Comparison of Different Approaches for the Model-Based Design of Experiments

Before starting an experiment it is wise to make close investigations on the structure to be examined and the possible designs of the experiment. The optimal design of the measurement setups is acquired by using mathematical optimization methods which are supported by numerical simulations of the structure and its behavior. The numerical model can then be used to verify the resulting design of the experiment and both of them can be validated by monitoring an existing structure. The aim of this paper is to undertake this first step to confirm the optimal design of experiments which is gained by three different approaches and testing them on a numerical model. These methods are: the reduction of parameter uncertainties by using the Fisher Information Matrix, the second approach is made by the minimization of the mean-squared errors between the assumed true solution and the solution of the inverse problem and as third approach the so-called sigma-point method is used where for the mean-squared error biased estimators for the parameter identification problem are used.