The cure stage of thermosetting composites production is critical for the overall process duration and manufacturing costs. While process simulations are commonly used to estimate cure behaviour, real-time predictive capabilities in resin transfer moulding (RTM) remain limited, primarily due to the computational cost of finite element (FE) methods. To address this gap and accurately estimate cure process parameters in RTM, this study proposes a surrogate cure simulation approach based on two state-of-the-art machine learning (ML) voting ensemble models – XGBoost and Light Gradient Boosting Machine – designed to predict cure time and temperature overshoot. To train the model, the cure of an epoxy/carbon fibre flat plate was simulated using the FE solver Marc, providing data over a wide range of conditions. The predictions of temperature overshoot and cure time demonstrate remarkable consistency and high accuracy (R2 values up to 98%) with execution times under 30 ms for both variables. Performance was validated against unseen simulation data and further verified through RTM manufacturing trials and differential scanning calorimetry (DSC), confirming cure completion and a final glass transition temperature of 191°C–194°C. Unlike existing studies that remain simulation-focused, this approach bridges process simulation and data-driven modelling, offering a practical tool for real-time optimisation in industrial RTM applications.
RaiNPitchumaniR. Optimal cure cycles for the fabrication of thermosetting-matrix composites. Polym Compos1997; 18: 566–581.
2.
CarlonePPalazzoGS. A simulation based metaheuristic optimization of the thermal cure cycle of carbon-epoxy composite laminates. In: AIP Conference Proceedings, 2011.
3.
MawardiAPitchumaniR. Optimal temperature and current cycles for curing of composites using embedded resistive heating elements. J Heat Transfer2003; 125: 126–136.
4.
SkordosAPartridgeI. Inverse heat transfer for optimization and on-line thermal properties estimation in composites curing. Sci. Eng2004; 12: 157–172.
5.
AleksendrićDCarlonePVelimirĆ. Optimization of the temperature-time curve for the curing process of thermoset matrix composites. Appl Compos Mater2016; 23: 1047–1063.
6.
StruzzieroGTeuwenJJE. Effect of convection coefficient and thickness on optimal cure cycles for the manufacturing of wind turbine components using VARTM. Compos Part Appl Sci Manuf2019; 123: 25–36.
7.
MatsuzakiRYokoyamaRKobaraT, et al.Multi-objective curing optimization of carbon fiber composite materials using data assimilation and localized heating. Compos Part Appl Sci Manuf2019; 119: 61–72.
8.
TifkitsisKIMesogitisTSStruzzieroG, et al.Stochastic multi-objective optimisation of the cure process of thick laminates. Compos Part Appl Sci Manuf2018; 112: 383–394.
9.
CassolaSDuhovicMSchmidtT, et al.Machine learning for polymer composites process simulation – a review. Composites Part B2022; 246: 110208.
10.
MendikuteJPlazaolaJBaskaranM, et al.Impregnation quality diagnosis in resin transfer moulding by machine learning. Composites Part B2021; 221: 108973.
11.
StieberSSchröterNSchiendorferA, et al.FlowFrontNet: improving carbon composite manufacturing with CNNs. In: DongYMladenićDSaundersC (eds). Machine Learning and Knowledge Discovery in Databases. Applied Data Science Track. Springer International Publishing, 2021, pp. 411–426.
12.
LavaggiTSamizadehMNiknafs KermaniN, et al.Theory-guided machine learning for optimal autoclave co-curing of sandwich composite structures. Polym Compos2022; 43: 5319–5331.
13.
ZobeiryNEeDFloydAP. Theory-guided machine learning composites processing modelling for manufacturability assessment in preliminary design, 2019.
14.
WangYXuSBwarKH, et al.Application of machine learning for composite moulding process modelling. Compos Commun2024; 48: 101960.
15.
MujtabaAIslamFKaedingP, et al.Machine-learning based process monitoring for automated composites manufacturing. J Intell Manuf2025; 36: 1095–1110.
16.
WuCZhangRZhaoP, et al.Curing simulation and data-driven curing curve prediction of thermoset composites. Sci Rep2024; 14: 31860.
17.
ZhangFGuoHWangW, et al.Fast prediction of curing residual stress in polymer-matrix composites by a deep learning method. Polym Compos2025; 46: 8424–8441.
