Finite element modeling and computerized structural analysis are heavily relied upon in the design and development of engineered structures. Frequently, these structures are tested as a means of verifying the structural models. Vibration tests are performed to measure modal characteristics, which are then compared with computed frequencies and mode shapes. These comparisons contain basic information on structural uncertainty, which is herein defined as the difference between analytical and experimental estimates.
This paper presents a method for modeling structural uncertainty for generic classes of structures, based on actual analysis and test experience. It shows how the resulting generic statistical models can be applied to other structures of the same class, as a practical way of quantifying the accuracy of predicted frequencies, mode shapes and forced response, prior to testing. A realistic numerical example is presented to illustrate the general method. Statistical correlation is preserved throughout the process, and is traced through the numerical example.