Fatigue Feature Classification for Automotive Strain Data

Fatigue strain signal were analysed using data segmentation and data clustering. For data segmentation, value of fatigue damage and global statistical signal analysis such as kurtosis was obtained using specific software. Data clustering were carried out using K-Mean clustering approaches. The objective function was calculated in order to determine the best numbers of groups. This method is used to calculate the average distance of each data in the group from its centroid. Finally, the fatigue failure indexes of metallic components were generated from the best number of group that has been acquired. Based on four data collect from two different roads which are D1, D2, the index value generated is not the same for all of data because due to K-Mean clustering, the best group is different for each of the data used. The maximum indexes generated are different for two types of road and namely the index 4 for D1 and index 5 for D2. Due to the road surface condition, higher distributions of the best groups give higher values of index and reflect to higher fatigue damage experienced by the system.