Predicting the glass transition temperature of polymer based on generative adversarial networks and automated machine learning

Abstract

Solution styrene-butadiene rubber (SSBR) finds wide applications in high performance tire design and various other fields. This study aims to create a quantitative structure–property relationship (QSPR) model linking SSBR’s glass transition temperature (T_g) to its structural properties. A dataset of 68 sets of data from published literature was compiled to develop a predictive machine learning model for SSBR’s structural design and synthesis using small sample sizes. To tackle small sample sizes, a framework combining generative adversarial networks (GAN) and the Tree-based Pipeline Optimization Tool (TPOT) is proposed. GAN is first used to generate additional samples that mirror the original dataset’s distribution, expanding the dataset. The TPOT is then applied to automatically find the best model and parameter combinations, creating an optimal predictive model for the mixed dataset. Experimental results show that using GAN to enlarge the dataset and TPOT regression models significantly enhances model performance, increasing the R² value from 0.745 to 0.985 and decreasing the RMSE from 7.676 to 1.569. The proposed GAN–TPOT framework demonstrates the potential of combining generative models with automated machine learning to improve materials science research. This combination accelerates research and development processes, enhances prediction and design accuracy, and introduces new perspectives and possibilities for the field.

Keywords

generative adversarial networks / glass transition temperature / solution styrene-butadiene rubber / treebased pipeline optimization tool / virtual sample generation

