Multiphysics Modeling of Voice Coil Actuators With Recurrent Neural Network

Abstract

In order to accurately model the behaviors of a voice coil actuator (VCA), the three-dimensional (3-D) method is preferred over a lumped model. However, building a 3-D model of a VCA is often very computationally expensive. The computation efficiency can be limited by the spatial discretization, the multiphysics nature, and the nonlinearities of the VCA. In this paper, we propose incorporating the recurrent neural network (RNN) into the multiphysics simulation to enhance its computation efficiency. In the proposed approach, the multiphysics problem is first solved with the finite element method (FEM) at full 3-D accuracy within a portion of the required time steps. An RNN is then trained and validated with the obtained transient solutions. Once the training completes, the RNN can make predictions on the transient behaviors of the VCA in the remaining portion of the required time steps. With the proposed approach, it avoids solving the 3-D multiphysics problem at all time steps such that a significant reduction of computation time can be achieved. The training cost of the RNN model can be amortized when a longer duration of transient behaviors is required. A loudspeaker example is used to demonstrate the enhancement of the computation efficiency by using RNN in the multiphysics modeling. Various structures of neural networks and tunable parameters are investigated with the numerical example in order to optimize the performance of the RNN model.