Data-Driven State Estimation of Carbon Nanotube Field Effect Transistor with Smart RBF Network

Abstract

Since 1993, Devices based on CNTs have applicationsranging from nanoelectronics to optoelectronics. Thechallenging issue in designing these devices is that thenonequilibrium Green's function (NEGF) method has tobe employed to solve the Schrödinger and Poissonequations, which is complex and time consuming. In thepresent study, a novel smart and optimal algorithm ispresented for fast and accurate modeling of CNT fieldeffecttransistors (CNTFETs) based on an artificial neuralnetwork. A new and efficient way is presented forincrementally constructing radial basis function (RBF)networks with optimized neuron radii to obtain theestimator network. An incremental extreme learningmachine (I-ELM) algorithm is used to train the RBFnetwork. To ensure the optimal radii for incrementalneurons, this algorithm utilizes a modified version of anoptimization algorithm known as the Nelder-Meadsimplex algorithm. Results confirm that the proposedapproach reduces the network size for faster errorconvergence while preserving the estimation accuracy.

Keywords

• cntfet
• modeling
• nonlinear system
• rbf
• state estimation