Leptokurtic signals in random control vibration testing

In several industrial sectors, some components are subjected to mechanical vibrations which may lead to a premature failure. To ensure that they operate properly during their service life, the utilization of qualification tests has been consolidated over the years. It is often required to carry out accelerated tests for obvious reasons as feasibility and cost: the aim is to limit the duration of tests. The Test Tailoring procedure requires an appropriate definition for vibratory test profiles to be utilized as an excitation in terms of motion generated by vibrating tables or shakers. The synthesis of such profiles requires that signals be measured in real environments and then that their most important characteristics be reproduced in a laboratory, in particular their spectral content and damage potential.The conventional procedures permit the synthesis of an accelerated test profile in terms of a Power Spectral Density, which is characterized by a Gaussian distribution of the corresponding timeseries values. Such a kind of synthesis might be unfit to represent the real environment signal taken as a reference, owing to the latter’s usual non-Gaussianity. As a consequence, reliability could be compromised since the “nature” of the real signal is not preserved. Typical examples of non-Gaussian signals coming forth in real applications are the so-called Leptokurtic signals, whose high amplitude peaks originate a strongly non Gaussian probability distribution. A parameter called kurtosis is often employed to represent the number and severity of the peaks of the signal. A common reference is made to “kurtosis control” whenever it is required that the synthesized and the measured signal have not only the same spectral content but the same kurtosis value as well. In this work some novel Mission Synthesis algorithms are proposed, which generate test profiles by controlling precisely the kurtosis value and complying with the spectral content of the reference signal.