This is an unedited manuscript accepted for publication and provided as an Article in Press for early access at the author’s request. The article will undergo copyediting, typesetting, and galley proof review before final publication. Please be aware that errors may be identified during production that could affect the content. All legal disclaimers of the journal apply.
M.L.S.N.S Lakshmi,
Janani S,
Dinesh P S,
C. Ramya,
P.Senthil Kumar,
M.Sahaya Sheela,
- Professor, Department of Electronics and Communication Engineering, Ramachandra college of Engineering, Eluru, Andhra Pradesh, India
- Assistant Professor, Department of Computer Science and Engineering, KGISL Institute of Technology, Coimbatore, Tamil Nadu, India
- Associate Professor, Department of Computer Science and Engineering , Bannari Amman Institute of Technology, Sathyamangalam, Tamil Nadu, India
- Assistant Professor, Department of Information Technology, Nandha College of Technology, Perundurai, Tamil Nadu, India
- Professor, Department of Mechanical Engineering, Maha Barathi Engineering College, Chinna salem, Kallakurichi (Dt), Tamil Nadu, India
- Associate Professor, Department of Electronics and Communication Engineering, Vel Tech Rangarajan Dr.Sagunthala R&D Institute of Science and Technology Chennai, Tamil Nadu, India
Abstract
The prediction of the elasticity and mechanical strength of functional polymer materials is one of the key requirements for characterizing the structure–property relationship of such materials, which can greatly speed up polymer design, reduce the need for experimental characterization and enable the development of materials with mechanical properties appropriate to their functions. The existing works are based on the traditional machine learning model or deep learning model, which cannot be applied to model the joint effect of the information of chemical sequence, molecular connectivity, and long-range structural dependency. In addition, the prediction variability and the nonlinear structure–property relationships are hard to overcome as well as complex polymer topologies for accurate estimation of mechanical properties. In this paper, a hybrid deep learning framework is introduced, which integrates a Molecular Transformer and Graphormer, to give a comprehensive view of polymer structure–property. To tackle these shortcomings, a hybrid deep learning framework is presented here that integrates a Molecular Transformer and Graphormer to create comprehensive polymer structure–property representation. To ensure the connectivity, branching and structural relationships of the polymers are maintained, a Topology-Aware Graph Encoding (TAGE) mechanism was added. The Molecular Transformer captures the sequence level chemical dependency while the Graphormer is trained topologically-aware without relying on 3D spatial data. The complementary representations thus provide for a simultaneous prediction of elasticity and mechanical strength. Experimental validation shows a 98.42% prediction accuracy which is higher than that of D-MPNN (91.84%), GIN (93.12%), GATv2 (94.26%), Graphormer (95.31%) and Molecular Transformer (95.74%). The proposed model achieves the elasticity RMSE of 0.071, mechanical-strength MAE of 0.082, elasticity R² of 0.986 and mechanical-strength R² of 0.982.
Keywords: Prediction of Elasticity, Polymer Materials, Deep Learning, Mechanical Strength, Topology-Aware Graph Encoding (TAGE), and Molecular Transformer.
References
1. Y. Liang, X. Wei, Y. Peng, X. Wang, and X. Niu, “A review on recent applications of machine learning in mechanical properties of composites,” Polymer Composites, vol. 46, pp. 1939-1960, 2025.
2. N. Brimo, D. Ç. Serdaroğlu, and H. Muhittin, “Artificial Intelligence and Machine Learning for Material Characterizations and Mechanical Properties,” in Nano-Biomaterials in Tissue Repair and Regeneration: Materials Characterization, ed: Springer, 2024, pp. 55-72.
3. C. Liu, R. Guo, and Y. Su, “A deep learning based prediction model for effective elastic properties of porous materials,” Scientific Reports, vol. 15, p. 6707, 2025.
4. S. Ferdousi, Z. Demchuk, W. Choi, R. C. Advincula, and Y. Jiang, “A deep learning and finite element approach for exploration of inverse structure–property designs of lightweight hybrid composites,” Composite Structures, vol. 365, p. 119179, 2025.
