Advantage of using the fourier components pair instead of power spectral density for fatigue crack diagnostics

A new generic approach to condition monitoring and diagnostic feature-extraction is advocated. This generic approach is optimal for the analysis of frequency domain data for health monitoring and damage detection. The approach consists of using simultaneously new diagnostic features; real and imaginary parts of the Fourier transform. Thus approach is more genetic than conventional approaches based on power spectral density, and it can be shown that the power spectral density approach is a particular case of the proposed new generic approach. Numerical examples are given based on the analysis of synthetic signals generated using the nonlinear model of a diagnosis object. The synthetic signals simulated the forced vibroacoustical oscillation of a cracked part of machinery under narrow band Gaussian excitation. Our new generic approach is used for detecting damage consisting of a fatigue crack of various relative sizes. The numerical examples show that the detection, based on Fisher's criterion, was definitely more effective when using our new diagnostic features than when using the conventional power spectral density feature. The proposed new generic approach to damage detection offers a clear effectiveness improvement over the conventional approach based on the power spectral density.