Machine learning guided automatic recognition of crystal boundaries in bainitic/martensitic alloy and relationship between boundary types and ductile-to-brittle transition behavior

doi:10.1016/j.jmst.2020.12.024

Abstract

Abstract:

Gradient boosting decision tree (GBDT) machine learning (ML) method was adopted for the first time to automatically recognize and conduct quantitative statistical analysis of boundaries in bainitic microstructure using electron back-scatter diffraction (EBSD) data. In spite of lack of large sets of EBSD data, we were successful in achieving the desired accuracy and accomplishing the objective of recognizing the boundaries. Compared with a low model accuracy of <50 % as using Euler angles or axis-angle pair as characteristic features, the accuracy of the model was significantly enhanced to about 88 % when the Euler angle was converted to overall misorientation angle (OMA) and specific misorientation angle (SMA) and considered as important features. In this model, the recall score of prior austenite grain (PAG) boundary was ~93 %, high angle packet boundary (OMA>40°) was ~97 %, and block boundary was ~96 %. The derived outcomes of ML were used to obtain insights into the ductile-to-brittle transition (DBTT) behavior. Interestingly, ML modeling approach suggested that DBTT was not determined by the density of high angle grain boundaries, but significantly influenced by the density of PAG and packet boundaries. The study underscores that ML has a great potential in detailed recognition of complex multi-hierarchical microstructure such as bainite and martensite and relates to material performance.

Key words: Machine learning, Feature engineering, Automatic recognition, Lath structure, Crystallography

X.C. Li, J.X. Zhao, J.H. Cong, R.D.K. Misra, X.M. Wang, X.L Wang, C.J. Shang. Machine learning guided automatic recognition of crystal boundaries in bainitic/martensitic alloy and relationship between boundary types and ductile-to-brittle transition behavior[J]. J. Mater. Sci. Technol., 2021, 84: 49-58.

Figures/Tables 13

Fig. 1. Microstructural hierarchy of lath martensite structure (a), six crystallographic variants (V1-V6) for the K-S orientation relationship that evolves on a (111) austenite plane (b), and standard pole figure showing orientations of 24 variants transformed from one austenite grain with K-S orientation relationship (c) [17].

Fig. 2. (a) Inverse pole figure color map of bainitic microstructure in experimental steel, (b) {001} pole figure, (c) variants in the selected PAG and (d) different types of boundaries: PAG boundaries (black lines), packet boundaries (blue lines), block boundaries (red lines) and sub-block boundaries (green lines) within the selected PAG.

Table 1 Parameters for different variant pairs with K-S orientation relationship.

Variant Pair	OMA (^o)	Axis	SMA (^o)					Boundary Type
Variant Pair	OMA (^o)	Axis	{100}	{110}	{111}	{112}	{123}	Boundary Type
V1-V1	-	-						-
V1-V2	60.00	[0.577 0.577 -0.577]	48.19	0.00	0.00	0.00	0.00	Block boundary
V1-V3	60.00	[0.000 0.707 0.707]	41.41	0.00	10.48	5.28	3.46	Block boundary
V1-V4	10.52	[0.000 -0.707 -0.707]	7.44	0.00	6.08	5.25	3.44	Sub-block boundary
V1-V5	60.00	[0.000 -0.707 -0.707]	41.41	0.00	10.58	5.18	3.37	Block boundary
V1-V6	49.48	[0.000 0.707 0.707]	34.42	0.00	21.07	0.00	6.84	Block boundary
V1-V7	49.47	[-0.577 -0.577 0.577]	39.95	10.52	0.00	9.81	2.29	Packet boundary
V1-V8	10.53	[0.577 0.577 -0.577]	8.59	6.07	0.00	3.50	3.95
V1-V9	50.51	[-0.615 0.186 -0.767]	31.80	10.53	20.91	5.84	3.12
V1-V10	50.51	[-0.739 -0.462 0.490]	33.43	10.53	10.58	6.03	5.89
V1-V11	14.88	[0.933 0.354 0.065]	5.35	6.14	9.30	5.33	4.43
V1-V12	57.22	[-0.357 0.603 0.714]	39.20	6.15	14.18	5.33	4.43
V1-V13	14.88	[0.354 -0.933 -0.065]	5.35	6.15	9.30	5.38	4.41
V1-V14	50.51	[-0.490 0.462 -0.739]	33.43	10.53	10.48	6.20	5.54
V1-V15	57.21	[-0.738 -0.246 0.628]	37.67	7.44	20.12	6.05	4.30
V1-V16	20.60	[0.659 -0.659 -0.363]	15.45	7.43	4.97	6.05	1.75
V1-V17	51.73	[-0.659 0.363 -0.659]	38.31	9.56	12.10	4.95	0.38
V1-V18	47.12	[-0.719 -0.302 0.626]	32.24	14.21	14.21	3.07	4.76
V1-V19	50.51	[-0.186 0.767 0.615]	31.80	10.53	20.94	5.88	2.67
V1-V20	57.21	[0.357 0.714 -0.603]	39.19	6.15	14.21	5.36	4.41
V1-V21	20.60	[0.955 0.000 -0.296]	6.06	9.57	14.21	8.92	4.48
V1-V22	47.12	[-0.302 0.626 0.719]	32.25	14.21	14.18	3.11	4.48
V1-V23	57.21	[-0.246 -0.628 -0.738]	37.68	7.44	20.12	6.08	4.35
V1-V24	21.05	[0.912 -0.410 0.000]	8.59	7.44	13.55	8.57	6.23

Fig. 3. The flowchart of ML model for boundary recognition.

