Skip to main content
SpringerLink
Log in

A Computer Vision Approach to Evaluate Powder Flowability for Metal Additive Manufacturing

  • Technical Article
  • Published:
Integrating Materials and Manufacturing Innovation Aims and scope Submit manuscript

Abstract

Additive manufacturing (AM) is a transformative technology to many industries that enables the fabrication of parts with complex geometries. A vast majority of powder-bed metal AM techniques use powder as feedstock. The powder packing behavior and flowability significantly influence the defect density of as-built parts and, eventually, affect their reliability and mechanical performance. The experimental characterization methods of powder flowability, for example, Hausner ratio, Carr index, and angle of repose, are rather time-consuming and cost-inefficient. Here, we show a rapid-deployed, low-cost, and reliable computer vision approach to evaluate powder flowability based on scanning electron microscopy images. We have trained seven machine learning models using 2,212 SEM images from 16 types of commonly used plasma-atomized metal powders in AM. Our results indicate that the vector of locally aggregated descriptors model with speedup robust features performs best among the models, represented by about 12 ± 7%. Mean absolute percentage error value is lower than traditional convolutional neural network model. The image analysis model can be implemented without a powerful computing system. The performance of such model is robust to the changes of image brightness. This study also demonstrates that our model can successfully predict the flowability of metal powder that does not exist in the original dataset. Such a computer vision approach provides an effective and efficient tool to evaluate and predict the powder flowability for AM.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10

Similar content being viewed by others

References

  1. Sachs E (1992) Three dimensional printing: rapid tooling and prototypes directly from a CAD model. J Eng Ind 10(1115/1):2900701

    Google Scholar 

  2. Sames WJ (2016) The metallurgy and processing science of metal additive manufacturing. Annu Rev Mater Res 61(5):315–360

    Google Scholar 

  3. Lewandowski JJ, Seifi M (2016) Metal additive manufacturing: a review of mechanical properties. Annu Rev Mater Res. https://doi.org/10.1146/annurev-matsci-070115-032024

    Article  Google Scholar 

  4. Meiners W, Wissenbach KD, Gasser AD (1998) Shaped body especially prototype or replacement part production. U.S. patent DE19649849C1

  5. Gibson I, Rosen DW, Stucker B (2014) Additive manufacturing technologies. Springer, New York

  6. Nguyen QB (2017) Characteristics of inconel powders for powder-bed additive manufacturing. Eng 3(5):695–700

    Article  CAS  Google Scholar 

  7. Sutton AT (2020) Characterization of AISI 304L stainless steel powder recycled in the laser powder-bed fusion process. Addit Manuf. https://doi.org/10.1016/j.addma.2019.100981

    Article  Google Scholar 

  8. Zhai W (2020) Characterization of nanoparticle mixed 316 L powder for additive manufacturing. J Mater Sci Technol 47:162–168

    Article  Google Scholar 

  9. Gu DD (2012) Laser additive manufacturing of metallic components: materials, processes and mechanisms. Int Mater Rev 57(3):33–164

    Article  CAS  Google Scholar 

  10. Averardi A (2020) Effect of particle size distribution on the packing of powder beds: A critical discussion relevant to additive manufacturing. Mater Today Commun. https://doi.org/10.1016/j.mtcomm.2020.100964

    Article  Google Scholar 

  11. Seyda V, Herzog D, Emmelmann (2017) Relationship between powder characteristics and part properties in laser beam melting of Ti–6Al–4V and implications on quality. J Laser Appl. https://doi.org/10.2351/1.4983240

    Article  Google Scholar 

  12. Spierings AB (2015) Powder flowability characterisation methodology for powder-bed-based metal additive manufacturing. Prog Additi Manuf 1(1–2):9–20 https://doi.org/10.1007/s40964-015-0001-4

    Article  Google Scholar 

  13. Ballard DH, Brown CM (1982) Computer vision. Englewood cliffs, Prentice Hall, NJ

  14. Sonka M, Hlavac V, Boyle R (2014) Image processing, analysis, and machine vision. Cengage Learning

  15. LeCun Y, Bengio Y, Hinton G (2015) Deep learning. Nature 521(7553):436–444. https://doi.org/10.1038/nature14539

    Article  CAS  Google Scholar 

  16. Li Q (2004) Interparticle van der Waals force in powder flowability and compactibility. Int J Pharm. https://doi.org/10.1016/j.ijpharm.2004.05.001

