This study proposes a continuous convolution method combined with memoryless nonlinear transformation for multi-input multi-output stationary non-Gaussian random vibration tests. The challenge of the multi-shaker non-Gaussian random vibration test lies in the coupling problems that are manifested in the inherent physical system and in the existence of cross-spectral densities. In the presented method, the independent stationary Gaussian random signals pass through a designed finite impulse response filter with a convolution manipulation first, and then the resulting signals are transformed to the non-Gaussian random signals by the memoryless nonlinear transformation method. The desired drive signals are obtained by the input–output relationship in the frequency domain. The finite impulse response filter is constructed by the frequency sampling technique in which the amplitude characteristics of the filter are determined by the predefined reference power spectral densities. A new monotonic nonlinear transformation function with an approximate kurtosis solution is provided. It only contains one parameter for kurtosis control both in sub-Gaussian and super-Gaussian cases. The memoryless nonlinear transformation is used to maintain the cross-spectral densities, although some distortions are introduced to the power spectra during the transformation process. The inverse system method is used to overcome the coupling problem caused by the inherent physical system. A simulation example and a triaxial vibration test are carried out, and the results indicate the validity and feasibility of the proposed method.
Read full abstract