Random Vibration Testing Beyond PSD Limitations

The traditional approach to random vibration testing using fast-Fourier-transform simulation has become out-of-date, since it is restricted to consideration of the power spectral density only. The latter means that the FFT approach is based on a Gaussian random signal model. However, the MIL-STD810F standard establishes that “care must be taken to examine field-measured probability density for non-Gaussian behavior.” A test engineer is now required to “ensure that test and analysis hardware and software are appropriate when non-Gaussian distributions are encountered.” There is widespread belief that the time waveform replication can address the non-Gaussian issue. However the TWR method is not a simulation, since the replication test is representative of just one road sample measured, not of a road type like the simulation test. This difference between replication and simulation is discussed here. Two methods of non-Gaussian simulation (polynomial function transformation and special phase selection) based on kurtosis and skewness characteristics are considered and examples of simulating various field data are given.