    Article  Google Scholar 

  17. DeCost BL, Holm EA (2017) Characterizing powder materials using keypoint-based computer vision methods. Comput Mater Sci 12:438–445. https://doi.org/10.1016/j.commatsci.2016.08.038

    Article  Google Scholar 

  18. DeCost BL (2016) Computer vision and machine learning for autonomous characterization of AM powder feedstocks. Jom 69(3):456–465. https://doi.org/10.1007/s11837-016-2226-1

    Article  Google Scholar 

  19. Scime L, Beuth J (2018) Anomaly detection and classification in a laser powder bed additive manufacturing process using a trained computer vision algorithm. Addit Manuf 19:114–126. https://doi.org/10.1016/j.addma.2017.11.009

    Article  Google Scholar 

  20. Scime L, Beuth JJAM (2018) A multi-scale convolutional neural network for autonomous anomaly detection and classification in a laser powder bed fusion additive manufacturing process. Additi Manuf 24:273–286

    Article  Google Scholar 

  21. Liang Z (2019) A particle shape extraction and evaluation method using a deep convolutional neural network and digital image processing. Powder Technol 353:156–170. https://doi.org/10.1016/j.powtec.2019.05.025

    Article  CAS  Google Scholar 

  22. Hausner HHJPT (1981) Powder characteristics and their effect on powder processing. Powder Technol 30(1):3–8

    Article  CAS  Google Scholar 

  23. Bowker MJ, Stahl PH (2008) Preparation of water-soluble compounds through salt formation. The Practice of Medicinal Chemistry, Elsevier, pp 747–766

    Chapter  Google Scholar 

  24. Mehta A, Barker GJRoPiP (1994) The dynamics of sand. 57(4):383

  25. Jillavenkatesa A, Dapkunas SJ, Lum HL-S (2001) Particle size characterization, vol 960, No. 1. National Institute of Standards and Technology

  26. Eshel G, Levy GJ, Mingelgrin U, Singer MJ (2004) Critical evaluation of the use of laser diffraction for particle-size distribution analysis. Soil Sci Soc Am J 68(3):736–743

    Article  CAS  Google Scholar 

  27. Keck CM, Müller RHJIjop (2008) Size analysis of submicron particles by laser diffractometry—90% of the published measurements are false. Int J Pharm 355(1–2):150–163

    Article  CAS  Google Scholar 

  28. International Organization for Standardization ISO 2008. Representation of results of particle size analysis—part 6: descriptive and quantitative representation of particle shape and morphology. In: Protocol ISO 9276–6. International Organization for Standardization, Geneva

  29. Wadell HJTJoG (1935) Volume, shape, and roundness of quartz particles. J Geol 43(3):250–280

    Article  Google Scholar 

  30. Kaleem MA (2020) An experimental investigation on accuracy of Hausner ratio and Carr index of powders in additive manufacturing processes. Met Powder Rep. https://doi.org/10.1016/j.mprp.2020.06.061

    Article  Google Scholar 

  31. Zegzulka J (2020) Characterization and flowability methods for metal powders. Sci Rep 10(1):21004. https://doi.org/10.1038/s41598-020-77974-3

    Article  CAS  Google Scholar 

  32. Muñiz-Lerma JA (2018) A comprehensive approach to powder feedstock characterization for powder bed fusion additive manufacturing: a case study on AlSi7Mg. Mater 11(12):2386

    Article  Google Scholar 

  33. Snow Z, Martukanitz R, Joshi S (2019) On the development of powder spreadability metrics and feedstock requirements for powder bed fusion additive manufacturing. Addit Manuf 28:78–86

    CAS  Google Scholar 

  34. ASTM B212-17 (2017) Standard test method for apparent density of free-flowing metal powders using the Hall Flowmeter funnel. ASTM International, West Conshohocken, PA

  35. ASTM B527–20 (2020) Standard test method for tap density of metal powders and compounds. ASTM International: West Conshohocken, PA

  36. ASTM B213-13 (2013) Standard test methods for flow rate of metal powders using the Hall Flowmeter funnel. ASTM International, West Conshohocken, PA

