Heterogeneous sensing and scientific machine learning for quality assurance in laser powder bed fusion – A single-track study

Gaikwad, Aniruddha; Giera, Brian; Guss, Gabe; Forien, Jean‐Baptiste; Matthews, Manyalibo J.; Rao, Prahalada

doi:10.1016/j.addma.2020.101659

Cited by 56 publications

(20 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Supervised learning algorithms refer to machine learning methods in which models are trained using labels. Typical supervised learning methods used in digital twin include supper vector machine (SVM) [ 195 , 201 ], decision trees [ 86 , 93 , 94 ], k-nearest neighbors [ 102 ], convolutional neural networks (CNN) [ 103 , 135 , 202 , 206 , 270 ] and recurrent neural networks (RNN) [ 202 ]. In practice, data labeling can be an expensive task.…”

Section: Discussionmentioning

confidence: 99%

“…As Gaikward et al in [ 201 ] presented the temperature distribution of parts, predicted based on a graph–theoretical computational heat transfer approach and subsequently combined it with an SVM model in order to detect potential quality faults in printing processes. Another attempt of grey box modeling for build quality in dependency of process parameters and in situ sensor signatures was proposed in [ 202 ] by Gaikward et al, where the a priori knowledge of physical processes was incorporated into three sequentially connected shallow ANNs and consequently achieved better performance in comparison with purely data-driven methods (CNN, LSTM, RNN, among others). With respect to DfX and knowledge engineering, Ko et al employed CART to predict additive manufacturability, which was further fed back to a knowledge-query formulation phase in order to continuously construct and broaden an AM knowledge base [ 203 ] ( EG -factor).…”

Section: Sustainable Resilient Manufacturingmentioning

confidence: 99%

See 1 more Smart Citation

A Survey on AI-Driven Digital Twins in Industry 4.0: Smart Manufacturing and Advanced Robotics

Huang

Shen²,

et al. 2021

Sensors

162

View full text Add to dashboard Cite

Digital twin (DT) and artificial intelligence (AI) technologies have grown rapidly in recent years and are considered by both academia and industry to be key enablers for Industry 4.0. As a digital replica of a physical entity, the basis of DT is the infrastructure and data, the core is the algorithm and model, and the application is the software and service. The grounding of DT and AI in industrial sectors is even more dependent on the systematic and in-depth integration of domain-specific expertise. This survey comprehensively reviews over 300 manuscripts on AI-driven DT technologies of Industry 4.0 used over the past five years and summarizes their general developments and the current state of AI-integration in the fields of smart manufacturing and advanced robotics. These cover conventional sophisticated metal machining and industrial automation as well as emerging techniques, such as 3D printing and human–robot interaction/cooperation. Furthermore, advantages of AI-driven DTs in the context of sustainable development are elaborated. Practical challenges and development prospects of AI-driven DTs are discussed with a respective focus on different levels. A route for AI-integration in multiscale/fidelity DTs with multiscale/fidelity data sources in Industry 4.0 is outlined.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Sustainable Resilient Manufacturingmentioning

confidence: 99%

A Survey on AI-Driven Digital Twins in Industry 4.0: Smart Manufacturing and Advanced Robotics

Huang

Shen²,

et al. 2021

Sensors

162

View full text Add to dashboard Cite

show abstract

“…Yuan et al employed a machine learning based monitoring of laser powder bed fusion of 316L stainless steel based on in-situ laser single-track videos, enabling the prediction of laser powder bed fusion track widths with an R 2 of 0.93 [23]. Gaikwad et al developed machine learning based predictive models of single tracks of 316L stainless steel using pyrometer and high-speed video camera data [24]. Their best ML prediction for meltpool width is achieved by a sequential decision analysis neural network (SeDANN) model with an R 2 of 0.87.…”

Section: Introductionmentioning

confidence: 99%

MeltpoolNet: Melt pool Characteristic Prediction in Metal Additive Manufacturing Using Machine Learning

Parand¹,

Ogoke²,

Kao³

et al. 2022

Preprint

View full text Add to dashboard Cite

Characterizing meltpool shape and geometry is essential in metal Additive Manufacturing (MAM) to control the printing process and avoid defects. Predicting meltpool flaws based on process parameters and powder material is difficult due to the complex nature of MAM process. Machine learning (ML) techniques can be useful in connecting process parameters to the type of flaws in the meltpool. In this work, we introduced a comprehensive framework for benchmarking ML for melt pool characterization. An extensive experimental dataset has been collected from more than 80 MAM articles containing MAM processing conditions, materials, meltpool dimensions, meltpool modes and flaw types. We introduced physics-aware MAM featurization, versatile ML models, and evaluation metrics to create a comprehensive learning framework for meltpool defect and geometry prediction. This benchmark can serve as a basis for melt pool control and process optimization. In addition, data-driven explicit models have been identified to estimate meltpool geometry from process parameters and material properties which outperform Rosenthal estimation for meltpool geometry while maintaining interpretability.

show abstract

“…Li et al studied a deep learning-based process monitoring method and a quality identification method for the metal AM process 15 , 16 . Gaikwad et al studied heterogeneous sensing and machine learning for single-track quality in laser powder bed fusion 17 . Feenstra et al demonstrated the complex interactions and relationships between the parameters using artificial neural networks in DED processes for Inconel 625, Hastelloy X, and stainless steel 316 L 18 .…”

Section: Introductionmentioning

confidence: 99%

Selection of effective manufacturing conditions for directed energy deposition process using machine learning methods

Lim

Lee

et al. 2021

Sci Rep

View full text Add to dashboard Cite

In the directed energy deposition (DED) process, significant empirical testing is required to select the optimal process parameters. In this study, single-track experiments were conducted using laser power and scan speed as parameters in the DED process for titanium alloys. The results of the experiment confirmed that the deposited surface color appeared differently depending on the process parameters. Cross-sectional view, hardness, microstructure, and component analyses were performed according to the color data, and a color suitable for additive manufacturing was selected. Random forest (RF) and support vector machine multi-classification models were constructed by collecting surface color data from a titanium alloy deposited on a single track; the accuracies of the multi-classification models were compared. Validation experiments were performed under conditions that each model predicted differently. According to the results of the validation experiments, the RF multi-classification model was the most accurate.

show abstract

Heterogeneous sensing and scientific machine learning for quality assurance in laser powder bed fusion – A single-track study

Cited by 56 publications

References 45 publications

A Survey on AI-Driven Digital Twins in Industry 4.0: Smart Manufacturing and Advanced Robotics

A Survey on AI-Driven Digital Twins in Industry 4.0: Smart Manufacturing and Advanced Robotics

MeltpoolNet: Melt pool Characteristic Prediction in Metal Additive Manufacturing Using Machine Learning

Selection of effective manufacturing conditions for directed energy deposition process using machine learning methods

Contact Info

Product

Resources

About