Structure Optimization Design of Cantilever Beam in the Piezoresistive Acceleration Sensor Based on Artificial Neural Network

Piezoresistive micro-mechanical acceleration sensor is one of the earliest developed silicon micro-mechanical acceleration sensors in MEMS. Its cantilever beam is the main component. So the structure optimization design of the cantilever beam becomes the key. During the process of the former structure optimization of the cantilever beam, the working way between the program of FEM analysis and the optimizes belong to a kind of series connection, so we must design interface program in order to realize the data transmit between FEM and optimize program. This paper tries to replace FEM with neural network and carries on structure analysis on the cantilever beam in the piezoresistive sensor, FEM can be interpreted as a kind of relation of connecting with the program of optimizing in parallel, once after training the essential sample of network with FEM acquisition, the FEM program has no relations with the optimizing program. The next work is to train the network, put the well-trained network to the optimizing program and put into operation. Obtaining the FEM training samples, training the network, structure optimization are separated and interrelated. The whole procedure is very succinct.