Autonomous Agentic AI for Adaptive Cure Optimization and Defect Prevention in Thermoset Polymer Composite Manufacturing

Notice

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.

Year : 2026 | Volume : 14 | 02 | Page :
    By

    Manish Kumar Jha,

  1. PhD, Department of Mathematics (Computer Science Research), Patliputra University, Patna, Bihar, India

Abstract

Thermoset polymer composites occupy a central position in modern structural manufacturing, from aircraft fuselages to wind-turbine blades. Despite progress in resin chemistry and fiber architecture, the “cure process” that transforms compliant preforms into load-bearing structures remains difficult to manage. Manufacturers encounter ‘voids’, “interlaminar delaminations”, and “spring-back distortion” when curing complex or thick-section parts. The cause is not ignorance of the relevant physics, but rather that ‘temperature’, ‘chemistry’, ‘rheology’, and ‘mechanics’ couple in ways that no fixed cure schedule can anticipate across the full range of production variability.

This paper presents the “Autonomous Agentic Artificial Intelligence System for Adaptive Cure Optimization (AAAI-ACO)”, a framework designed to close this gap. Rather than prescribing a cure cycle in advance, the system watches the evolving process state through embedded sensors – “fiber Bragg gratings”, “multi-frequency dielectric probes”, ‘thermocouples’, and “ultrasonic transducers” – fuses those measurements in real time using an Extended Kalman Filter, and feeds the result into a continuously running digital twin built around a physics-informed neural network (PINN). The PINN re-produces the Kamal-Sourour cure kinetics and the coupled energy equation at roughly 200 times the speed of a full finite-element model, making it fast to project process trajectories forward by 30 minutes and evaluate candidate control actions before committing to them.

Decisions are taken by a hierarchy of four software agents: spatially distributed sensing agents for local data quality and anomaly detection; a defect prediction agent running gradient-boosting and LSTM models for probabilistic void, delamination, and residual stress risk estimation; a coordination agent that queries a Proximal Policy Optimization (PPO) deep reinforcement learning policy; and a supervisor agent that enforces hard safety limits throughout. The PPO agent was pre-trained across an initial 50,000 simulated cure cycles before full training extended to 250,000 episodes with varied resin batch properties and geometric conditions, building a policy that generalizes rather than memorizes.

Conceptual validation against Hexcel 8552/AS4 laminate cure data from the published literature projects reductions of approximately 47% in void content, 37% in spring-back distortion, and 20% in energy consumption relative to the manufacturer’s standard cure cycle, alongside an 18.6% reduction in cycle time. These are simulation-derived projections, and the author acknowledges that hardware-in-the-loop testing will be required to confirm framework performance in a real autoclave. The results are internally consistent, benchmarked against comparable studies, and grounded in established control theory and materials science. This work provides a coherent conceptual foundation for researchers working toward autonomous cure management in safety-critical composite production.

Keywords: Thermoset composite curing; Autonomous agentic AI; Reinforcement learning; Digital twin; Defect prevention; Adaptive process control; Multi-agent systems; Industry 4.0.

How to cite this article:
Manish Kumar Jha. Autonomous Agentic AI for Adaptive Cure Optimization and Defect Prevention in Thermoset Polymer Composite Manufacturing. Journal of Polymer & Composites. 2026; 14(02):-.
How to cite this URL:
Manish Kumar Jha. Autonomous Agentic AI for Adaptive Cure Optimization and Defect Prevention in Thermoset Polymer Composite Manufacturing. Journal of Polymer & Composites. 2026; 14(02):-. Available from: https://journals.stmjournals.com/jopc/article=2026/view=239907


