<p>This study proposes a novel, Artificial Intelligence (AI)-driven inverse design methodology for selecting constituent materials in brazed ceramic–metal composites, which, to the best of our knowledge, has not been reported before. Multiple AI algorithms, including Linear Regression (LR), Polynomial Regression (PR), Random Forest (RF), Artificial Neural Network (ANN), and a multi-output auto-encoder (AE) model, are developed. Eight input–output feature configurations are evaluated to select the single and multi-output parameters. The input–output feature comprises material properties, namely, the coefficient of thermal expansion and Young's modulus of brazed ceramic–metal composite materials obtained from literature, the strength parameter (average global stress represented by <InlineEquation ID="IEq1"> <EquationSource Format="TEX">${VMS}_{avg}$</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mrow> <mi mathvariant="italic">VMS</mi> </mrow> <mrow> <mi mathvariant="italic">avg</mi> </mrow> </msub> </math></EquationSource> </InlineEquation>) estimated from Finite Element Method (FEM) simulation for joint assembly structure containing porosity, and effective CTE value, <InlineEquation ID="IEq2"> <EquationSource Format="TEX">${\alpha }_{eff}$</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi>α</mi> <mrow> <mi mathvariant="italic">eff</mi> </mrow> </msub> </math></EquationSource> </InlineEquation>. The autoencoder (AE) model with a 32–16-32 structure outperforms LR, PR, RF, and ANN, achieving an Absolute Percentage Error (APE) of ~ 0.125–4.5%, compared to literature-reported values for predicting unseen data. The developed AE model accurately selects both single- and multi-output parameters, while a slightly lower performance is observed for ANN and PR in predicting low-importance and multi-output features. The selected brazed ceramic–metal materials from the developed AE model for a given <InlineEquation ID="IEq3"> <EquationSource Format="TEX">${VMS}_{avg}$</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mrow> <mi mathvariant="italic">VMS</mi> </mrow> <mrow> <mi mathvariant="italic">avg</mi> </mrow> </msub> </math></EquationSource> </InlineEquation> of 87.08 MPa corresponds to the alumina ceramic, Ag–Cu-Ti braze, and Kovar metal alloy.</p>

Artificial intelligence-driven methodology for predicting brazed ceramic–metal composite materials

期刊论文
出版模式：

开放获取
发表日期：

2026年4月28日

查看全文

Sunita Khod,
Vinay Kamma,
Ravi Kumar Varma,
Mayank Goswami

摘要

This study proposes a novel, Artificial Intelligence (AI)-driven inverse design methodology for selecting constituent materials in brazed ceramic–metal composites, which, to the best of our knowledge, has not been reported before. Multiple AI algorithms, including Linear Regression (LR), Polynomial Regression (PR), Random Forest (RF), Artificial Neural Network (ANN), and a multi-output auto-encoder (AE) model, are developed. Eight input–output feature configurations are evaluated to select the single and multi-output parameters. The input–output feature comprises material properties, namely, the coefficient of thermal expansion and Young's modulus of brazed ceramic–metal composite materials obtained from literature, the strength parameter (average global stress represented by ${VMS}_{avg}$ ${VMS}_{avg}$ ) estimated from Finite Element Method (FEM) simulation for joint assembly structure containing porosity, and effective CTE value, ${\alpha }_{eff}$ $α_{eff}$ . The autoencoder (AE) model with a 32–16-32 structure outperforms LR, PR, RF, and ANN, achieving an Absolute Percentage Error (APE) of ~ 0.125–4.5%, compared to literature-reported values for predicting unseen data. The developed AE model accurately selects both single- and multi-output parameters, while a slightly lower performance is observed for ANN and PR in predicting low-importance and multi-output features. The selected brazed ceramic–metal materials from the developed AE model for a given ${VMS}_{avg}$ ${VMS}_{avg}$ of 87.08 MPa corresponds to the alumina ceramic, Ag–Cu-Ti braze, and Kovar metal alloy.