Fig. 4. The accuracy of GBDT model with different feature combinations.

Fig. 5. Accuracy of model predictions using different feature combinations compared with manual recognition.

Fig. 6. The importance of each feature analyzed with GBDT model.

Fig. 7. Evaluation of recognition accuracy of different types of boundaries.

Fig. 8. Boundary recognition in experiment steel by ML model. (a) Inverse pole figure color map of the bainite structure in experimental steel, (b) boundaries (black lines: PAG boundaries, blue lines: packet boundaries, red lines: block boundaries, green lines: sub-block boundaries), and (c) PAG boundaries skeleton map.

Fig. 9. Quantitative statistical results of interface obtained by ML model.

Fig. 10. Frequency distribution of different types of boundaries with different specific misorientation angles: (a) OMA, (b) {100}-SMA, (c) {110}-SMA, (d) {111}-SMA, (e) {112}-SMA, (f) {123}-SMA.

Fig. 11. Using Kernel density estimation to analyze the difference between different types of boundaries in different features. (a) {100}-SMA and {110}-SMA, (b) OMA and {111}-SMA, (c) {112}-SMA and {123}-SMA.

Fig. 12. The change of ductile-to-brittle transition temperature (DBTT) with density of different types of boundaries.

References 52

[1]	C.J. Shang, X.M. Wang, S.W. Yang, X.L. He, H.B. Wu, Acta Metall. Sin. 39 (2003) 1019-1024 (in Chinese).
[2]	J.R. Yang, C.Y. Huang, S.C. Wang, Mater. Des. 13 (1992) 335-338.
[3]	C. Wang, M. Wang, J. Shi, W. Hui, H. Dong, Scr. Mater. 58 (2008) 492-495. DOI URL
[4]	J.W. Morris, Z. Guo, ISIJ Int. 43 (2003) 410-419. DOI URL
[5]	J.W. Morris, Science 320 (2008) 1022-1023. DOI PMID
[6]	Z. Guo, C.S. Lee, J.W. Morris, Acta Mater. 52 (2004) 5511-5518. DOI URL
[7]	X.J. Sun, C.J. Shang, P.F. Miao, L.L. Dong, Trans. Mater. Heat Treat. 41 (2020) 50-56.
[8]	X. Li, X. Ma, S.V. Subramanian, C. Shang, R.D.K. Misra, Mater. Sci. Eng. A 616 (2014) 141-147. DOI URL
[9]	E. Bouyne, H.M. Flower, T.C. Lindley, A. Pineau, Scr. Mater. 39 (1998) 295-300. DOI URL
[10]	C.F. Wang, M.Q. Wang, J. Shi, W.J. Hui, H. Dong, Mater. Sci. Technol. 23 (2007) 659-664.
[11]	C.F. Wang, M.Q. Wang, J. Shi, W.J. Hui, H. Dong, Iron Steel 11 (2007) 60-63.
[12]	D.S. Liu, M. Luo, B.G. Cheng, R. Cao, J.H. Chen, Metall. Mater. Trans. A 49 (2018) 4918-4936. DOI URL
[13]	S. Morito, H. Yoshida, T. Maki, X. Huang, Mater. Sci. Eng. A 438 (2006) 237-240.
[14]	Z. Luo, J. Shen, H. Su, Y. Ding, C. Yang, X. Zhu, Iron Steel Res. Int. 17 (2010) 40-48.
[15]	H.W. Luo, X.H. Wang, Z.B. Liu, Z.Y. Yang, J. Mater. Sci. Technol. 51 (2020) 130-136. DOI URL
[16]	S.L. Long, Y.L. Liang, Y. Jiang, Y. Liang, M. Yang, Y.L. Yi, Mater. Sci. Eng. A 676 (2016) 38-47. DOI URL
[17]	H. Kitahara, R. Ueji, N. Tsuji, Y. Minamino, Acta Mater. 54 (2006) 1279-1288. DOI URL
[18]	T. Furuhara, H. Kawata, S. Morito, G. Miyamoto, T. Maki, Metall. Mater. Trans. A 39 (2008) 1003-1013. DOI URL
[19]	X.L. Wang, Z.Q. Wang, L.L. Dong, C.J. Shang, X.P. Ma, S.V. Subramanian, Mater. Sci. Eng. A 704 (2017) 448-458. DOI URL
[20]	X.L. Wang, Z.Q. Wang, X.P. Ma, S.V. Subramanian, Z.J. Xie, C.J. Shang, X.C. Li, Mater. Charact. 140 (2018) 312-319. DOI URL
