The effect of noise on a direct finite element updating based on measured input–output

This paper tries to establish some `safety conditions' for the use of direct updating procedures. Such procedures are based on some kind of force balance (input error), and can be very sensitive to noise in the measured quantities. So, in spite of computational advantages, direct procedures have not overwhelmed the ones based on response comparison (output error), which, due to the non-linear relationship between responses and physical parameters, are much more time consuming. The updating procedure under test is first described, and a reliability indicator useful for theoretical work is developed. Then a statistical model of the noise is defined and its effect on the results is examined. Through the defined indicator, the various factors entering the procedure, such as the selection of working frequencies and the use of it a priori information, are assessed.