers can reduce the burden for developing efficient control strategies for suspension systems. According to the Duhem hysteresis model, a model-based unsupervised classification approach is proposed in this paper for recognizing hysteresis features of vehicle dampers. Through constructing Duhem models for different force-displacement hysteresis loops of the same sample, significant hysteresis features are extracted for multivariate time series of vehicle dampers. In terms of model parameters of samples, the clustering method is used to group dampers into different classes. Clustering results of 700 samples show that model-based features can efficiently describe the hysteresis features of vehicle dampers. K-means and Fuzzy-C Mean (FCM) are more appropriate for clustering the damper model feature data than hierarchical agglomerative clustering (HAC).