112 J. P. Winkel et al. Fig. 12.15 Drive-Point FRF Check Showing No Phase Shifting Outside 180 Degrees Fig. 12.16 FRF and Stability Diagram using Uncorrelated Broadband Forces The FRFs matched very well for the first half of the frequency range of interest, but above that the SOL 111 FRFs started to diverge as shown in Fig. 12.17. The SOL 111 FRF exhibited a frequency shift, which grew steadily as the frequency increased. The following three potential sources of error were looked at: 1. Round off error in the time domain data. 2. Restarting the MSC NASTRAN decks without recalculating the eigenvalues every time. 3. Differences in the residual vectors being calculated between the SOL 111 and SOL112 solutions. First, if the issue was round off error, was investigated by lowering the sample frequency of the uncorrelated broad band forces being generated, which resulted in the SOL 111 FRF (orange curve) starting to diverge at a low frequency, as shown in Fig. 12.18. It should be noted that large field format was utilized to minimize round off error. Next, could the issue be due to not setting up the NASTRAN restart decks correctly? A solution was run without any restarts and compared back to the original. No improvement on the results were seen. Finally, could the issue lie with the fact that the residual vectors being used in both solutions be different from each other? This was verified not to be the case by purposefully not calculating as many residual vectors on one analytical case and seeing the effect to the data. The resulted in slight in a very slight change magnitudes, but not frequencies, as shown in Fig. 12.19.
RkJQdWJsaXNoZXIy MTMzNzEzMQ==