A review on physics-informed machine learning for process-structure-property modeling in additive manufacturing - state-of-the-art review

Aug 5, 2026

This article presents a state-of-the-art review of physics-informed machine learning (PIML) models for process-structure-property modeling in additive manufacturing. Additive manufacturing enables complex, customized geometries, yet quality assurance requires accurate prediction of process-structure-property relationships. Pure physics-based models face high computational cost, large data demands, and limited interpretability, while ML can be data-hungry and black-box. PIML fuses physical knowledge with ML to improve robustness and interpretability. The review categorizes PIML approaches by how physical knowledge is integrated—feature engineering, architecture design, and physics-based loss modification—and compares studies and future research directions.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages32

Reading Time1 h 21 min

AccessFree to download

Introduction

Motivation and rationale

Journal of Manufacturing Processes 133 (2025) 524–555 Contents lists available at ScienceDirect Journal of Manufacturing Processes journal [homepage:](homepage: www.elsevier.com/locate/manpro)[ www.elsevier.com/locate/manpro](homepage: www.elsevier.com/locate/manpro) | Review article
A review on physics-informed machine learning for
process-structure-property modeling in additive manufacturing | | | | | --- | --- | --- | --- | | Meysam Faegha, Suyog Ghungrada, Jo˜ao Pedro Oliveira b, Prahalada Raoc, Azadeh Haghighia,*
a Department of Mechanical and Industrial Engineering, University of Illinois Chicago, Chicago, IL, USA
b CENIMAT/I3N, Department of Materials Science, NOVA School of Science and Technology, Universidade NOVA de Lisboa, Caparica 2829-516, Portugal c Grado Department of Industrial and Systems Engineering, Virginia Tech, Blacksburg, VA, USA | | | | | A R T I C L E I N F O | | A B S T R A C T | | | Keywords
Additive manufacturing
Physics-informed machine learning Process-structure-property relationships Physics-based feature engineering
Physics-based architecture Physics-based loss function | | This article presents a state-of-the-art review of the emerging field of physics-informed machine learning (PIML) models in additive manufacturing for process-structure-property modeling. Additive manufacturing processes hold immense potential for fabricating intricate and complex geometries across diverse applications and material classes. From a quality assurance standpoint, appropriate modeling of process-structure-property relationships of additive manufacturing processes using either physics-based or machine learning (ML)-based approaches has been a topic of intensive research. As an example, ML of data acquired from in-situ sensors is related to flaw formation, e.g., porosity, cracking, or deformation. In recent years, the computational burden of pure physicsbased models, the large data set requirement, and their black-box nature, i.e., the lack of interpretability of ML models, have prompted researchers to turn to PIML models. In PIML models, physical insights of the additive manufacturing process gained from various means are integrated with ML models, resulting in a more robust and interpretable framework for both process and microstructure evolution. A key delineator is the source of physical knowledge to be fused into PIML models, which can be obtained either from governing physical equations, datacentric feature extraction without implementing any physical equations, or a hybrid of the two foregoing. Within this review, we stratify PIML models based on the method used for the fusion of physical knowledge to ML models, into three categories, namely: (i) physics-based feature engineering, (ii) physics-based architecture shaping of ML models, and (iii) physics-based modification of the loss function of the ML models. For each of these categories, we further delineate the source of physical knowledge, ML models, integration approach, and data-set requirement, among others. A comparative analysis of the reviewed studies is presented and critically discussed, while the potential research gaps, along with future research directions on developing PIML models for different AM technologies are outlined. | | 1. Introduction 1.1. Motivation and rationale Additive manufacturing (AM) allows for the direct transformation of digital designs into physical objects through layer-by-layer deposition, joining, or solidification of materials without the need for molds, tools, or extensive manual labor [1]. In comparison with conventional manufacturing techniques, AM offers numerous advantages, includingthe ability to manufacture intricate, complex, and customized geometries at a significantly reduced cost, minimizing production time-tomarket, and realization of novel material compositions and designs with tailored functionalities [2]. Additionally, AM has demonstrated significant potential

FAQs

Why are physics-informed machine learning (PIML) models used for additive manufacturing process-structure-property modeling?

PIML models address limitations of pure physics-based models, such as high computational burden and large data requirements, and limitations of conventional ML, including black-box behavior and lack of interpretability. They integrate physical insights with ML for a more robust and interpretable framework.

What are the three main categories of PIML models discussed in the review?

The review stratifies PIML models into three categories: (i) physics-based feature engineering, (ii) physics-based architecture shaping of ML models, and (iii) physics-based modification of the loss function.

How does the source of physical knowledge affect the design of PIML models?

The physical knowledge to be fused can come from governing physical equations, data-centric feature extraction without explicit physical equations, or a hybrid of both. The review discusses how this choice influences integration approach and data-set requirements across categories.

Aug 5, 2026

Upload by Seraphina