[21]	B.B. Wu, X.L. Wang, Z.Q. Wang, J.X. Zhao, Y.H. Jin, C.S. Wang, C.J. Shang, R.D.K. Misra, Mater. Sci. Eng. A 745 (2019) 126-136. DOI URL
[22]	B.B. Wu, S. Huang, Z.Q. Wang, J.X. Zhao, C.S. Wang, C.J. Shang, R.D.K. Misra, Mater. Lett. 250 (2019) 55-59. DOI
[23]	L. Germain, N. Gey, R. Mercier, P. Blaineau, M. Humbert, Acta Mater. 60 (2012) 4551-4562. DOI URL
[24]	Z.Q. Wang, T.T. Song, Y. You, X.L. Wang, C.J. Shang, K. Liu, B. Chen, Chin. J. Eng. 40 (2018) 945-953.
[25]	M. Abbasi, T.W. Nelson, C.D. Sorensen, L.Y. Wei, Mater. Charact. 66 (2012) 1-8. DOI URL
[26]	M. Abbasi, D. Kim, T.W. Nelson, M. Abbasi, Mater. Charact. 95 (2014) 219-231. DOI URL
[27]	G. Miyamoto, N. Iwata, N. Takayama, T. Furuhara, ISIJ Int. 51 (2011) 1174-1178. DOI URL
[28]	C.T. Wu, H.T. Chang, C.Y. Wu, S.W. Chen, S.Y. Huang, M.X. Huang, Y.T. Pan, P. Bradbury, J. Chou, H.W. Yen, Mater. Today 34 (2020) 41-45. DOI URL
[29]	C. Wen, Y. Zhang, C. Wang, D. Xue, Y. Bai, S. Antonov, L. Dai, T. Lookman, Y. Su, Acta Mater. 170 (2019) 109-117. DOI URL
[30]	C. Shen, C. Wan, X. Wei, Y. Li, S. van der Zwaag, W. Xu, Acta Mater. 179 (2019) 201-214. DOI URL
[31]	W.Y. Wang, S.L. Shang, Y. Wang, F. Han, K.A. Darling, Y. Wu, X. Xie, O.N. Senkov, J. Li, X.D. Hui, K.A. Dahmen, P.K. Liaw, L.J. Kecskes, Z.K. Liu, Nature Comput. Mater. 3 (2017) 1-10.
[32]	A. Chowdhury, E. Kautz, B. Yener, D. Lewis, Comput. Mater. Sci. 123 (2016) 176-187.
[33]	B.L. DeCost, E.A. Holm, Comput. Mater. Sci. 110 (2015) 126-133. DOI URL
[34]	S.M. Azimi, D. Britz, M. Engstler, M. Fritz, F. Mücklich, Sci. Rep. 8 (2018) 1-14.
[35]	S. Gupta, J. Sarkar, M. Kundu, N.R. Bandyopadhyay, S. Ganguly, Measurement 151 (2020), 107224. DOI URL
[36]	J.H. Friedman, Ann. Math. Stat. 29 (2001) 1189-1232.
[37]	A. Jalal, Int. J. Intell. Syst. 22 (2018) 177-193.
[38]	P. Domingos, Commun. ACM 55 (2012) 78-87.
[39]	H. Beladi, G.S. Rohrer, A.D. Rollett, V. Tari, P.D. Hodgson, Acta Mater. 63 (2014) 86-98. DOI URL
[40]	S. Morito, H. Tanaka, R. Konishi, T. Furuhara, T. Maki, Acta Mater. 51 (2003) 1789-1799. DOI URL
[41]	S. Morito, H. Saito, T. Ogawa, T. Furuhara, T. Maki, ISIJ Int. 45 (2005) 91-94. DOI URL
[42]	T. Furuhara, H. Kawata, S. Morito, T. Maki, Mater. Sci. Eng. A 431 (2006) 228-236. DOI URL
[43]	X.C. Li, J.X. Zhao, J.L. Wang, X.L. Wang, S.L. Liu, C.J. Shang, Mater. Lett. 259 (2020), 126841. DOI URL
[44]	L. Breiman, Mach. Learn. 45 (2001) 5-32. DOI URL
[45]	C.L. Miao, C.J. Shang, H.S. Zurob, G.D. Zhang, S.V. Subramanian, Metall. Mater. Trans. A 43 (2012) 665-676. DOI URL
[46]	C. Shorten, T.M. Khoshgoftaar, J. Big Data 6 (2019) 1-48. DOI URL
[47]	P. Corcoran, C. Costacke, V. Varkarakis, J. Lemley, IEEE Consum. Electron. Mag. 9 (2020) 48-54.
[48]	M. Rosenblatt, J. Ann. Math. Stat. 27 (1956) 832-837.
[49]	E. Parzen, Ann. Math. Stat. 33 (1962) 1065-1076. DOI URL
[50]	A. Lambert, X. Garat, T. Sturel, Scr. Mater. 43 (2000) 161-166. DOI URL
[51]	A.F. Gourgues, H.M. Flower, T.C. Lindley, Mater. Sci. Tech. 16 (2000) 26-40. DOI URL
[52]	Y. You, C. Shang, W. Nie, S. Subramanian, Mater. Sci. Eng. A 558 (2012) 692-701. DOI URL