Detection of anomalous meltpool conditions using deep learning in laser powder bed fusion

Abstract

Laser powder bed fusion for metals (PBF-LB/M) is an advanced additive manufacturing process in which metal powder is selectively melted and fused layer by layer. Maintaining consistent meltpool conditions is critical to ensure the production of fully dense, defect-free parts. While process parameters, such as laser power or scan speed, are usually held constant during a build, effective variations in meltpool conditions arising from geometry, thermal history, and energy redistribution during the build can lead to porosity formation defects such as lack of fusion or keyholing porosity, which are often detected only after fabrication. To address this challenge, in-situ monitoring techniques have been implemented to observe meltpool dynamics in real time. During fabrication process for nickel alloy 625 samples, high-speed infrared and off-axis imaging systems capture grayscale images of the process, which are converted into contour maps highlighting regions of interest, including the meltpool, meltpool vicinity or plume, and spatter. These images are analyzed using deep learning (DL) methods to detect potential anomalies. In particular, convolutional neural networks (CNNs) are employed for feature extraction and classification of meltpool conditions. The CNN model is trained on a labeled dataset of meltpool images and validated on a separate test set. It achieves a classification accuracy of 97.87% across three condition categories: acceptable normal meltpool, overheated meltpool and its vicinity, and underheated meltpool and its vicinity. This trained DL model is ready to be implemented in a real-time monitoring framework that enables effective detection of conditions resulting in defects during the build process and supports dynamic adjustment of process parameters, such as energy input, to reduce defect formation. By enabling real-time feedback and control, the system would enhance the meltpool consistency during the PBF-LB/M process.

Keywords

Additive manufacturing / Laser powder bed fusion / Meltpool condition / Process monitoring / Deep learning (DL) / Image processing

References

[1] Kladovasilakis N, Charalampous P, Kostavelis I, et al.. Impact of metal additive manufacturing parameters on the powder bed fusion and direct energy deposition processes: a comprehensive review. Prog Addit Manuf, 2021, 6(3): 349-365

[2] Zhang Z, Sahu CK, Singh SK, et al.. Machine learning based prediction of melt pool morphology in a laser-based powder bed fusion additive manufacturing process. Int J Prod Res, 2024, 62(5): 1803-1817

[3] Gaikwad A, Williams RJ, de Winton H, et al.. Multi phenomena melt pool sensor data fusion for enhanced process monitoring of laser powder bed fusion additive manufacturing. Mater Des, 2022, 221 ArticleID: 110919

[4] Zhang B, Jaiswal P, Rai R, et al.. Convolutional neural network-based inspection of metal additive manufacturing parts. Rapid Prototyp J, 2019, 25(3): 530-540

[5] Qin J, Hu F, Liu Y, et al.. Research and application of machine learning for additive manufacturing. Addit Manuf, 2022, 52 ArticleID: 102691

[6] Gaikwad A, Yavari R, Montazeri M, et al.. Toward the digital twin of additive manufacturing: integrating thermal simulations, sensing, and analytics to detect process faults. IISE Trans, 2020, 52(11): 1204-1217

[7] Mukherjee T, DebRoy T. A digital twin for rapid qualification of 3D printed metallic components. Appl Mater Today, 2019, 14: 59-65

[8] Wang H, Li B, Zhang S, et al.. Traditional machine learning and deep learning for predicting melt-pool cross-sectional morphology of laser powder bed fusion additive manufacturing with thermographic monitoring. J Intell Manuf, 2025, 36: 2079-2104

[9] Grasso M, Colosimo BM. Process defects and in situ monitoring methods in metal powder bed fusion: a review. Meas Sci Technol, 2017, 28(4 ArticleID: 044005

[10] Özel T. A review on in-situ process sensing and monitoring systems for fusion-based additive manufacturing. Int J Mechatron Manuf Syst, 2023, 16(2/3): 115-154

[11] Gu Z, Mani Krishna KV, Parsazadeh M, et al.. Deep learning-based melt pool and porosity detection in components fabricated by laser powder bed fusion. Prog Addit Manuf, 2025, 10: 53-70

[12] Fu Y, Downey AR, Yuan L, et al.. Machine learning algorithms for defect detection in metal laser-based additive manufacturing: a review. J Manuf Process, 2022, 75: 693-710