18.
HumfeldKDKimGYJeonJH, et al.Co-training of multiple neural networks for simultaneous optimization and training of physics-informed neural networks for composite curing. Compos Part Appl Sci Manuf2025; 193: 108820.
19.
TifkitsisKIWinistoerferASkordosA. Online optimisation and active control of the cure process of thick composite laminates. J Manuf Process2023; 87: 221–230.
20.
KarkanasPPartridgeI. Cure modeling and monitoring of epoxy/amine resin systems. I. Cure kinetics modeling. J Appl Polym Sci2000; 77: 1419–1431.
21.
PascaultJPWilliamsRJJ. Relationships between glass transition temperature and conversion. Polym Bull1990; 24: 115–121.
22.
MesogitisTSkordosALongAC. Stochastic simulation of the influence of cure kinetics uncertainty on composites cure. Compos Sci Technol2015; 110: 145–151.
23.
StruzzieroGSkordosA. Multi-objective optimisation of the cure of thick components. Compos. Part Appl. Sci. Manuf2017; 93: 126–136.
24.
FarmerJDCovertEE. Thermal conductivity of a thermosetting advanced composite during its cure. J Thermophys Heat Tran1996; 10: 467–475.
25.
Marc® 2023.2 volume A: theory and user information. Hexagon, 2023.
26.
Marc® 2023.2 volume B: element library. Hexagon, 2023.
27.
Marc® 2023.2 volume D: user subroutines and special routines. Hexagon, 2023.
28.
SkordosAPartridgeI. Cure kinetics modeling of epoxy resins using a non-parametric numerical procedure. Polym Eng Sci2001; 41: 793–805.
29.
Hexcel® RTM 6 180 0C epoxy system for resin transfer moulding monocomponent system product data, 2009. https://www.hexcel.com/
30.
MesogitisTSkordosALongA. Stochastic heat transfer simulation of the cure of advanced composites. J Compos Mater2016; 50: 2971–2986.
31.
Hexcel® RTM 6 180 0C epoxy system for resin transfer moulding monocomponent system product data, 2016. https://www.hexcel.com/
García FernándezFde PalaciosPGarcía-IruelaA, et al.Using bootstrapping to determine artificial neural network confidence intervals—case study of particleboard internal bond determined from production data. Appl Sci2025; 15: 4554.
34.
FriedmanJH. Greedy function approximation: a gradient boosting machine. Ann Stat2001; 29: 1189–1232.
35.
OmotehinwaTOOyewolaDO. Hyperparameter optimization of ensemble models for spam email detection. Appl Sci2023; 13: 1971.
36.
WangBZhaoJGuoZ, et al.Failure envelope prediction of 2D SiCf/SiC composites based on XGBoost model. Compos Part Appl Sci Manuf2024; 185: 108287.
37.
ZhangLGuoZTaoQ, et al.XGBoost-based short-term prediction method for power system inertia and its interpretability. Energy Rep2023; 9: 1458–1469.
38.
WortsmanMIlharcoGGadreSY, et al.Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time. In: Proceedings 39th International Conference on Machine Learning. PMLR, 2022, pp. 23965–23998. https://proceedings.mlr.press/v162/wortsman22a.html
39.
ShahhosseiniMHuGPhamH. Optimizing ensemble weights and hyperparameters of machine learning models for regression problems. Mach Learn Appl2022; 7: 100251.
40.
LeeJDuhovicMMayD, et al.Physics-informed neural networks for real-time simulation of transverse liquid composite moulding processes and permeability measurements. Compos Part Appl Sci Manuf2025; 193: 108857.
41.
MulliganTGnaniahSSimsG. Measurement good practice guide no. 62: thermal analysis techniques for composites and adhesives. 2nd ed.National Physical Laboratory.
MesogitisTKratzJSkordosA. Heat transfer simulation of the cure of thermoplastic particle interleaf carbon fibre epoxy prepregs. J Compos Mater2019; 53: 2053–2064.
44.
GuoYYangBAlqawzaiS, et al.Machine learning-based and interpretable models for predicting the resistance and probability of progressive collapse of steel frame-composite floor structures. Eng Struct2025; 332: 120089.