References

  1. Soutis C. Fibre reinforced composites in aircraft construction. Prog Aerosp Sci. 2005;41(2):143-151.
  2. Marsh G. Airbus takes wing with composite aircraft. Reinf Plast. 2005;49(3):26-32.
  3. Grand View Research. Composites Market Size, Share & Trends Analysis Report. San Francisco, CA: Grand View Research; 2023.
  4. Twardowski TE, Lin SE, Geil PH. Curing in thick composite laminates: experiment and simulation. J Compos Mater. 1993;27(3):216-250.
  5. White SR, Hahn HT. Cure cycle optimization for the reduction of processing-induced residual stresses in composite materials. J Compos Mater. 1993;27(14):1352-1378.
  6. Abou Msallem Y, Jacquemin F, Boyard N, Poitou A, Delaunay D, Chatel S. Material characterization and residual stresses simulation during the manufacturing process of epoxy matrix composites. Compos Part A. 2010;41(1):108-115.
  7. Bogetti TA, Gillespie JW. Process-induced stress and deformation in thick-section thermoset composite laminates. J Compos Mater. 1992;26(5):626-660.
  8. Johnston AA. An Integrated Model of the Development of Process-Induced Deformation in Autoclave Processing of Composite Structures [PhD dissertation]. Vancouver: University of British Columbia; 1997.
  9. Wood JR, Bader MG. Void control for polymer-matrix composites (1): theoretical and experimental methods for determining the growth and collapse of gas bubbles. Compos Manuf. 1994;5(3):139-147.
  10. Kim YK, Daniel IM. Cure cycle effect on composite structures manufactured by resin transfer molding. J Compos Mater. 2002;36(14):1725-1743.
  11. Wijskamp S, Akkerman R, Lamers EAD. Residual stresses in non-symmetrical carbon/epoxy laminates. In: Proceedings of ECCM-10; 2002; Bruges, Belgium.
  12. Pascault JP, Sautereau H, Verdu J, Williams RJJ. Thermosetting Polymers. New York: CRC Press; 2002.
  13. Hexcel Corporation. HexPly 8552 Epoxy Matrix Product Data Sheet. Stamford, CT: Hexcel Corporation; 2020.
  14. Lu Y. Industry 4.0: a survey on technologies, applications and open research issues. J Ind Inf Integr. 2017;6:1-10.
  15. Tao F, Cheng J, Qi Q, Zhang M, Zhang H, Sui F. Digital twin-driven product design, manufacturing and service with big data. Int J Adv Manuf Technol. 2018;94(9-12):3563-3576.
  16. Pillai VL, Beris AN, Dhurjati PS. Implementation of model-based optimal temperature profiles for autoclave curing of composites using a knowledge-based system. Ind Eng Chem Res. 1994;33(10):2443-2452.
  17. Wooldridge M. An Introduction to MultiAgent Systems. Chichester: John Wiley & Sons; 2009.
  18. Wang J, Ma Y, Zhang L, Gao RX, Wu D. Deep learning for smart manufacturing: methods and applications. J Manuf Syst. 2018;48:144-156.
  19. Sutton RS, Barto AG. Reinforcement Learning: An Introduction. 2nd ed. Cambridge, MA: MIT Press; 2018.
  20. Pantelelis N, Bistekos E. Process optimisation for the production of CFRP aerospace structures. Compos Sci Technol. 2006;66(7-8):949-962.
  21. Schapira Y, Asher R, Azoulay O. A digital twin for composite curing process monitoring. In: Proceedings of the 21st ICCM; 2017.
  22. Li H, Lim SH, Park J. Machine learning prediction of void content in epoxy composites. J Compos Mater. 2021;55(8):1087-1098.
  23. Jordan MI, Mitchell TM. Machine learning: trends, perspectives, and prospects. Science. 2015;349(6245):255-260.
  24. Wuest T, Weimer D, Irgens C, Thoben KD. Machine learning in manufacturing: advantages, challenges, and applications. Prod Manuf Res. 2016;4(1):23-45.
  25. Chen L, Papandreou G, Kokkinos I, Murphy K, Yuille AL. DeepLab: semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs. IEEE Trans Pattern Anal Mach Intell. 2018;40(4):834-848.
  26. Hochreiter S, Schmidhuber J. Long short-term memory. Neural Comput. 1997;9(8):1735-1780.
  27. Shahriari B, Swersky K, Wang Z, Adams RP, de Freitas N. Taking the human out of the loop: a review of Bayesian optimization. Proc IEEE. 2016;104(1):148-175.
