Application of Machine Learning Method for Calculating the Instability Voltage of Nano- Beam Actuator in the Presence of Electrostatic Force and Considering vdW and Casimir Forces

Abstract

Abstract Considering the significant importance of investigating of pull-in instability and determining the pull-in voltage in nano-sensors and nano-actuators, in this article, the determination of the pull-in voltage of these structures is discussed based on the machine learning method. MLP neural network and SVR methods, which have good ability to estimate data and regression, are considered for this purpose. In this regard, the number of 500 data have been considered for training of these approaches. For the training process, the pull-in voltage has been set as the target and the physical and geometric characteristics of nanostructures are considered as inputs. The exact value of pull-in voltage for training has been determined using the SSLM together with Galerkin methods; where is a reliable procedure to determine of pull-in voltage. The type of employed MLP is feed forward back propagation and its utilized learning process is the Levenberg Marquardt. The number of layers and neurons for each layer have been checked by practicing different configurations, the most optimal mode includes two hidden layers and the number of 10 and 8 neurons in the first and second hidden layers. Also, SVR with RBF kernel has been also used. By comparing the performance of two methods, it was found that MLP has a relatively good ability to estimate the pull-in voltage. Also, the capability of neural networks in determining the pull-in voltage has been examined according to the results presented in the previous reference, and the characteristics of these structures were not considered in the training process of the machine learning method. The obtained results show the very good capability of the neural network in determining the pull-in voltage of nanostructures in previous study as well.