  37. Lowe David G (2004) Distinctive image features from scale-invariant keypoints. Int J Comp Vision 60(2):91–110

    Article  Google Scholar 

  38. Bay H (2008) Speeded-up robust features (SURF) Comput Vision Image Underst. 110(3):346–359

    Article  Google Scholar 

  39. Vedaldi A, Fulkerson B (2010) Vlfeat: An Open and Portable Library of Computer Vision Algorithms. In: Proceedings of the 18th ACM international conference on Multimedia https://doi.org/10.1145/1873951.1874249

  40. Corke P (2017) Robotics, vision and control: fundamental algorithms in MATLAB® second, completely revised. Springer, Cham

  41. Aloise D (2009) NP-hardness of Euclidean sum-of-squares clustering. Mach Learn 75(2):245–248

    Article  Google Scholar 

  42. Sivic J, Zisserman A (2003) Video Google: a text retrieval approach to object matching in videos in null. IEEE

  43. Perronnin F, Dance C (2007) Fisher kernels on visual vocabularies for image categorization. In 2007 IEEE conference on computer vision and pattern recognition

  44. Jégou H et al (2010) Aggregating local descriptors into a compact image representation. In 2010 IEEE computer society conference on computer vision and pattern recognition

  45. Kohonen TJN (1988) An introduction to neural computing. Neural Netw 1(1):3–16

    Article  Google Scholar 

  46. Deng L, Yu DJF (2014) Deep learning: methods and applications FNT Signal Process. 7(3–4):197–387

    Article  Google Scholar 

  47. Goodfellow I et al (2016) Deep learning. MIT press, Cambridge

  48. Aloysius N, Geetha M (2017) A review on deep convolutional neural networks. in 2017 International Conference on Communication and Signal Processing (ICCSP)

  49. LeCun Y, Bottou L, Bengio Y, Haffner P (1998) Gradient-based learning applied to document recognition. Proceedings of the IEEE 86(11):2278–2324

    Article  Google Scholar 

  50. Krizhevsky A, Sutskever A, Hinton GE (2017) Imagenet classification with deep convolutional neural networks. Commun ACM 60(6):84–90

    Article  Google Scholar 

  51. Simonyan K, Zisserman AJ (2014) Very deep convolutional networks for large-scale image recognition. arXiv preprint

  52. Szegedy C et al (2015) Going deeper with convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition

  53. He K et al (2016) Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition

  54. Stavrou SG (2020) Investigation of powder flowability at low stresses: influence of particle size and size distribution. Powder Technol 364:98–114. https://doi.org/10.1016/j.powtec.2020.01.068

    Article  CAS  Google Scholar 

  55. Brika SE (2020) Influence of particle morphology and size distribution on the powder flowability and laser powder bed fusion manufacturability of Ti-6Al-4V alloy. Addit Manufact. https://doi.org/10.1016/j.addma.2019.100929

    Article  Google Scholar 

  56. McGlinchey D (2009) Bulk solids handling: equipment selection and operation. John Wiley & Sons

  57. Bay H, Tuytelaars T, Van Gool L (2006) Surf: speeded up robust features. in European conference on computer vision, Springer

  58. Kodinariya TM, Makwana PRJIJ (2013) Review on determining number of Cluster in K-Means Clustering. Computer Vision – ECCV 2006: 9th European Conference on Computer Vision, Graz, Austria, Proceedings, Part I. Berlin, Heidelberg

Download references

Acknowledgements

We greatly acknowledge the financial support from Natural Sciences and Engineering Research Council of Canada (NSERC Discovery Grant # RGPIN-2018-05731), Connaught New Researcher Award, and Dean’s Spark Assistant Professorship in the Faculty of Applied Science & Engineering at the University of Toronto. J.Z., Z.L., and Y.Z. acknowledge the access to the SEM system in Prof. Yu Sun’s lab at University of Toronto. Y.Z. and Q.S. acknowledge the support from New Frontiers in Research Fund—Exploration (NFRFE-2019-00603).

