Consideration and Verification of Accuracy Evaluation Indices for Operational TPA

抄録

In this study, we considered accuracy evaluation indices for operational transfer path analysis (TPA) that employs the principal component regression method, and we verified their reliability. To evaluate the accuracy of TPA, the consistency of the response signals, in which the calculated response signal from the TPA model is compared with the measured response signal, has been frequently used in the past. Also, some cases use the condition number that indicates the inverse matrix condition used in the calculation procedure of the acceleration transfer function. In addition to these accuracy evaluation indices, a correlated principal component number (CPCN), which indicates the number of principal components that correlate with the response signal generated at the running test, has been proposed. The reliability of these accuracy evaluation indices has been verified through a simple simulation. The results show that the two conventional indices (the consistency method and the condition number) do not satisfactorily evaluate the accuracy. However, a CPCN can indicate the frequency bands where the TPA accuracy is high or low. Consequently, the new index was found to be a suitable index for evaluating the accuracy of the operational TPA.