Synthesis of Vibration Signals with Prescribed Power Spectral Density and Kurtosis Value

In the field of vibration qualification testing, random excitations are typically set as input in terms of a PSD profile. The physical motion at the shaker head is obtained through the application of the Inverse Fourier Transform in combination with randomized phases. The overall probability distribution of the input signal tends toward Gaussian, whereas distinctive peaks are often present in real-life random excitations, causing the probability distribution to be non-Gaussian. The parameter known as kurtosis is usually exploited to quantify the feature of non-Gaussianity. Several methods have been proposed to control kurtosis, still maintaining the desired PSD profile, in order to synthesize more realistic signals. However, kurtosis control implemented by some of these methods may be ineffective. In fact, in some cases, the response of a lightly damped system can prove closer to Gaussian than the applied excitation. This work presents two novel algorithms to effectively control kurtosis in random vibration tests are proposed.