Evaluation of model reduction methods using modal test data

Model reduction algorithms used to generate the test analysis model for direct correlation with modal survey data are evaluated for accuracy using measured modal test data. Using test data allows some variables, such as the instrumentation calibration uncertainty and the noise contained in the dynamic testing environments that normally are not easy to simulate, to be included in the assessment. Five commonly used model reduction methods, Guyan (static) reduction, improved reduced system method, modal reduction, hybrid reduction, and Craig-Bampton reduction are evaluated based on their performances on the mass orthogonality matrix and the cross orthogonality matrix. The selection of the mass orthogonality and the cross orthogonality matrices as the evaluation parameters is based on the fact that only the reduced mass matrix is deteriorated by the reduction transformation matrix. The evaluation results of three Space Station hardware modal survey test data indicate that the Guyan reduction method provides the better data correlation. It is also the easiest method to implement as well as low sensitivity to test-analysis errors.