Thermal Calculation Method Using 1D-CAE Coupled With a Surrogate Model for Thermal Printing Process

Abstract

This study proposes a method to integrate a surrogate model using neural networks into 1D-CAE (1 dimensional computer aided engineering) to accelerate thermal calculations in the thermal printing process. Thermal printing, which uses heat from resistive heating elements to print, requires precise and efficient computational models. In this research, we constructed a computational model of the thermal printing process using Modelica, creating head and media as separate components. The media component incorporates an RNN (recurrent neural network) using GRU (gated recurrent unit) to predict the complex causal relationships of heat flow and print indicator from head temperature and media speed. The RNN was trained using training data generated from an FEM (finite element method) model, completing the 1D-CAE computational model. This approach significantly reduces computation time compared to traditional FEM, providing an effective tool for examining complex thermal history control in thermal printing processes.