  28. Fernandez R, Manzanal D, Larripa M. Neural network modeling of epoxy cure kinetics from dielectric data. Polym Eng Sci. 2019;59(S2):E214-E221.
  29. Struzziero G, Barbezat M, Skordos AA. Optimal cure cycles for thick laminates. Compos Part A. 2019;124:105436.
  30. Mnih V, Kavukcuoglu K, Silver D, et al. Human-level control through deep reinforcement learning. Nature. 2015;518(7540):529-533.
  31. Silver D, Huang A, Maddison CJ, et al. Mastering the game of Go with deep neural networks and tree search. Nature. 2016;529(7587):484-489.
  32. Haarnoja T, Zhou A, Abbeel P, Levine S. Soft actor-critic: off-policy maximum entropy deep reinforcement learning with a stochastic actor. In: Proceedings of ICML; 2018.
  33. Kim J, Lee S, Park C. Reinforcement learning for injection molding process control. Polym Eng Sci. 2021;61(3):789-801.
  34. Bortot P, Gottardo P, Bernardi L. Model-based reinforcement learning for extrusion process optimization. J Manuf Syst. 2022;63:501-514.
  35. Gonzalez AM, Weyrauch F, Rolfes R. Reinforcement learning-based cure cycle adaptation for thick composites. Compos Sci Technol. 2020;197:108275.
  36. Jennings NR, Sycara K, Wooldridge M. A roadmap of agent research and development. Auton Agents Multi-Agent Syst. 1998;1(1):7-38.
  37. Kratz J, Hsiao K, Fernlund G, Hubert P. Thermal models for MTM45-1 and Cycom 5320 OOA prepregs. J Compos Mater. 2013;47(3):341-352.
  38. Leitao P, Restivo F. ADACOR: a holonic architecture for agile and adaptive manufacturing control. Comput Ind. 2006;57(2):121-130.
  39. Trentesaux D, Pach C, Bekrar A, Sallez Y, Berger T, Bonte T, et al. Benchmarking flexible job-shop scheduling and control systems. Control Eng Pract. 2013;21(9):1204-1225.
  40. Park S, Lee J, Chung I. Multi-agent reinforcement learning for autoclave thermal management. IEEE Trans Ind Inform. 2022;18(9):5987-5998.
  41. Kamal MR, Sourour S. Kinetics and thermal characterization of thermoset cure. Polym Eng Sci. 1973;13(1):59-64.
  42. Johnston A, Vaziri R, Poursartip A. A plane strain model for process-induced deformation of laminated composite structures. J Compos Mater. 2001;35(16):1435-1469.
  43. Fernlund G, Poursartip A, Twigg G, Albert C. Residual stress, spring-in and warpage in autoclaved composite parts. In: Proceedings of ICCM-13; 2001; Beijing.
  44. Loos AC, Springer GS. Curing of epoxy matrix composites. J Compos Mater. 1983;17(2):135-169.
  45. Gutowski TG, Cai Z, Bauer S, Boucher D, Kingery J, Wineman S. Consolidation experiments for laminate composites. J Compos Mater. 1987;21(7):650-669.
  46. Mesogitis TS, Skordos AA, Long AC. Stochastic simulation of the influence of cure kinetics variability on composites curing. Compos Sci Technol. 2014;110:145-154.
  47. Potter K, Khan B, Wisnom M, Bell T, Stevens J. Variability, fibre waviness and misalignment in the determination of the properties of composite materials and structures. Compos Part A. 2008;39(9):1343-1354.
  48. Grunenfelder LK, Nutt SR. Void formation in composite prepregs: effect of dissolved moisture. Compos Sci Technol. 2010;70(16):2304-2309.
  49. Centea T, Hubert P. Out-of-autoclave prepreg consolidation under deficient pressure conditions. J Compos Mater. 2014;48(16):2033-2045.
  50. Abdelal G, Robotham A, Cantwell W. Autoclave cure simulation of composite structures applying implicit and explicit FE techniques. Int J Mech Mater Des. 2013;9(1):55-63.
  51. Ersoy N, Ganapathi M, Wisnom M, Potter K. Development of spring-in angle during cure of a thermosetting composite. Compos Part A. 2005;36(12):1700-1706.
  52. Kravchenko OG, Kravchenko SG, Casares A, Pipes RB. Digital image correlation measurement of resin chemical shrinkage in a composite laminate. Compos Struct. 2014;108:849-856.
  53. Grieves M. Digital twin: manufacturing excellence through virtual factory replication [White Paper]. 2014.
  54. Rosen R, von Wichert G, Lo G, Bettenhausen KD. About the importance of autonomy and digital twins for the future of manufacturing. IFAC-PapersOnLine. 2015;48(3):567-572.
