Christian Adams

EFFICIENT EXPERIMENTAL VALIDATION OF STOCHASTIC SENSITIVITY ANALYSES OF SMART SYSTEMS

A method for the efficient experimental validation of stochastic sensitivity analyses is proposed and tested using a smart system for vibration reduction. Stochastic analyses are needed to assess the reliability and robustness of smart systems. A model-based design of experiments combines an experimental design with the results of a previous numerical sensitivity analysis. To test this method, a system of structural dynamics is used. Active suppression of disturbing vibrations of a cantilever beam by means of active piezoelectric elements is considered. The observed target variables are the level of vibration reduction at the beam’s end and the fundamental frequency considering five uncertain system variables. Based on a numerical model of the piezoelectric beam, a variance-based sensitivity analysis is performed to determine each design variable’s impact on the target variables. According to these numerical results, a model-based experimental design is established and the experiments are conducted. In comparison to a fully five-factor factorial experimental design, the model-based approach reduced the experimental effort by 50%, without great loss of information.