Fatigue Feature Clustering of Modified Automotive Strain Signals for Saving Testing Time

Abstract

This paper discusses on the clustering of damaging fragments from a fatigue data editing leading to fatigue damage using the fuzzy C-means. In this work, strain signals were measured from an automotive coil spring which involved vehicle movements on three types of road surfaces at different speeds. Next, the extractions of higher amplitude cycle utilising the wavelet transform were conducted. From the results, it was obtained that the transformation was able to shorten the strain signal time up to 81.5 % removing 81.5 % of lower amplitude cycles, which these cycles theoretically contribute to minimum fatigue damage. Therefore, retaining of fatigue damage by more than 90.4 % was obtained. Furthermore, the resulted fragments from the extraction processes had been clustered utilising the fuzzy C-means, providing a significant coefficient of determination, reaching 0.9036. The testing time was successfully decreased up to 81.4 % using the edited strain signals. In conclusion, the technique could be successfully used to shorten a strain signal without changing the main history.