Clustering of decomposed strain signal energy for durability classification

This paper presents clustering of automotive spring fatigue life for failure classification based on K-means approach. For safety promotion of buses, fatigue life prediction of the spring is needed to be classified for maintenance. In this analysis, the strain signals of a heavy vehicle leaf spring were collected from two common roads and analyzed using Hilbert Huang transform. The strain amplitude was used to obtain fatigue life of the leaf spring. Subsequently, the instantaneous frequencies, energies and fatigue lives were clustered into three groups according to the K-means approach. Numerous classification trees were trained with the clustered group as target while the instantaneous frequencies, energies and fatigue lives datasets as input. The trained classification trees were evaluated using receiver operating characteristic curve which shown an acceptable prediction of classes. This classification tree serves as a tool to evaluation automotive leaf spring design for fatigue failure prevention without destroying the component.