The Accuracy of Wavelet Coefficients in Identifying Fatigue Features Using a Fuzzy C-Means-based Integrated Clustering Algorithm

This study discussed the accuracy of wavelet coefficients in identifying fatigue features. The accuracy was based on the correlation coefficient and coefficient of determination obtained from the scattering of wavelet coefficients and the fatigue life developed using Fuzzy C-Means-based integrated clustering algorithm. During the analysis, fatigue-based strain data were simulated within a range of 200 to 2,000 µε of using both constant and variable amplitudes, with mean values set to negative, zero, and positive. Wavelet transforms included were Morlet wavelet and the 4th, 12th, 20th, and 30th orders of the discrete Daubechies wavelet, due to each fame in the field of fatigue. Meanwhile, fatigue life was decided by Coffin-Manson, Morrow, as well as Smith-Watson-Topper models. The findings revealed an inverse correlation between wavelet coefficients and fatigue life. A higher wavelet coefficient was correlated with lower fatigue life, and vice versa. The most significant correlation was found between the 20th-order discrete Daubechies wavelet and the Coffin-Manson model, with a value of -0.623 and a coefficient of determination of 0.388. These results suggest that the 20th order of the discrete Daubechies wavelet might be more reliable for use in fatigue life estimation models due to the stronger association with lower wavelet coefficients, which tended to correlate with longer fatigue life. In conclusion, these outcomes were expected to improve the understanding of fatigue studies and provide valuable perceptions for industries in assessing the useful life of their products.