[13] Smoqi Z, Gaikwad A, Bevans B, et al.. Monitoring and prediction of porosity in laser powder bed fusion using physics-informed meltpool signatures and machine learning. J Mater Process Technol, 2022, 304 ArticleID: 117550

[14] Guo S, Agarwal M, Cooper C, et al.. Machine learning for metal additive manufacturing: towards a physics-informed data-driven paradigm. J Manuf Syst, 2022, 62: 145-163

[15] Özel T. Deep learning-based applications in metal additive manufacturing processes: challenges and opportunities-a review. Int J Lightweight Mater Manuf, 2025, 8(4): 453-468

[16] Caggiano A, Zhang J, Alfieri V, et al.. Machine learning-based image processing for on-line defect recognition in additive manufacturing. CIRP Ann, 2019, 68: 451-454

[17] Liu J, Ye J, Silva Izquierdo D, et al.. A review of machine learning techniques for process and performance optimization in laser beam powder bed fusion additive manufacturing. J Intell Manuf, 2023, 34(8): 3249-3275

[18] Williams RJ, Sing SL. Spatiotemporal analysis of powder bed fusion melt pool monitoring videos using deep learning. J Intell Manuf, 2025, 36: 2409-2422

[19] Yuan B, Giera B, Guss G et al (2019) Semi-supervised convolutional neural networks for in-situ video monitoring of selective laser melting. In: 2019 IEEE winter conference on applications of computer vision (WACV), Waikoloa, HI, USA, pp 744–753. https://doi.org/10.1109/WACV.2019.00084

[20] Li Z, Liu F, Yang W, et al.. A survey of convolutional neural networks: analysis, applications, and prospects. IEEE Trans Neural Netw Learn Syst, 2022, 33(12): 6999-7019

[21] Jamnikar ND, Liu S, Brice C, et al.. In-process comprehensive prediction of bead geometry for laser wire-feed DED system using molten pool sensing data and multi-modality CNN. Int J Adv Manuf Technol, 2022, 121(1/2): 903-917

[22] Parsazadeh M, Sharma S, Dahotre N. Towards the next generation of machine learning models in additive manufacturing: a review of process dependent material evolution. Prog Mater Sci, 2023, 135 ArticleID: 101102

[23] Scime L, Beuth J. Anomaly detection and classification in a laser powder bed additive manufacturing process using a trained computer vision algorithm. Addit Manuf, 2018, 19: 114-126

[24] Williams G, Meisel NA, Simpson TW, et al.. Design repository effectiveness for 3D convolutional neural networks: application to additive manufacturing. J Mech Des, 2019

[25] Mao Z, Feng W, Ma H, et al.. Continuous online flaws detection with photodiode signal and melt pool temperature based on deep learning in laser powder bed fusion. Opt Laser Technol, 2023, 158 ArticleID: 108877

[26] Balu A, Lore KG, Young G et al (2016) A deep 3D convolutional neural network based design for manufacturability framework. arXiv:1612.02141. https://doi.org/10.48550/arXiv.1612.02141

[27] Shen Z, Shang X, Zhao M, et al.. A learning-based framework for error compensation in 3D printing. IEEE Trans Cybern, 2019, 49(11): 4042-4050

[28] Lee KH, Lee HW, Yun GJ. A defect detection framework using three-dimensional convolutional neural network (3D-CNN) with in-situ monitoring data in laser powder bed fusion process. Opt Laser Technol, 2023, 165 ArticleID: 109571

[29] Kim J, Yang Z, Ko H, et al.. Deep learning-based data registration of melt-pool-monitoring images for laser powder bed fusion additive manufacturing. J Manuf Syst, 2023, 68: 117-129

[30] Qi X, Chen G, Li Y, et al.. Applying neural-network-based machine learning to additive manufacturing: current applications, challenges, and future perspectives. Engineering, 2019, 5(4): 721-729

[31] Sun C, Shi VG. PhysiNet: a combination of physics-based model and neural network model for digital twins. Int J Intell Syst, 2022, 37(8): 5443-5456

[32] Eshkevari SS, Takáč M, Pakzad SN, et al.. DynNet: physics-based neural architecture design for linear and nonlinear structural response modeling and prediction. Eng Struct, 2021, 229 ArticleID: 111582

