A new practical and intuitive method for kurtosis control in random vibration testing

摘要

Experimental simulations of random vibrations are important for automotive and aerospace component testing. Shaker excitation signals are typically generated by applying the Inverse Fast Fourier Transform (IFFT) to the desired Power Spectral Density (PSD) profile in combination with randomized phase spectra. However, signals generated in this way will always have a Gaussian probability distribution, while real-life excitations typically contain distinctive peaks exceeding the average level of vibration, i.e. non-Gaussian features. The statistical parameter known as "kurtosis" can be used to characterize these distinctive peaks, so that so-called "kurtosis control" is regarded as an approach to solve this problem. In this article, a new method to modify the input signal kurtosis by manipulating phase spectra is proposed. The method also manages to maintain a high output kurtosis, i.e. after passing through a system transfer function. This can be achieved while still preserving the desired PSD (as in the traditional random vibration control), because only the phase spectra are manipulated by the method. Moreover, it will be demonstrated that a practical and intuitive control strategy with reasonable precision and high convergence speed can be realized.