Abstract
CAD and CAE approach have accelerated sensor design and optimization. But physical sensors tend to be accompanied with multi-physics interaction, making it hard to perform accurate 3D simulation of the sensors. Compared to 3D, 2D simulation’s computing expenses have decreased by 85%, while bringing about a maximum deviation up to 92%. In this paper, a methodology to decrease the deviation while simplifying 3D simulations of cantilever-based sensors into 2D ones are proposed and verified with a piezoelectric current sensor. To make up for the deviation, firstly, comprehensive 2D and 3D simulations are performed within a broad design space. Then, the deviation between 2D and 3D simulation under different design parameters are calculated. Finally, the design parameters and the deviation are used to train a neural network. After training, when new design parameters are input into the network, the corresponding deviation can quickly be predicted and used to revise 2D coarse simulation result. As a result, the deviation between 2D and 3D simulation is below 0.34%, which is a two order of magnitude decreasing compared to the result before being revised. This methodology can not only be used for the design and optimization of piezoelectric current sensor’s, but also can be potentially useful for other cantilever-based sensors, like piezoelectric mass and force sensor.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call