Author information

Authors and Affiliations

  1. Department of Materials Science and Engineering, University of Toronto, Toronto, ON, M5S 3E4, Canada

    Jiahui Zhang, Zhiying Liu, Tianyi Lyu & Yu Zou

  2. AP&C Advanced Powder and Coating, a GE Additive Company, General Electric, Montreal, QC, Canada

    Mahdi Habibnejad-korayem

  3. Department of Statistical Sciences, University of Toronto, Toronto, ON, M5S 3G3, Canada

    Qiang Sun

Authors
  1. Jiahui Zhang
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Mahdi Habibnejad-korayem
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Zhiying Liu
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Tianyi Lyu
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Qiang Sun
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Yu Zou
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Yu Zou.

Ethics declarations

Conflict of interest

The authors declare that they have no conflict of interest. All co-authors have seen and agree with the contents of the manuscript and there is no financial interest to report. We certify that the submission is original work and is not under review at any other publication.

Appendices

Appendix 1

See Table 4.

Table 4 Quantity of SEM images for each type of metal powders
Full size table

Appendix 2

The Detailed Procedures to Generate Compact Code for Each Image

We first extracted predictive features, also referred to key features, from the images. The scale-invariant feature transform (SIFT) [37] and speeded up robust feature (SURF) [57] algorithms are two powerful approaches to extract such features because they are invariant to scale, rotate, translation, illumination, and blur. To calculate SIFT and SURF features, there were four major steps: scale-space extrema detection, key-point localization, orientation assignment, and key-point descriptor. For each image, we applied various detectors to detect distinctive key points and computed the corresponding SIFT and SURF features by VLFeat library [39] and Machine Vision Toolbox [40], respectively. For SIFT features, we applied Harris–Laplace (HL) and difference of Gaussian (DoG) operators to detect distinctive blob-like and corner-like image features. For SURF features, the algorithm used an integer approximation of the determinant of the Hessian blob detector. After extracting SIFT and SURF features from each image, k-means clustering was used to cluster all the features into k clusters, in which each feature is clustered to the cluster whose centroid is closest to the feature [41]. Those distinct clusters could be regarded as distinct visual words of the dictionary formed by all key-point features. The number of k is determined by the elbow algorithm [58], as shown in Fig. 11. Finally, all features extracted from the image were aggregated to generate compact code (descriptor) for each image. The bag of visual words (BOVW) [42], Fisher kernel (FV) [43], and the vector of locally aggregated descriptors (VLAD) [44] representation methods were applied for achieving this purpose. The BOVW representation was calculated by assigning each key-point feature in the image to the cluster whose centroid is closest to the feature and counting the occurrence frequency histogram. In addition to the 0-order statistics of the distribution of descriptors as collected by BOVW, the FV representation also collected some high-order statistics. The VLAD representation was a simplified non-probabilistic version of FV, calculated by accumulating the residual of each descriptor to its assigned cluster. One example of constructing a feature-based representation is illustrated in Fig. 12.

Fig. 11
figure 11

The selection of the optimized number of clusters through the “elbow” method. a k = 80 for SURF features (indicated by the dotted line); b k = 80 for SIFT features (indicated by the dotted line)

Full size image
Fig. 12
figure 12

Illustration of an example of the generation of feature-based representations. a Key-point detection, b SIFT descriptor calculation, c k-means clustering, d BOVW representation, e FV representation, and f VLAD representation

Full size image

Appendix 3

Figures for ANN and CNN Training Process

See Figs. 13 and 14

Fig. 13
figure 13

ANN training curves (image representations generated by model ii as the input and the median size as the output): a regression plot for training dataset; b regression plot for validation dataset; c regression plot for test dataset; and d regression plot for total dataset

Full size image
Fig. 14
figure 14

Performance vs. training epoch for the VGG16 CNN models: a1 root-mean-square error (RMSE) on training the median size, a2 root-mean-square error (RMSE) on training the HR, a3 root-mean-square error (RMSE) on training the CI, a4 root-mean-square error (RMSE) on training the AOR, b1 loss on training the median size, b2 loss on training the HR, b3 loss on training the CI, b4 loss on training the AOR

Full size image

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Zhang, J., Habibnejad-korayem, M., Liu, Z. et al. A Computer Vision Approach to Evaluate Powder Flowability for Metal Additive Manufacturing. Integr Mater Manuf Innov 10, 429–443 (2021). https://doi.org/10.1007/s40192-021-00226-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s40192-021-00226-3

Keywords

Search

Navigation