[33] Karniadakis GE, Kevrekidis IG, Lu L, et al.. Physics-informed machine learning. Nat Rev Phys, 2021, 3(6): 422-440

[34] Criales LE, Arisoy YM, Lane B, et al.. Laser powder bed fusion of nickel alloy 625: experimental investigations of effects of process parameters on melt pool size and shape with spatter analysis. Int J Mach Tools Manuf, 2017, 121: 22-36

[35] Yang L, Lo L, Ding S, et al.. Monitoring and detection of meltpool and spatter regions in laser powder bed fusion of super alloy Inconel 625. Prog Addit Manuf, 2020, 5: 367-378

[36] Otsu N. A threshold selection method from gray-level histograms. IEEE Trans Syst Man Cybern, 1979, 9(1): 62-66

[37] Thanki A, Goossens L, Ompusunggu AP, et al.. Melt pool feature analysis using a high-speed coaxial monitoring system for laser powder bed fusion of Ti-6Al-4V grade 23. Int J Adv Manuf Technol, 2022, 120(9): 6497-6514

[38] Wang C, Tan XP, Tor SB, et al.. Machine learning in additive manufacturing: state-of-the-art and perspectives. Addit Manuf, 2020, 36 ArticleID: 101538

[39] Valizadeh M, Wolff SJ. Convolutional neural network applications in additive manufacturing: a review. Adv Ind Manuf Eng, 2022, 4 ArticleID: 100072

[40] Kumar S, Gopi T, Harikeerthana N, et al.. Machine learning techniques in additive manufacturing: a state of the art review on design, processes and production control. J Intell Manuf, 2023, 34(1): 21-55

[41] Glorot X, Bengio Y (2010) Understanding the difficulty of training deep feedforward neural networks. In: Proceedings of the thirteenth international conference on artificial intelligence and statistics, pp 249–256

[42] Orr GB, Müller KR. Neural networks: tricks of the trade. Lect Notes Comput Sci, 1998, 1524: 6

[43] Nair V, Hinton GE (2010) Rectified linear units improve restricted boltzmann machines. In: International conference on machine learning. pp 807–814

[44] Bridle JSSoulié FF, Hérault J. Probabilistic interpretation of feedforward classification network outputs, with relationships to statistical pattern recognition. Neurocomputing. NATO ASI series, 1990, Berlin, Springer

[45] Gu J, Wang Z, Kuen J, et al.. Recent advances in convolutional neural networks. Pattern Recognit, 2018, 77: 354-377

[46] Gorach T. Deep convolutional neural networks-a review. Int Res J Eng Technol, 2018, 5(7): 439-452

[47] Srivastava N, Hinton G, Krizhevsky A, et al.. Dropout: a simple way to prevent neural networks from overfitting. J Mach Learn Res, 2014, 15(1): 1929-1958

[48] Laurent C, Pereyra G, Brakel P et al (2016) Batch normalized recurrent neural networks. 2016 IEEE international conference on acoustics, speech and signal processing (ICASSP), Shanghai, China, pp 2657–2661. https://doi.org/10.1109/ICASSP.2016.7472159

[49] Bjorck N, Gomes CP, Selman B et al (2018) Understanding batch normalization. arXiv:1806.02375.https://doi.org/10.48550/arXiv.1806.02375

[50] Awad M, Khanna R (2015) Support vector machines for classification. In: Efficient learning machines: theories, concepts, and applications for engineers and system designers. Apress, Berkeley, CA, pp 39–66. https://doi.org/10.1007/978-1-4302-5990-9

[51] He K, Zhang X, Ren S et al (2016) Deep residual learning for image recognition. In: IEEE conference on computer vision and pattern recognition (CVPR), Las Vegas, NV, USA, pp 770–778. https://doi.org/10.1109/CVPR.2016.90

RIGHTS & PERMISSIONS

The Author(s)

Just Accepted

This article has successfully passed peer review and final editorial review, and will soon enter typesetting, proofreading and other publishing processes. The currently displayed version is the accepted final manuscript. The officially published version will be updated with format, DOI and citation information upon launch. We recommend that you pay attention to subsequent journal notifications and preferentially cite the officially published version. Thank you for your support and cooperation.