  55. Grieves M, Vickers J. Digital twin: mitigating unpredictable, undesirable emergent behavior in complex systems. In: Kahlen FJ, Flumerfelt S, Alves A, editors. Transdisciplinary Perspectives on Complex Systems. Cham: Springer; 2017.
  56. Tao F, Zhang H, Liu A, Nee AYC. Digital twin in industry: state-of-the-art. IEEE Trans Ind Inform. 2019;15(4):2405-2415.
  57. Schapira Y, Azoulay O, Cohen Y. Digital twin for autoclave cure process monitoring using embedded FBG sensors. Compos Part A. 2021;148:106508.
  58. Raju B, Hiremath SR, Mahapatra DR. A review of micromechanics based models for effective elastic properties of reinforced polymer matrix composites. Compos Struct. 2018;204:607-619.
  59. Chern BC, Moon TJ, Howell JR, Tan W. New experimental data and revised cure kinetics for Hercules 3501-6 epoxy. J Compos Mater. 2002;36(17):2061-2072.
  60. Russell S, Norvig P. Artificial Intelligence: A Modern Approach. 4th ed. Hoboken, NJ: Pearson; 2020.
  61. Bellman R. A Markovian decision process. J Math Mech. 1957;6(5):679-684.
  62. Kang HK, Park SO, Kim CS, Hong CS. Cure monitoring of composite laminates using fiber optic sensors. Smart Mater Struct. 1998;7(4):459-469.
  63. Kranbuehl D, Delos S, Hoff M, Weller L, Haverty P, Seeley J. Dynamic dielectric analysis: nondestructive testing for cure state and moisture in thermosets. Polym Eng Sci. 1987;27(1):93-96.
  64. Gill A, Bousie E, Partridge I. Process monitoring of thermoset composites by in-situ ultrasonic instrumented tooling. Compos Part A. 2012;43(5):886-892.
  65. Kalman RE. A new approach to linear filtering and prediction problems. J Basic Eng. 1960;82(1):35-45.
  66. Raissi M, Perdikaris P, Karniadakis GE. Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J Comput Phys. 2019;378:686-707.
  67. Brunke L, Greeff M, Hall AW, Yuan Z, Zhou S, Panerati J, et al. Safe learning in robotics: from learning-based control to safe reinforcement learning. Annu Rev Control Robot Auton Syst. 2022;5:411-444.
  68. Schulman J, Wolski F, Dhariwal P, Radford A, Klimov O. Proximal policy optimization algorithms. arXiv:1707.06347 [Preprint]. 2017.
  69. Raju KVSN, Mouzakis DE, Karger-Kocsis J. An experimental investigation of cure kinetics of Hexcel 8552/AS4 for composite process simulation. Polym Compos. 2014;35(2):285-293.
  70. Kim BS, White SR. Process-induced residual stress analysis of AS4/3501-6 composite materials. Mech Adv Mater Struct. 1997;4(1):29-72.
  71. Kaufman M, Klotzsche M. Cost drivers in composite manufacturing: reducing the assembly burden in aerospace structures. SAMPE J. 2017;53(4):18-28.
  72. Xu X. From cloud computing to cloud manufacturing. Robot Comput-Integr Manuf. 2012;28(1):75-86.
  73. Reifsnider K, Case S, Duthoit J. The mechanics of composite structures: a science-based approach to the digital thread for composite structural integrity. Compos Sci Technol. 2000;60(12-13):2539-2551.
  74. Kang MK, Lee EH, Kim HS, Kim CG. Development of MEMS-based strain gauges for in-situ composite cure monitoring. Sens Actuators A Phys. 2018;273:30-38.
  75. Schilling PJ, Karedla BR, Tatiparthi AK, Verges MA, Herrington PD. X-ray computed microtomography of internal damage in fiber reinforced polymer matrix composites. Compos Sci Technol. 2005;65(14):2071-2078.
  76. McMahan HB, Moore E, Ramage D, Hampson S, Arcas B. Communication-efficient learning of deep networks from decentralized data. In: Proceedings of AISTATS; 2017.
  77. Ruiz E, Achim V, Soukane S, Trochu F, Breard J. Optimization of injection flow rate to minimize micro/macro-voids formation in resin transfer molded composites. Compos Sci Technol. 2006;66(3-4):475-486.
  78. Korda N, La P. On Thompson sampling for reinforcement learning and bandits. IEEE Trans Inf Theory. 2012;59(7):4220-4243.
Ahead of Print Subscription Original Research
Volume 14
02
Received 20/03/2026
Accepted 27/03/2026
Published 10/04/2026
Publication Time 21 Days
6

Login


My IP

PlumX Metrics