Data-Driven Modeling for Multiphysics Parametrized Problems-Application to Induction Hardening Process

Khouloud Derouiche,Monzer Daoud,Khalil Traidi,Sevan Garois,Francisco Chinesta,Victor Champaney

doi:10.3390/met11050738

Khouloud Derouiche, Monzer Daoud + Show 4 more

Open Access

https://doi.org/10.3390/met11050738

Copy DOI

Abstract

Data-driven modeling provides an efficient approach to compute approximate solutions for complex multiphysics parametrized problems such as induction hardening (IH) process. Basically, some physical quantities of interest (QoI) related to the IH process will be evaluated under real-time constraint, without any explicit knowledge of the physical behavior of the system. Hence, computationally expensive finite element models will be replaced by a parametric solution, called metamodel. Two data-driven models for temporal evolution of temperature and austenite phase transformation, during induction heating, were first developed by using the proper orthogonal decomposition based reduced-order model followed by a nonlinear regression method for temperature field and a classification combined with regression for austenite evolution. Then, data-driven and hybrid models were created to predict hardness, after quenching. It is shown that the results of artificial intelligence models are promising and provide good approximations in the low-data limit case.

Highlights

The aeronautical and automotive industries would eventually like to lighten the mechanical structures in order to meet the relative environmental concerns to carbon dioxide emissions and fuel economy while maintaining optimal mechanical properties and performances
Based on the set of computed snapshots, collected at some sparse points in the space domain and for different values of frequency and power, a dimensionality reduction approach based on proper orthogonal decomposition (POD) [13] followed by a machine learning algorithm for regression was applied in order to create an approximate solution for the temperature field evolution
The standardization of the input parameters was applied to avoid problems related to units and different scaled variables, the datasets were split into training and testing subsets (75% of data to build the models and 25% to evaluate the prediction accuracy)

Summary

Introduction

The aeronautical and automotive industries would eventually like to lighten the mechanical structures in order to meet the relative environmental concerns to carbon dioxide emissions and fuel economy while maintaining optimal mechanical properties and performances. For many applications, only superficial layer material properties play an important role In this context, surface treatments of industrial components by mechanical, thermal, or thermochemical means are suitable [2,3]. Two main steps are at the origin of the IH: an electromagnetic induction heating and subsequent quenching. It has the advantage of providing a very short surface heat-up time, a good fatigue performance, a precise control of the treated zone, a good reproducibility, and an operating mode compatible with severe environmental requirements, unlike thermochemical treatments such as carburizing and carbonitriding [6]. From the industrial point of view, the challenges behind the development of such methodologies are:

Methods

Results

Discussion

Conclusion