A Physics-Informed Neural Network for RRAM modeling

Sha，Yanliang; Ouyang，Lingyun; Chen，Quan

10.23919/ACES-China52398.2021.9581858

2021-07-28

978-1-6654-3377-8

2021 International Applied Computational Electromagnetics Society Symposium, ACES-China 2021, Proceedings

1-2

28-31 July 2021

Chengdu, China

Modeling of RRAM devices has been challenging due to the existence of the flux-dependent internal variables. Extra differential equations or capacitive circuits are often needed to model the evolution of the internal state variable and its impacts on the device responses, rendering the RRAM model difficult to develop and slow to evaluate. In this work, we propose a simple yet viable alternative to build steady-state RRAM compact models using physics-informed neural networks that do not involve internal state components. The central idea is to utilize a sequence of currents that are generated before as inputs, to account for the effect of the flux history that affects the current under the present voltage. The accuracy of the proposed model is applied to and verified by experimentally measured data.

Current measurement Computational modeling Neural networks Differential equations Rendering (computer graphics) Mathematical models Data models

第一

英语

EI

20214711198869

Computational electromagnetics ; Differential equations

Electricity and Magnetism:701 ; Data Storage, Equipment and Techniques:722.1 ; Calculus:921.2 ; Numerical Methods:921.6

2-s2.0-85119371620

Scopus

School of Microelectronics,Southern University of Science and Technology,China

Sha，Yanliang,Ouyang，Lingyun,Chen，Quan. A Physics-Informed Neural Network for RRAM modeling[C],2021:1-2.