4 Hybrid Substructure Assembly Techniques for Efficient and Robust. . . 41 Error Magnituge 5.0e-01 8.0e-01 2.0e+00 4.0e+00 7.0e+00 1.3e+01 3.0e+01 5.0e+01 1.0e+02 2.0e+02 5.0e+02 2.0e+04 1.0e+05 1.0e+11 1.0e+05 2.0e+04 5.0e+02 2.0e+02 1.0e+02 5.0e+01 3.0e+01 1.3e+01 7.0e+00 4.0e+00 2.0e+00 8.0e-01 5.0e-01 Output SNR 5.0e-01 8.0e-01 2.0e+00 4.0e+00 7.0e+00 1.5e+01 3.0e+01 5.0e+01 1.0e+02 2.0e+02 5.0e+02 2.0e+04 1.0e+05 1.0e+11 1.0e+05 2.0e+04 5.0e+02 2.0e+02 1.0e+02 5.0e+01 3.0e+01 1.5e+01 7.0e+00 4.0e+00 2.0e+00 8.0e-01 5.0e-01 Input SNR -8 -6 -4 -2 0 2 4 6 Error Phase 5.0e-01 8.0e-01 2.0e+00 4.0e+00 7.0e+00 1.3e+01 3.0e+01 5.0e+01 1.0e+02 2.0e+02 5.0e+02 2.0e+04 1.0e+05 1.0e+11 1.0e+05 2.0e+04 5.0e+02 2.0e+02 1.0e+02 5.0e+01 3.0e+01 1.3e+01 7.0e+00 4.0e+00 2.0e+00 8.0e-01 5.0e-01 Output SNR 5.0e-01 8.0e-01 2.0e+00 4.0e+00 7.0e+00 1.5e+01 3.0e+01 5.0e+01 1.0e+02 2.0e+02 5.0e+02 2.0e+04 1.0e+05 1.0e+11 1.0e+05 2.0e+04 5.0e+02 2.0e+02 1.0e+02 5.0e+01 3.0e+01 1.5e+01 7.0e+00 4.0e+00 2.0e+00 8.0e-01 5.0e-01 Input SNR -7 -6 -5 -4 -3 -2 -1 Correlation Magnituge 5.0e-01 8.0e-01 2.0e+00 4.0e+00 7.0e+00 1.3e+01 3.0e+01 5.0e+01 1.0e+02 2.0e+02 5.0e+02 2.0e+04 1.0e+05 1.0e+11 1.0e+05 2.0e+04 5.0e+02 2.0e+02 1.0e+02 5.0e+01 3.0e+01 1.3e+01 7.0e+00 4.0e+00 2.0e+00 8.0e-01 5.0e-01 Output SNR 5.0e-01 8.0e-01 2.0e+00 4.0e+00 7.0e+00 1.5e+01 3.0e+01 5.0e+01 1.0e+02 2.0e+02 5.0e+02 2.0e+04 1.0e+05 1.0e+11 1.0e+05 2.0e+04 5.0e+02 2.0e+02 1.0e+02 5.0e+01 3.0e+01 1.5e+01 7.0e+00 4.0e+00 2.0e+00 8.0e-01 5.0e-01 Input SNR 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Correlation Phase 5.0e-01 8.0e-01 2.0e+00 4.0e+00 7.0e+00 1.3e+01 3.0e+01 5.0e+01 1.0e+02 2.0e+02 5.0e+02 2.0e+04 1.0e+05 1.0e+11 1.0e+05 2.0e+04 5.0e+02 2.0e+02 1.0e+02 5.0e+01 3.0e+01 1.3e+01 7.0e+00 4.0e+00 2.0e+00 8.0e-01 5.0e-01 Output SNR 5.0e-01 8.0e-01 2.0e+00 4.0e+00 7.0e+00 1.5e+01 3.0e+01 5.0e+01 1.0e+02 2.0e+02 5.0e+02 2.0e+04 1.0e+05 1.0e+11 1.0e+05 2.0e+04 5.0e+02 2.0e+02 1.0e+02 5.0e+01 3.0e+01 1.5e+01 7.0e+00 4.0e+00 2.0e+00 8.0e-01 5.0e-01 Input SNR 0 0.2 0.4 0.6 0.8 Fig. 4.3 Evaluation results of the identification process of a two mass oscillator for different signal-to-noise ratios at in- and output, evaluated for ETFE, SPA, SID and PEM identification unreliable, since the evaluation shows no clear tendency for dropping SNRs. In contrast to the SID, more accurate results are generated with the PEM for a larger range of SNR values. Assembly The results shown in Fig. 4.4 are generated in the same way as in Fig. 4.3. In this case, the assembled system of the identified models and the exact assembly is compared. Except for the phase error, all other criteria show that the inaccuracies made in the identification are increased by the assembly process. The limits of in- and output SNRs in which accurate results are achieved decrease significantly or are not present at all. Additionally, the decrease in the achieved accuracy exhibits a sharp drop with decreasing SNRs in case of the PEM. This underlines that hybrid substructuring demands a high accuracy of the identification process in order to acquire useful results in the assembled system. For the present example, the PEM methods show by far the highest accuracy. Comparing Figs. 4.3 and 4.4 indicates that only the accurate results obtained by PEM deliver accurate results for the assembly. The assessment by an error measure does include a qualitative comparison. At this point, it is necessary to compare the acquired results on FRF level, which is done in the next section. 4.4.1.2 Acoustic Testbench In this section, the results of the identification and subsequent assembly process of the acoustic testbench are discussed. Additionally, only the methods SPA and PEM are considered, in order to provide clear plots. As shown in the previous section, these methods are more promising than ETFE or SID. The identification was conducted as presented in Sect. 4.3.2. The SNR on the input and output is set to 100. The evaluation of all FRF of the identified and subsequently assembled 10×6 system in terms of the used error measures is gathered in Tables 4.1 and 4.2. These findings correspond to the expectation that was indicated by the results presented in Sect. 4.4.1. For the chosen SNR values, the PEM shows better results compared to the SPA in terms of the used error measures, for both identification and assembly. The comparison criteria that have been applied can be used to detect changes for entire systems when the
RkJQdWJsaXNoZXIy MTMzNzEzMQ==