Mechanistic machine learning for metamaterial fatigue strength design from first principles in additive manufacturing - Abstract

Aug 3, 2026

Upload by Sage

Digital control in manufacturing processes generates rich metadata from thermal and optical measurements, enabling property grading and improved feedback for fault detection. This study investigates metadata-driven design of fatigue-resistant structures by combining physically grounded models—density functional theory, cyclic plasticity, and fracture mechanics—with machine learning. ML models learn efficiently within their physical feature space, while mechanistic fatigue modeling is computationally demanding. An energy-based mechanistic function quantifies flaw effects on fatigue lifetime under specified boundary loads and supports probabilistic lifetime regression at larger prediction scales. Selective laser melting is analyzed using its digital control and in-situ metadata availability.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages18

Reading Time45 min

AccessFree to download

Introduction

Background and prior work on fatigue modeling

Mechanistic and machine-learning framework

Energy-based mechanistic fatigue function

Probabilistic regression across scales

Application to selective laser melting

Materials & Design 241 (2024) 112889 Contents lists available at ScienceDirect Materials & Design journal [homepage: www.elsevier.com/locate/matdes](homepage: www.elsevier.com/locate/matdes) | Mechanistic machine learning for metamaterial fatigue strength design from ﬁrst principles in additive manufacturing
Mustafa Awd a,∗ , Lobna Saeedb, Sebastian Münstermann c, Matthias Faes d, Frank Walther a | | | | | --- | --- | --- | --- | | a Chair of Materials Test Engineering (WPT), TU Dortmund University, Baroper Str. 303, Dortmund, D-44227, Germany
b Crystallography and Geomaterials Research Group, Faculty of Geosciences, University of Bremen, Klagenfurter Str. 2-4, Bremen, D-28359, Germany c Institute for Metal Forming, RWTH Aachen University, Intzestr. 10, Aachen, D-52072, Germany
d Chair for Reliability Engineering (CRE), TU Dortmund University, Leonhard-Euler Str. 5, Dortmund, D-44227, Germany | | | | | A R T I C L E I N F O | | A B S T R A C T | | | Keywords:
Density functional theory (DFT)
Functional grading Machine learning (ML) Fatigue strength
Cohesion energy
Additive manufacturing (AM) Lifetime prediction Bayesian statistics
Process monitoring Artiﬁcial intelligence (AI) | | Digital control in manufacturing processes produces signiﬁcant amounts of metadata. The production process metadata, such as thermal and optical measurements, enables a higher degree of property grading than uninstrumented manufacturing and feedback for fault detection. This study explores how metadata can design fatigue-resistant structures using physically grounded models such as density functional theory, cyclic plasticity, and fracture mechanics that train machine learning algorithms. Machine learning models work very eﬃcientlyin their trained physical space. In comparison, mechanistic models are computationally costly for complex phenomena such as fatigue. We show how fatigue can be administered consistently at all scales by energy-based criteria and how a mechanistic function can be built based on this concept. The energy mechanistic function allows exact quantiﬁcation of the eﬀect of the existing ﬂaws from manufacturing on fatigue lifetime under certain load boundary conditions. Since the mechanistic function is local and subscale to the prediction scale of the machine learning model, it can be used to build density functions for probabilistic regression of the fatigue property on the scale above. The analysis is applied to the selective laser melting process due to the availability of digital control and metadata generation during deposition. | | 1. Introduction There are beneﬁts to additive manufacturing (AM) over conventional subtractive manufacturing methods, but more structural qualiﬁcation is needed for technical alloys’ cyclic deformation and fatigue characteristics. Increasing high-cycle fatigue resistance in metallic materials under cyclic loads is essential for structurally demanding applications. Often, eﬀorts to improve fatigue damage resistance using traditional strengthening methods and hierarchical nanostructural alterations are unsuccessful. However, physics-based methods appearing in the recent literature promise signiﬁcant breakthroughs when coupled with high throughput machine learning and digital twinning techniques. In pure Cu, low-angle dislocation barriers are introduced at the nanoscale, improving tensile strength and fatigue limit. The increased fatigue life results from the high density of built-in low-angle dislocation cells, which reduces surface roughening and crack initiation [1]. The fatigue life of materials with minor ﬂaws was estimated when subjected to multiple loads, and a notch factor and Murakami crite- * Corresponding author. [E-mail address:](E-mail address: mustafa.awd@tu-dortmund.de)[ mustafa.awd@tu-dortmund.de](E-mail address: mustafa.awd@tu-dortmund.de) (M. Awd). rion calibrated the Fatemi-Socie, Smith-Watson-Topper, Susmel, and Lazarin critical plane

FAQs

How does the study use manufacturing metadata to design fatigue-resistant structures?

It leverages thermal and optical measurements produced by digital control to support property grading and to enable feedback for fault detection, which then feeds physically grounded modeling and machine learning for fatigue-resistant design.

Why combine density functional theory and mechanistic fatigue criteria with machine learning?

Machine learning models work efficiently within the trained physical feature space, while mechanistic fatigue criteria are costly for complex phenomena. The approach uses physically grounded models to provide transferable physical meaning and to build an energy-based mechanistic function for fatigue lifetime quantification.

What is the role of the energy-based mechanistic function in fatigue lifetime prediction?

It provides exact quantification of how manufacturing-induced flaws affect fatigue lifetime under given load boundary conditions. Because the function is local and operates below the ML prediction scale, it enables density functions for probabilistic regression at higher scales.

Aug 3, 2026

Upload by Sage