iterative algorithm. The improvement with respect to the uncoupled basis is fair. The error levels indicated by Δrepresent the error levels obtained with the vectors that result from IRCA on the set tolerance level. The improvement on the latter is better for the considered number of mode vectors as one should expect. The application of the two-sided Craig-Bampton scheme outlined in Section 3.1 (equations 26 and 27) is also given for comparison purposes and labeled as tsCb. 0 5 10 15 20 25 30 10−7 10−6 10−5 10−4 10−3 10−2 10−1 Mode number Δ f/f Relative frequency error comparison Unc basis tsCb 1 Iteration 2 Iteration Fig. 3 Scheme section 3.1, 1 iteration vs. 2 iterations Figure 4 reports the relative frequency error results of the proposed scheme in Section 3.2. As mentioned in the theoretical exposition, this scheme has the advantage of the assembly of the generalized dofs on the full reduced system level for the integration of the global modes into the reduction basis. The comparison of the current results with the two-sided Craig-Bampton scheme is also provided for these results. The results show fair accuracy for kept number of vectors in the reduction bases. Figure 5 represents the relative frequency error results of the proposed scheme in Section 3.3. On this result figure, the combination of the mode vectors has been changed for comparison purposes. Namely, F mark in the legend represents the number of fixed interface modes, similarly the I mark represents the integrated number of global pseudo vectors into the reduction basis (e.g. 25F-5I represents the fact that 25 Fixed interface modes with 5 Irca modes are used in the mode block of the basis.). It should be mentioned that for this test, the ultimate results of the iterative correction algorithm is used. In general, increase of the global mode count results in a better accuracy. Umut Tabak and Daniel J. Rix 276
RkJQdWJsaXNoZXIy MTMzNzEzMQ==