Influence of PSD Estimation Parameters on Fatigue Life Prediction in Spectral Method

Abstract

The paper analyzes in detail the influence of the Welch method on the estimation of fatigue life in the frequency domain. The aim is to examine how the setting of parameters, such as window size, weighting and overlap, affects the calculation of power spectral density (PSD) and thus lifetime. The paper uses knowledge from the theory of signal processing, statistics and fatigue life estimation in the frequency domain. In order to demonstrate the general validity of the conclusions, four different reference PSD spectra are generated and examined. The spectra are converted to time histories using an inverse fast Fourier transform and then used as input (measured) data for the Welch method. The PSD thus obtained are subsequently subjected to lifetime estimation using the Dirlik and Tovo–Benasciutti models. Based on the obtained results, it was found that the best configuration of the parameters of the Welch method for fatigue calculations is the selection of the smallest possible time window (maximum 60% of the observation length) and the largest possible overlap (minimum 70% of the window size).

Keywords

• power spectral density
• Welch method
• spectral fatigue
• probability density function
• sensitivity analysis