A multi-scale convolutional neural network for autonomous anomaly detection and classification in a laser powder bed fusion additive manufacturing process

Luke Scime; Jack Beuth

doi:10.1016/j.addma.2018.09.034

A multi-scale convolutional neural network for autonomous anomaly detection and classification in a laser powder bed fusion additive manufacturing process

Luke Scime, Jack Beuth

Research output: Contribution to journal › Article › peer-review

279 Scopus citations

Abstract

In-situ detection of processing defects is a critical challenge for Laser Powder Bed Fusion Additive Manufacturing. Many of these defects are related to interactions between the recoater blade, which spreads the powder, and the powder bed. This work leverages Deep Learning, specifically a Convolutional Neural Network (CNN), for autonomous detection and classification of many of these spreading anomalies. Importantly, the input layer of the CNN is modified to enable the algorithm to learn both the appearance of the powder bed anomalies as well as key contextual information at multiple size scales. These modifications to the CNN architecture are shown to improve the flexibility and overall classification accuracy of the algorithm while mitigating many human biases. A case study is used to demonstrate the utility of the presented methodology and the overall performance is shown to be superior to that of methodologies previously reported by the authors.

Original language	English
Pages (from-to)	273-286
Number of pages	14
Journal	Additive Manufacturing
Volume	24
DOIs	https://doi.org/10.1016/j.addma.2018.09.034
State	Published - Dec 2018
Externally published	Yes

Funding

The build used as a case study in Section 5 was funded under a US Department of Energy grant (DE-FE0024064) and performed for Dr. Samikshya Subedi and Prof. Anthony Rollett of CMU's Materials Science and Engineering Department in collaboration with Oregon State University and Prof. Vinod Narayanan at the University of California Davis. The authors would like to thank Dr. Colt Montgomery and Todd Bear (CMU) for their role in running many of the builds analyzed during the development of the ML methodologies as well as Nicholas Jones, Edgar Mendoza, and Rahi Patel (CMU) for their assistance in user-testing the powder bed monitoring software. Finally, the authors would like to thank Dr. Brian Fisher, Prof. Elizabeth Holm, Prof. Levent Burak Kara, and Dr. Sneha Prabha Narra (CMU) for their continuing advice on this topic. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Funders	Funder number
U.S. Department of Energy	DE-FE0024064

Keywords

Additive manufacturing
Computer vision
Convolutional Neural Network (CNN)
In-situ process monitoring
Machine learning

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1016/j.addma.2018.09.034

Cite this

@article{3059bc6e394c4996a625e67a9ee346c7,

title = "A multi-scale convolutional neural network for autonomous anomaly detection and classification in a laser powder bed fusion additive manufacturing process",

abstract = "In-situ detection of processing defects is a critical challenge for Laser Powder Bed Fusion Additive Manufacturing. Many of these defects are related to interactions between the recoater blade, which spreads the powder, and the powder bed. This work leverages Deep Learning, specifically a Convolutional Neural Network (CNN), for autonomous detection and classification of many of these spreading anomalies. Importantly, the input layer of the CNN is modified to enable the algorithm to learn both the appearance of the powder bed anomalies as well as key contextual information at multiple size scales. These modifications to the CNN architecture are shown to improve the flexibility and overall classification accuracy of the algorithm while mitigating many human biases. A case study is used to demonstrate the utility of the presented methodology and the overall performance is shown to be superior to that of methodologies previously reported by the authors.",

keywords = "Additive manufacturing, Computer vision, Convolutional Neural Network (CNN), In-situ process monitoring, Machine learning",

author = "Luke Scime and Jack Beuth",

note = "Publisher Copyright: {\textcopyright} 2018 Elsevier B.V.",

year = "2018",

month = dec,

doi = "10.1016/j.addma.2018.09.034",

language = "English",

volume = "24",

pages = "273--286",

journal = "Additive Manufacturing",

issn = "2214-8604",

publisher = "Elsevier",

}

TY - JOUR

T1 - A multi-scale convolutional neural network for autonomous anomaly detection and classification in a laser powder bed fusion additive manufacturing process

AU - Scime, Luke

AU - Beuth, Jack

PY - 2018/12

Y1 - 2018/12

N2 - In-situ detection of processing defects is a critical challenge for Laser Powder Bed Fusion Additive Manufacturing. Many of these defects are related to interactions between the recoater blade, which spreads the powder, and the powder bed. This work leverages Deep Learning, specifically a Convolutional Neural Network (CNN), for autonomous detection and classification of many of these spreading anomalies. Importantly, the input layer of the CNN is modified to enable the algorithm to learn both the appearance of the powder bed anomalies as well as key contextual information at multiple size scales. These modifications to the CNN architecture are shown to improve the flexibility and overall classification accuracy of the algorithm while mitigating many human biases. A case study is used to demonstrate the utility of the presented methodology and the overall performance is shown to be superior to that of methodologies previously reported by the authors.

AB - In-situ detection of processing defects is a critical challenge for Laser Powder Bed Fusion Additive Manufacturing. Many of these defects are related to interactions between the recoater blade, which spreads the powder, and the powder bed. This work leverages Deep Learning, specifically a Convolutional Neural Network (CNN), for autonomous detection and classification of many of these spreading anomalies. Importantly, the input layer of the CNN is modified to enable the algorithm to learn both the appearance of the powder bed anomalies as well as key contextual information at multiple size scales. These modifications to the CNN architecture are shown to improve the flexibility and overall classification accuracy of the algorithm while mitigating many human biases. A case study is used to demonstrate the utility of the presented methodology and the overall performance is shown to be superior to that of methodologies previously reported by the authors.

KW - Additive manufacturing

KW - Computer vision

KW - Convolutional Neural Network (CNN)

KW - In-situ process monitoring

KW - Machine learning

UR - http://www.scopus.com/inward/record.url?scp=85054683767&partnerID=8YFLogxK

U2 - 10.1016/j.addma.2018.09.034

DO - 10.1016/j.addma.2018.09.034

M3 - Article

AN - SCOPUS:85054683767

SN - 2214-8604

VL - 24

SP - 273

EP - 286

JO - Additive Manufacturing

JF - Additive Manufacturing

ER -

A multi-scale convolutional neural network for autonomous anomaly detection and classification in a laser powder bed fusion additive manufacturing process

Abstract

Funding

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this