The problem of structural model updating has gained much interest as finite element model capabilities and modal testing have become more mature areas of structural dynamics (e.g., [1][2][3]). The need for model updating arises because there are always errors associated with the process of constructing a theoretical model of a structure, and this leads to uncertain accuracy in predictive response. Moreover, a model updating methodology is useful in predicting the structural damage by continually updating the structural model using vibration data [4][5][6][7][8][9][10][11][12]. Such updated models obtained periodically throughout the lifetime of the structure can be further used to update the response predictions and lifetime structural reliability based on available data [13][14][15][16]. The updated information about the condition of the structure can be used to identify potentially unsafe structures, to schedule inspection intervals, repairs or maintenance, or to design retrofitting or control strategies for structures that are found to be vulnerable to possible future loads.Structural model updating is an inverse problem according to which a model of a structure, usually a finite element model, is adjusted so that either the calculated time histories, frequency response functions, or modal parameters best match the corresponding quantities measured or identified from the test data. This inverse process aims at providing updated models and their corresponding uncertainties based on the data. These updated models are expected to give more accurate response predictions to future loadings, as well as allow for an estimation of the uncertainties associated with such response predictions. In practice, the inverse problem of model updating is usually ill-conditioned due to insensitivity of the response to changes in the model parameters, and nonunique [17-20] because of insufficient available data relative to
-2Engineering Design Reliability Handbook the desired model complexity. Additional difficulties associated with the development of an effective model updating methodology include: (1) model error present due to the fact that the chosen class of structural models is unable to exactly model the actual behavior of the structure; (2) measurement noise in the dynamic data, especially for higher modes; (3) incomplete set of observed DOFs (degrees of freedom) due to the limited number of sensors available and the limited accessibility throughout the structure; (4) incomplete number of contributing modes due to limited bandwidth in the input and the dynamic response.This chapter presents a Bayesian framework for statistical modeling and updating of structural models utilizing measured dynamic data, along with its use in making informed structural response predictions and reliability assessments. The Bayesian approach to statistical modeling uses probability as a way of quantifying the plausibilities associated with the various models and the parameters of these models given the observed data. The Bayes rule uses a prior...