Analysis of synthesized non-Gaussian excitations for vibration-based fatigue life testing

Certain applications require that some critical components must undergo vibration qualification tests to check their suitability with respect to dynamic excitations. The common procedure is to consider measured field data as reference for the synthesis of random stationary signals to be used as input excitations for shakers or slip-tables. The current synthesis procedures usually generate test profiles in terms of PSD, corresponding to stationary random processes with Gaussian probability distribution of values. Such signals may be unrealistic in representing the characteristics of the reference data if the latter are not Gaussian. The Kurtosis parameter is often used to synthetically represent the amount and the amplitude of signal peaks. Its value is 3.0 for Gaussian signals, whereas higher values hold for signals featuring non-negligible high peaks, for example due to micro-shocks. In case of accelerated fatigue life tests, the synthesized signal must induce, in a limited duration, the same fatigue damage which the reference signal cause on the tested component throughout the expected lifetime of the latter. The Fatigue Damage Spectrum (FDS) is generally used to quantify the fatigue damage potential associated with the excitation. The test signal is synthesized targeting the same FDS computed for the reference profile. This paper presents two kurtosis-control algorithms for the synthesis of test profiles in combination with a technique able to match the prescribed FDS.