5. Y. Sowjanya, S. Gopalakrishnan, and R. D. Kumar, “Elevating IoT healthcare security using ProSRN and ICOM methodologies for effective threat management,” International Journal of Information Technology, vol. 17, pp. 5421-5439, 2025.
6. S. Akash, S. Tiwari, C. Bijudas, B. Santhosh, and P. Sunil, “Deep Learning Based Uncertainty Analysis on Elastic Properties of Unidirectional Composites,” Procedia Structural Integrity, vol. 71, pp. 180-187, 2025.
7. S. Gopalakrishnan, D. Dhinakaran, S. Edwin Raja, P. Raghavan, M. Girija, and R. Sugumar, “Fusion-Driven Medical Image Encryption Framework with Entropy-Calibrated Control and Integrity Assurance,” KSII Trans. Internet Inf. Syst., vol. 20, pp. 1017-1036, 2026.
8. M. Ganesh, P. Senthil, P. J. Jayarin, R. L. Priya, and S. Ramani, “An intuitive security paradigm based on novel deep learning model for the detection of cyber-assaults in IoHT networks,” Iranian Journal of Science and Technology, Transactions of Electrical Engineering, vol. 50, pp. 257-288, 2026.
9. G. Gebremaryam, K. Shahapurkar, V. Chenrayan, G. Kanaginahal, P. V. Rojas, K. P. Arunachalam, et al., “The structural and thermal integrity of novel bio-polymer composite processed from Ethiopian teff husk particles for constructional applications,” Journal of Polymer Research, vol. 32, p. 6, 2025.
10. P. Pandey, M. Verma, S. Lakhanpal, A. K. Bishoyi, R. Roopashree, M. Kaur, et al., “An updated review on the nanocarriers based co‐delivery system of chemo drug doxorubicin and phytocompounds,” Polymers for Advanced Technologies, vol. 36, p. e70050, 2025.
11. M. H. Rahman, M. Arifuzzaman, M. E. Haque, R. S. Naidu, M. E. Hoque, and M. A. Martuza, “Machine Learning-Based Prediction of Polymer Properties Using Structure–Property Relationship Modeling,” Polymers, vol. 18, p. 1320, 2026.
12. A. E. Korayem, C. K. R. Emani, S. Kundurthi, K. Venkatesan, S. Chen, J. Mapkar, et al., “Prediction of In‐Plane Shear Mechanical Properties Through Degree of Bonding Characterization as a Function of Automated Fiber Placement Processing Parameters for Carbon Fiber/PEKK Composite Laminates,” Polymer Composites, 2026.
13. L. Zhang, X. Sun, J. Li, L. Liu, T. Liao, Y. Lu, et al., “Machine learning prediction of mechanical properties of polyolefins from stress-strain curves,” Polymer, p. 129574, 2026.
14. S. Samine, M. Karouchi, M. Zemzami, N. Hmina, M. Lagache, and S. Belhouideg, “Deep Learning and Multiple Linear Regression for Predicting Mechanical Properties from Elastic Constants: Application to Aluminum-Based Alloys Used in 3D Printing,” Results in Engineering, p. 109602, 2026.
15. H. You, L. He, S. Zheng, and H. Li, “Facilitating small-data deep learning for predicting polymer mechanical properties using complex discrete networks via community detection algorithms,” Engineering Fracture Mechanics, p. 111984, 2026.
16. H. Khosravi, M. M. Pour, and M. Bahram, “Prediction of compressive strength of recycled concrete cylinders reinforced with fiber-reinforced polymer using a combined genetic algorithm and neural network approach,” Scientific Reports, 2026.
17. Dataset source: https://polymer.nims.go.jp/
Journal of Polymer & Composites
| Volume | 14 | |
| 05 | ||
| Received | 18/08/2026 | |
| Accepted | 06/10/2026 | |
| Published | 17/10/2026 | |
| Publication Time | 60 Days |