Industrial Chemistry
Open Access

Additive manufacturing; Machine learning; Defect prediction; Chemical composition; Quality control; Process optimization

Introduction

Additive manufacturing (AM), also known as 3D printing, has revolutionized the manufacturing industry by enabling the rapid production of complex parts with unprecedented design flexibility. However, one of the key challenges in additive manufacturing is the occurrence of defects, which can compromise the quality and integrity of printed components [1,2]. Addressing this challenge requires the development of robust predictive models capable of forecasting defect formation during the printing process. In recent years, machine learning (ML) techniques have emerged as powerful tools for defect prediction, with chemical composition serving as a critical input variable. This article explores the development and application of ML models that leverage chemical composition data to forecast defect occurrence in additive manufacturing processes [3,4]. Additive manufacturing (AM), commonly known as 3D printing, has emerged as a transformative technology in the manufacturing industry, offering unparalleled design freedom, reduced lead times, and increased efficiency compared to traditional manufacturing methods. However, despite its numerous advantages, AM processes are susceptible to the formation of defects, which can compromise the integrity and functionality of printed components [5]. Detecting and mitigating these defects is critical for ensuring the quality and reliability of AM-produced parts. One promising approach to address this challenge is the utilization of machine learning (ML) techniques to develop predictive models capable of forecasting defect occurrence in additive manufacturing processes [6,7]. These ML models leverage various input parameters, including chemical composition data, to identify patterns and correlations associated with defect formation. By analyzing the chemical composition of feedstock materials, ML models can predict the likelihood of defects such as porosity, cracking, and delamination during the printing process. In this introduction, we will explore the role of chemical composition in defect formation in additive manufacturing and discuss the motivation and objectives behind utilizing ML models for defect prediction [8]. Furthermore, we will provide an overview of the potential benefits of employing ML-based defect forecasting in additive manufacturing, as well as the challenges and considerations associated with developing and implementing such predictive models. Overall, this introduction sets the stage for understanding the significance of utilizing chemical composition data and ML techniques to forecast defect occurrence in additive manufacturing processes [9,10].

The role of chemical composition in defect formation

The chemical composition of feedstock materials used in additive manufacturing plays a crucial role in determining the likelihood of defect formation. Variations in composition, such as impurities, alloying elements, and processing additives, can influence material properties, microstructural characteristics, and ultimately, defect susceptibility. Common defects encountered in additive manufacturing processes include porosity, cracking, and delamination, which can arise due to inadequate melting, improper solidification, or residual stress accumulation. By analyzing the chemical composition of feedstock materials, it is possible to identify compositional factors that contribute to defect formation and develop predictive models to mitigate their occurrence.

Machine learning approaches for defect prediction

Machine learning techniques offer a data-driven approach to defect prediction by learning patterns and relationships from input data. In the context of additive manufacturing, ML models can be trained using chemical composition data, along with other process parameters and material properties, to predict the likelihood of defect occurrence in printed components. Supervised learning algorithms, such as random forests, support vector machines, and neural networks, are commonly used for defect classification and regression tasks. These models learn from labeled training data to make predictions on unseen test samples, allowing for the identification of defect-prone regions and the optimization of printing parameters to minimize defects.

Development and training of ml models

The development and training of ML models for defect prediction involve several key steps:

Data collection: Chemical composition data, along with relevant process parameters and defect labels, are collected from experimental measurements or simulation outputs.

Feature engineering: Input features, such as elemental compositions, alloy compositions, and processing parameters, are extracted from the raw data and preprocessed for model training.

Model selection: Suitable ML algorithms are selected based on the nature of the data and the specific prediction task, considering factors such as model complexity, interpretability, and performance.

Training and evaluation: The selected ML models are trained on a labeled dataset using optimization techniques such as gradient descent or genetic algorithms. Model performance is evaluated using metrics such as accuracy, precision, recall, and F1 score through cross-validation or holdout validation.

Model deployment: Trained ML models are deployed for real-time defect prediction in additive manufacturing processes, enabling proactive quality control and process optimization.

Applications and benefits of ml-based defect prediction

ML-based defect prediction models have wide-ranging applications in additive manufacturing, including:

Quality assurance: Early detection of defect-prone regions allows for real-time adjustments to printing parameters, minimizing defects and improving overall part quality.

Process optimization: ML models provide insights into the relationship between chemical composition, process parameters, and defect formation, enabling the optimization of printing conditions for enhanced performance and efficiency.

Material development: By correlating chemical composition with defect susceptibility, ML models can inform the design of new materials with improved printability and reliability.

Cost reduction: Predictive defect modeling reduces material waste, rework, and post-processing efforts, leading to cost savings and faster time-to-market for additive manufactured components.

Conclusion

Machine learning models utilizing chemical composition data offer a promising approach to forecast defect occurrence in additive manufacturing processes. By leveraging the rich information encoded in chemical compositions, ML models can provide actionable insights for quality assurance, process optimization, and material development in additive manufacturing. With continued advancements in data analytics, materials characterization, and computational modeling, ML-based defect prediction is poised to play a crucial role in advancing the reliability, efficiency, and scalability of additive manufacturing technologies.

References

1. Frölicher TL, Fischer E M, Gruber N (2018) Marine heatwaves under global warming. Nature 560: 360-364.

Indexed at, Google Scholar, Crossref

2. Kay J E (2020) Early climate models successfully predicted global warming. Nature 578: 45-46.

Indexed at, Google Scholar, Crossref

3. Traill LW, Lim LMM, Sodhi NS, BradshawCJA (2010) Mechanisms driving change: altered species interactions and ecosystem function through global warming. J Anim Ecol 79: 937-47.

Indexed at, Google Scholar, Crossref

4. Ross R (1986). The pathogenesis of atherosclerosis—an update.New England journal of medicine314: 488-500.

Indexed at, Google Scholar , Crossref

5. Duval C, Chinetti G, Trottein F, Fruchart J C, Staels B (2002) The role of PPARs in atherosclerosis.Trends Mol Med8: 422-430.

Indexed at, Google Scholar, Crossref

6. Libby P, Ridker P M, Maseri A (2002). Inflammation and atherosclerosis. Circulation 105: 1135-1143.

Google Scholar, Crossref

7. Kajinami K, Akao H, Polisecki E, Schaefer EJ (2005)Pharmacogenomics of statin responsiveness.Am J Cardiol 96: 65-70.

Indexed at, Google Scholar, Crossref

8. Kataoka Y, St John J, Wolski K, Uno K, Puri R, et al. (2015) Atheroma progression in hyporesponders to statin therapy. Arterioscler Thromb Vasc Biol 35: 990-995.

Indexed at, Google Scholar, Crossref

9. Zavodni AE, Wasserman BA, McClelland RL, Gomes AS, Folsom AR, et al .(2014) Carotid artery plaque morphology and composition in relation to incident cardiovascular events: the Multi‐Ethnic Study of Atherosclerosis (MESA). Radiology 271: 381–389.

Indexed at, Google Scholar Crossref

10. Polonsky TS, McClelland RL, Jorgensen NW, Bild DE, Burke GL, et al. (2010) Coronary artery calcium score and risk classification for coronary heart disease prediction. JAMA 303: 1610-1616.

Indexed at, Google Scholar, Crossref