Using fuzzy C-means in selecting a suitable wavelet transform for fatigue failure analyses

This study aimed to select a suitable wavelet transform for fatigue failure analyses. Strain signals of vehicle coil spring driven on road surfaces were divided into smaller segments. Subsequently, fatigue analyses were carried out using the Coffin-Manson, Morrow, and Smith-Watson-Topper models, which provided the highest fatigue damage of 1.49E-3 per block, with the shortest fatigue life of 6.73E2 reversals of block. Fatigue life was correlated with the Morlet and Daubechies wavelets using the fuzzy C-means. The clustering results showed a negative correlation, indicating that a higher wavelet coefficient was associated with lower fatigue life and vice versa. The strongest correlation was provided by the 30th order of discrete Daubechies wavelet and the coefficient of determination of 0.578 was obtained for the Smith-Watson-Topper model. These results showed that the 30th order of discrete Daubechies wavelet could enhance the understanding of fatigue failure and provide valuable insight into assessing the useful life.