TL;DR
This article presents how ML approaches may speed up material discovery, reduce trial-and-error, and enable individualized design solutions for 3D-printed polymers across a variety of contexts.
Machine learning (ML) techniques are increasingly being applied to establish correlations between input parameters and key process responses in the wire arc additive manufacturing (WAAM) process. Despite their potential, there remains limited understanding of how to develop an integrated ML framework that simultaneously considers both the dataset characteristics and the modeling approach to ensure accurate and reliable predictions. The present study addresses this gap by developing an integrated ML framework to predict the deposition behavior of Inconel 625 in WAAM. To capture nonlinear system behavior, three ML methods, namely artificial neural network (ANN), support vector machine (SVM), and adaptive neuro-fuzzy inference system (ANFIS), were developed and systematically evaluated for predictive modeling and process optimization, considering deposited geometry, area, and efficiency as the key output characteristics. The input parameters, i.e., voltage, wire feed rate, torch travel speed, and shielding gas flow rate, were identified as critical factors influencing the deposition process. The datasets were preprocessed to remove noise and analyzed to extract relevant features that captured the intrinsic physical behavior of the process. Performances of the ML models were evaluated using a separate test dataset, and predictions were assessed through mean absolute percentage deviation (MAPD). Results demonstrated that integrated ML framework could accurately represent intricate interdependencies among process parameters on deposition outcomes, providing a robust method of predictive modeling and parametric process optimization for Inconel 625 deposition by WAAM process. The ANN model demonstrated satisfactory performance for forward modeling with MAPD values of 12.24, 14.87, and 11.91 for deposition geometry, deposition area, and deposition efficiency, respectively. For inverse modeling, the ANN accurately predicted key inputs from outputs, with MAPD values of 1.39, 18.91, 12.25, and 19.36 for voltage, wire feed rate, torch speed, and shielding gas flow rate, respectively. Bidirectional predictive modeling keeps to set operating conditions to achieve desired depositions and process automations.
Significant advances have been made in the field of Additive Manufacturing (AM) across various manufacturing processes. Laser Powder Bed Fusion (LPBF) is one of the most widely adopted metal AM processes in the industry, which employs energy sources such as lasers to melt powder materials. While the LPBF process offers numerous advantages, such as the ability to produce complex geometries and multiple parts simultaneously, it also presents challenges in the form of defects such as porosity, residual stresses, and cracks that need to be addressed. Conducting experiments to identify and eliminate these defects can be prohibitively expensive, motivating the development of alternative predictive approaches. Numerical modeling and simulation techniques provide a cost-effective alternative to experimental approaches, enabling analyses ranging from thermal histories and microstructural evolution to mechanical property prediction and defect identification in fabricated components. However, such simulations can also be computationally intensive. In recent years, Machine Learning (ML) and Artificial Intelligence (AI) algorithms have emerged as viable alternative tools to accelerate process understanding, defect prediction, and parameter optimization. This review provides a comprehensive overview of ML-based approaches applied to LPBF, emphasizing the importance of understanding the underlying physical phenomena that significantly influence data collection, preprocessing, and feature engineering strategies essential for effective model training and validation. It offers a structured framework for researchers seeking to leverage ML methods to enhance predictive accuracy and computational efficiency, particularly through simulation-driven data generation for model development.
A novel, data-driven framework for predicting and optimizing the mechanical performance of 3D-printed polylactic acid (PLA) composites reinforced with date pit (DP) particles under controlled annealing conditions is presented, enabling simultaneous property prediction and design optimization from a minimal experimental dataset.
This study examines the application of ML techniques to composite materials, particularly for predicting fracture toughness, characterizing damage, and optimizing mechanical properties and reveals significant relationships between fracture toughness and important input parameters.
The paper presents an approach to constructing predictive models for the physical and mechanical properties of elastomeric composites using machine learning methods. The relevance of the study is driven by the need to accelerate the development of new materials and reduce the labor intensity of full-scale experiments. An automated machine learning algorithm is proposed, encompassing stages of input data unification, feature space formation, and comparative analysis of regression models (Random Forest, Gradient Boosted Decision Trees, Gradient Tree-Boosting Tweedie, Poisson Regression, Light Gradient Boosting Machine и Stochastic Dual Coordinate Ascent). During experimental validation on datasets containing formulation data with varying content of sulfur, natural rubber (NR), and zinc oxide, predictions were made for theoretical density, Karrer plasticity, brittleness temperature, and curing temperature. It was established that ensemble methods demonstrate the highest predictive capability; however, model accuracy significantly depends on sample representativeness. Intervals of the studied parameters (particularly the 140–150 °C range for curing temperature) characterized by increased prediction uncertainty were identified, requiring additional algorithm calibration. The obtained results confirm the effectiveness of the proposed approach for formulation optimization and identification of hidden dependencies in the “composition-property” system.
Accurate prediction and optimization of polymer composite properties is of paramount importance in the design of these lightweight, durable, and sustainable materials within renewable energy technologies. This work will provide a holistic machine learning-assisted framework that unites materials informatics with domain-specific features and state-of-the-art ML methodologies in the prediction of the mechanical properties of polymer composites, such as tensile strength. This includes embedding several ensemble models, including Random Forest and Gradient Boosting, kernel methods such as SVR, neural networks, graph-based approaches, while it applies principled hyperparameter tuning and uncertainty quantification, along with model-interpretability tools such as SHAP and systematic ablation to identify the most important material and processing factors. The proposed methodology is demonstrated with curated, publicly available datasets, and we discuss means for synthetic data augmentation, cross-validation protocols, assessment of model robustness, and best practices in reporting results in a reproducible way. The results show that data-driven models reduce prediction error and speed up the processes of materials selection in view of Net Zero targets, given limited available experimental iterations; this informs intelligent lightweighting of renewable energy components. Furthermore, the study highlights the capability of machine learning models to capture complex nonlinear relationships between composition, processing parameters, and mechanical response that are difficult to address using conventional trial-and-error approaches. By reducing reliance on extensive experimental campaigns, the proposed framework supports faster design cycles and more efficient utilization of material and energy resources in renewable energy applications.