Comparison of popular finite element model updating techniques

Finite element model updating techniques may be categorized with regard to the preservation of physically realizable mass and stiffness matrices. If such preservation is desired, updating based on "response error" is generally preferred. Two response error techniques are compared, one using mathematical programming and the other using Bayesian estimation for error reduction. The basis of comparison is a planar truss which has been tested and modelled. The two techniques are found to be essentially equally capable for the case studied. However, the Bayesian estimation method requires less ad hoc determination of "weighting" parameters which appear in all but the most simplistic update techniques. Problems encountered when an improper finite element model is updated are addressed.