Sensitivity analysis modelling for microscale multiphysics robust engineering design

Abstract

Sensitivity Analysis (SA) plays an important role in the development of any practical engineering model. It can help to reveal the sources and mechanisms of variability that provide the key to understanding system uncertainty. SA can also be used to calibrate simulation models for closer agreement with experimental results. Robust Engineering Design (RED) seeks to exploit such knowledge in the search for design solutions that are optimal in terms of performance in the face of variability. Microscale and multiphysics problems present challenges to modelling due to their complexity, which puts increased demands on computational methods. For example, in developing a model of a piezoelectric actuator, the process of calibration is prolonged by the number of parameters that are difficult to verify with the physical device. In the approach presented in this paper, normalised sensitivity coefficients are determined directly and accurately using the governing finite element model formulation, offering an efficient means of identifying parameters that affect the output of the model, leading to increased accuracy and knowledge of system performance in the face of variability.