Comparison and combination of different approaches for the model-based design of experiments under moderate errors
Ina Reichert, Peter Olney, Tom Lahmer
Bauhaus-Universitaet Weimar, Germany
In order to reduce for instance measurement errors while conducting experiments, it is wise to investigate the effect of multiple numbers of sensors as well as their positions to find the most suitable setup for the given purpose of the experiment. In order to accomplish this, mathematical optimization methods are used. These methods are employed on numerical simulations of structures and their behaviour. With the help of the numerical model the design of the experiment can be performed and later physical experiments can be used to validate the numerical solution. Within this extended abstract, two different procedures are followed in order to find the optimal experimental setup by using two different types of errors to create artificial measurements. First of all, the well-known Fisher Information Matrix is employed, where the parameter uncertainties are reduced. Secondly, the minimization of the mean-squared error between the estimated and the assumed true solution is applied. And finally, a combination of both methods is introduced, which leads to a constraint optimization problem.
ISSN 1611 - 4086 | © IKM 2015