Sensor Placement using Fisher Information Matrix for Robust Finite Element Model Updating

Three methods are presented to reduce the influence of measurement errors in parameter estimation in finite element model updating for structural health monitoring and damage assessment. First, a method using the Fisher information matrix is developed to choose an efficient set of measurement locations. This ensures efficient setup of a non-destructive test for finite element model updating for a given set of unknown parameters. Second, a normalization scheme is presented that, generally speaking, weighs data in a manner that varies inversely to the level of measurement error. This normalization results in a final estimate that is very close to the maximum likelihood estimator of the unknown parameters. Finally, the Fisher information matrix and its inverse, the CramerRao lower bound covariance matrix, are used to quantify the uncertainty in the final estimates. Numerical examples showed the proposed methods are effective in improving observability and accuracy of parameter estimates in finite element model updating. Minimizing the effect of measurement errors and their propagation in parameter estimates can greatly improve the finite element model updating for structural health monitoring.