Compression of Strain Load History Using Holder Exponents of Continuous Wavelet Transform

Abstract

This paper presents the compression of strain loading time history of automobile suspension spring by extracting singularities in the signal using Lipschitz regularity analysis. Time histories of suspension spring always contained redundant data that increase the size of the signal and are insignificant to durability analysis. Excessive signal data will cause the analysis to be time consuming and computationally expensive. Hence, elimination of insignificant data is important to improve the efficiency of durability analysis. A strain signal was captured from a suspension spring of a sedan car and analysed with continuous wavelet transform to identify the modulus maxima lines. Holder exponents of each singular point were estimated from the log-log plot of modulus maxima lines. The extracted singularities was compressed and compared against the original signal the determine durability and were compared statistical to determine the characteristics of the signal. A conventional time-domain-based fatigue data editing technique had been performed to compare the effectiveness of Lipschitz-based technique. Results showed that the Lipschitz-based edited signal was only quarter of the original signal length that could retain 100% of fatigue damage of the original signal with less than 5% of difference when compared in terms global statistics. Lipschitz-based technique had outperformed the time-domain-based technique which had shown unacceptable deviations in global statistics. This suggested that the Lipschitz-based singularities extracted using Lipschitz regularity analysis can sufficiently represent a strain loading history without compromising the data behaviours.