Verification of Mainframe Computer Structure Finite Element Model Under Vibration and Seismic Tests

Abstract

Shorter development design schedules and increasingly dense product designs create difficult challenges in predicting structural performance of a mainframe computer’s structure. To meet certain certification benchmarks such as the Telcordia Technologies Generic Requirements GR-63-CORE seismic zone 4 test profile, a physical test is conducted. This test will occur at an external location at the end of design cycle on a fully functional and loaded mainframe system. The ability to accurately predict the structural performance of a mainframe computer early in the design cycle is critical in shortening its development time. This paper discusses an improved method to verify the finite element analysis results predicting the performance of the mainframe computer’s structure long before the physical test is conducted. Sine sweep and random vibration tests were conducted on the frame structure but due to a limitation of the in-house test capability, only a lightly loaded structure can be tested. Evaluating a structure’s modal stiffness is key to achieving good correlation between a finite element (FE) model and the physical system. This is typically achieved by running an implicit modal analysis in a finite element solver and comparing it to the peak frequencies obtained during physical testing using a sine sweep input. However, a linear, implicit analysis has its limitations. Namely, the inability to assess the internal, nonlinear contact between parts. Thus, a linear implicit analysis may be a good approximation for a single body but not accurate when examining an assembly of bodies where the interaction (nonlinear contact) between the bodies is of significance. In the case of a nonlinear assembly of bodies, one cannot effectively correlate between the test and a linear, implicit finite element model. This paper explores a nonlinear, explicit analysis method of evaluating a structure’s modal stiffness by subjecting the finite element model to a vibration waveform and thereafter post processing its resultant acceleration using Fast Fourier Transformation (FFT) to derive the peak frequencies. This result, which takes into account the nonlinear internal contact between the various parts of the assembly, is in line with the way physical test values are obtained. This is an improved method of verification for comparing sine sweep test data and finite element analysis results. The final verification of the finite element model will be a successful physical seismic test. The tests involve extensive sequential, uniaxial earthquake testing in both raised floor and non-raised floor environments in all three directions. Time domain acceleration at the top of the frame structure will be recorded and compared to the finite element model. Matching the frequency content of these accelerations will be proof of the accuracy of the finite element model. Comparative analysis of the physical test and the modeling results will be used to refine the mainframe’s structural elements for improved dynamic response in the final physical certification test.