Scholarlink

<p>Additive manufacturing (AM) of lightweight alloys such as AlSi10Mg and Ti-6Al-4&#xa0;V has gained considerable attention in aerospace, biomedical, and automotive applications due to its ability to enable weight reduction, geometric complexity, and efficient material utilization. However, the quality and reliability of AM-fabricated components remain highly sensitive to nonlinear and strongly coupled process parameters, including laser power, scan speed, and layer thickness, which complicates process optimization using conventional empirical approaches. To address these challenges, artificial intelligence (AI) techniques and metaheuristic optimization methods most notably artificial neural networks (ANN), genetic algorithms (GA), and particle swarm optimization (PSO) have been increasingly adopted for process modelling, prediction, and control. This review critically examines both standalone approaches (ANN, GA, PSO) and hybrid frameworks (e.g., ANN–GA, PSO–SVM), with emphasis on their effectiveness in optimizing key outcomes such as density, porosity, mechanical performance, and thermal behaviour. Reported studies demonstrate strong predictive capability, with ANN-based models commonly achieving coefficients of determination (R²) in the range of 0.89–0.93 for porosity and density prediction, alongside defect or error reductions of up to approximately 30% in selected industrial case studies. Beyond conventional data-driven models, emerging directions are discussed, including physics-informed neural networks (PINNs), AI-enabled digital twin architectures that couple virtual and physical manufacturing systems, and lightweight Edge AI solutions for real-time process monitoring and control. By synthesizing insights from materials science, machine learning, and advanced manufacturing, this review highlights current limitations related to generalization, benchmarking, and interpretability, and outlines future opportunities in explainable AI (XAI), multi-objective optimization, and cross-domain model transfer. The work ultimately proposes a structured roadmap toward more robust, interpretable, and sustainable AI-driven optimization frameworks for next-generation additive manufacturing systems.</p> Graphical Abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

AI and Metaheuristic Optimization in Additive Manufacturing of Lightweight Alloys: A Critical Review

  • 期刊论文

  • 发表日期：

    2026年3月9日

查看全文
  • Aswin Karkadakattil

摘要

Additive manufacturing (AM) of lightweight alloys such as AlSi10Mg and Ti-6Al-4 V has gained considerable attention in aerospace, biomedical, and automotive applications due to its ability to enable weight reduction, geometric complexity, and efficient material utilization. However, the quality and reliability of AM-fabricated components remain highly sensitive to nonlinear and strongly coupled process parameters, including laser power, scan speed, and layer thickness, which complicates process optimization using conventional empirical approaches. To address these challenges, artificial intelligence (AI) techniques and metaheuristic optimization methods most notably artificial neural networks (ANN), genetic algorithms (GA), and particle swarm optimization (PSO) have been increasingly adopted for process modelling, prediction, and control. This review critically examines both standalone approaches (ANN, GA, PSO) and hybrid frameworks (e.g., ANN–GA, PSO–SVM), with emphasis on their effectiveness in optimizing key outcomes such as density, porosity, mechanical performance, and thermal behaviour. Reported studies demonstrate strong predictive capability, with ANN-based models commonly achieving coefficients of determination (R²) in the range of 0.89–0.93 for porosity and density prediction, alongside defect or error reductions of up to approximately 30% in selected industrial case studies. Beyond conventional data-driven models, emerging directions are discussed, including physics-informed neural networks (PINNs), AI-enabled digital twin architectures that couple virtual and physical manufacturing systems, and lightweight Edge AI solutions for real-time process monitoring and control. By synthesizing insights from materials science, machine learning, and advanced manufacturing, this review highlights current limitations related to generalization, benchmarking, and interpretability, and outlines future opportunities in explainable AI (XAI), multi-objective optimization, and cross-domain model transfer. The work ultimately proposes a structured roadmap toward more robust, interpretable, and sustainable AI-driven optimization frameworks for next-generation additive manufacturing systems.

Graphical Abstract