Artificial Neural Network modeling of MEMS cantilever resonator using Levenberg Marquardt algorithm

Modeling and simulation are two important stages in the design and development of MEMS devices. Accurate modeling of the device is very essential for the system level simulation. This paper presents the Artificial Neural Network (ANN) modeling of MEMS cantilever resonator. The geometrical parameters (length, thickness, width and gap) of the cantilever resonator and bias voltage are considered as the input parameters. The resonance frequency, displacement and motional current are taken as the output parameters. The variation of the output parameters with the variation in the input parameters is obtained from the MEMS simulation software Coventorware. Using this data an ANN model is developed for the cantilever resonator. The Levenberg-Marquardt (LM) algorithm is used for training the ANN model.