References

1	Sun C, Wen S, Ma H, et al. Improvement of silica dispersion in solution polymerized styrene–butadiene rubber via introducing amino functional groups. Ind Eng Chem Res. 2019;58(3):1454-1461.
2	Sun J, Song Y, Zheng Q, Tan H, Yu J, Li H. Nonlinear rheological behavior of silica filled solution-polymerized styrene butadiene rubber. J Polym Sci B Polym Phys. 2007;45(18):2594-2602.
3	Liu X, Zhao S, Zhang X, Li X, Bai Y. Preparation, structure, and properties of solution-polymerized styrene-butadiene rubber with functionalized end-groups and its silica-filled composites. Polymer. 2014;55(8):1964-1976.
4	Hou G, Tao W, Liu J, Zhang X, Dong M, Zhang L. Effect of the structural characteristics of solution styrene–butadiene rubber on the properties of rubber composites. J Appl Polym Sci. 2018;135(24):45749.
5	Hua J, Liu K, Wang Z, Geng J, Wang X. Effect of vinyl and phenyl group content on the physical and dynamic mechanical properties of HVBR and SSBR. J Appl Polym Sci. 2018;135(12):45975.
6	Wang B, Ma J-H. Wu Y-P. Application of multiple linear regression in predicting abrasion resistance of solution polymerized styrene-butadiene rubber based composites. China Synth Rubber Ind. 2013;36(2):123-126.
7	Li H, Yang D, Chen F, Zhou Y, Xiu Z. Application of artificial neural Networks in predicting abrasion resistance of solution polymerized styrene-butadiene rubber based composites. In: 2014 IEEE Workshop on Electronics, Computer and Applications; 2014:581-584.
8	Karuth A, Alesadi A, Xia W, Rasulev B. Predicting glass transition of amorphous polymers by application of cheminformatics and molecular dynamics simulations. Polymer. 2021;218:123495.
9	Liu Y, Wang J, Xiao B, Shu J. Accelerated development of hard highentropy alloys with data-driven high-throughput experiments. Journal of Materials Informatics. 2022;2(1):3.
10	Drucker, H;Burges, C;Kaufman, L;Smola, A;Vapnik, V Support vector regression machines;1996.
11	Shen L, Qian Q. A virtual sample generation algorithm supporting machine learning with a small-sample dataset: a case study for rubber materials. Comput Mater Sci. 2022;211:111475.
12	Yang J, Yu X, Xie Z-Q. Zhang J-P. A novel virtual sample generation method based on Gaussian distribution. Knowledge-Based Syst. 2011;24(6):740-748.
13	Li D, Liu J, Liu J. NNI-SMOTE-XGBoost: a novel small sample analysis method for properties prediction of polymer materials. Macromol Theory Simulations. 2021;30(5):2100010.
14	Chen W, Chen K. Nonlinear probabilistic virtual sample generation using Gaussian process latent variable model and fitting for rubber material. Comput Mater Sci. 2023;230:112477.
15	Liu Y, Yang Z, Yu Z, et al. Generative artificial intelligence and its applications in materials science: current situation and future perspectives. Journal of Materiomics. 2023;9(4):798-816.
16	Labaca-Castro R. Generative adversarial nets. In: Labaca-Castro R, ed. Machine Learning under Malware Attack. Springer Fachmedien;2023:73-76.
17	Marani A, Jamali A, Nehdi ML. Predicting ultra-high-performance concrete compressive strength using tabular generative adversarial Networks. Materials. 2020;13(21):4757.
18	Yang Z, Li S, Li S, Yang J, Liu D. A two-step data augmentation method based on generative adversarial Network for hardness prediction of high entropy alloy. Comput Mater Sci. 2023;220:112064.
19	Chen Q, Liu Z, Huang Y, et al. Predicting natural rubber crystallinity by a novel machine learning algorithm based on molecular dynamics simulation data. Langmuir. 2023;39(48):17088-17099.
20	Elshawi R, Maher M, Sakr S. Automated machine learning: state-ofthe-art and open challenges. arXiv. 2019.
21	Feurer, M; Klein, A; Eggensperger, K; Springenberg, JT; Blum, M; Hutter, F Efficient and robust automated machine learning;2015.
22	Jin H, Song Q, Hu X. Auto-keras: an efficient neural architecture search system. In: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining;KDD ’19. Association for Computing Machinery;2019:1946-1956.
23	Thornton C, Hutter F, Hoos HH, Leyton-Brown K. Auto-WEKA: combined selection and hyperparameter optimization of classification algorithms. In: Proceedings of the 19th ACM SIGKDD international conference on Knowledge discovery and data mining;2013:847-855.
24	Le TT, Fu W, Moore JH. Scaling tree-based automated machine learning to biomedical big data with a feature set selector. Bioinformatics. 2020;36(1):250-256.
25	Ding L. Structure and properties of solution polymerized butadiene styrene rubber, Chinese master’s theses full-text database. 2010. https://kns.cnki.net/kcms/detail/detail.aspx?dbname=CMFD2011&filename=2010252146.nh&dbcode=CMFD
26	Fu Y, Xu Q, Yan F. Comparison of structure and properties of different brands of solution-polymerized styrene-butadiene rubber application for tread compound. China Synth Rubber Ind. 2017;40(5):396-400.
27	Han Y, Yang G, Fei Y, Zhang S, Lv X. Comparison on microstructure and properties of SSBR. China Elastomerics. 2018;28(1):12-16.
28	Hu H, Xu D, Huang X, et al. Comparison of structure and properties of different brands of styrene-butadiene rubber. China Synth Rubber Ind. 2021;44(2):126-129.
29	Lai K, Zhao L, Leng Z. Performance of solution polymerized styrenebutadiene rubber 2506 and its application in tread compound of tire. China Synth Rubber Ind. 2017;40(2):152-156.
30	Qi T. Study on the processing and application of solution polymerized butadiene styrene rubber with different stucture from dushanzi petrochemical corp. Chinese Master’s Theses Full-text Database. 2020.
31	Yang Y. Study on the structures, properties, processing and application of solution-polymerized styrene-butadiene rubber. Chinese Master’s Theses Full-text Database. 2019. https://kns.cnki.net/kcms/detail/detail.aspx?dbname=CMFD2019&filename=1019855925.nh&dbcode=CMFD
32	Song Y, Zhan X, Yang L, Zhao Z, Tong L. Effect of coupling structure on properties of star-shaped structure solution-polymerized styrenebutadiene rubber. China Synth Rubber Ind. 2021;44(3):186-190.
33	Wang S. Technology development and application performance of high vinyl solution polymerized styrene-butadiene rubber. Chinese Master’s Theses Full-text Database. 2012. https://kns.cnki.net/kcms/detail/detail.aspx?dbname=CMFD2012&filename=1012414646.nh&dbcode=CMFD
34	Zhang J, Zhang J, Zhang X. Application of SSBR filled with green oil in the high performance tire compounds. China Elastomerics. 2013;23(3):53-58.
35	Zhang J, Wen Q, Wang Y, Zhang X. Influence of the molecular weight and its distribution on the processability of SSBR. World Rubber Industry. 2013;40(7):28-35.
36	Zhang X, Chen R, Chen M, Li H. Properties of domestic medium vinyl SSBR. Rubber Science and Technology. 2013;11(9):12-17.
37	Du Y. Study on the Structure and Properties of SSBR/NR Blend Vulcanizates. Chinese Master’s Theses Full-text Database;2009. https://kns.cnki.net/kcms/detail/detail.aspx?dbname=CMFD2009&filename=2008157560.nh&dbcode=CMFD
38	Loshchilov I, Hutter FSGDR. Stochastic gradient descent with warm restarts. arXiv. 2017.
39	Breiman L. Random forests. Mach Learn. 2001;45(1):5-32.
40	Friedman JH. Greedy function approximation: a gradient boosting machine. Ann Stat. 2001;29(5).
41	Chen T, Guestrin C. XGBoost: a scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining;KDD’16. Association for Computing Machinery;2016:785-794.
42	Ostroumova, L; Gusev, G; Vorobev, A; Dorogush, AV; Gulin, A CatBoost: unbiased boosting with categorical features;2017.
43	Parzen E. On estimation of a probability density function and mode. Ann Math Statistics. 1962;33(3):1065-1076.
44	Min A, Holzmann H, Czado C. Model selection strategies for identifying most relevant covariates in homoscedastic linear models. Comput Stat Data Anal. 2010;54(12):3194-3211.
45	Gideon S. Estimating the dimension of a model. Ann statistics. 1978;6(2):461-464.
46	Geurts P, Ernst D, Wehenkel L. Extremely randomized trees. Mach Learn. 2006;63(1):3-42.
47	Lundberg S, Lee S-I. A unified approach to interpreting model predictions. In: Proceedings of the 31st International Conference on Neural Information Processing Systems;2017. arXiv: Long Beach California USA.
48	Fox TGJ, Flory PJ. Further studies on the melt viscosity of polyisobutylene. J Phys Chem. 1951;55(2):221-234.