ePrints

Modelling the Degree of Cure in a Reactive Polymer Composite Moulding Process Through Hybrid Mechanistic Neural Network Models

Lookup NU author(s): Dr Jie ZhangORCiD

Downloads

• Published version [.pdf]

Licence

This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0).

Abstract

© The Author(s) 2025.This paper presents a hybrid mechanistic/neural network model for modelling an industrial fibre-reinforced reactive polymer composite curing process. Accurate predictions of the degree of cure are very important for the control and optimisation of polymer composite curing process. Hybrid models combine the advantages of both mechanistic model and neural network model. Two hybrid model structures are developed in this paper: one with parallel scheme and another one with the combination of series and parallel schemes. Parameters of the mechanistic model are first identified using process operation data through nonlinear least squares optimisation. Then prediction errors of the mechanistic model are compensated using a neural network. In the hybrid model with parallel structure, both the mechanistic model and the neural network model take the curing temperature and curing time as the model inputs. The neural network model predicts the mechanistic model errors and the predictions of both models are combined in order to reduce the overall model errors in predicting the degree of cure. In the combination of series and parallel scheme, the mechanistic model predicted degree of cure is taken as an additional input to the neural network which predicts the mechanistic model errors. It is found that the hybrid model with the combination of series and parallel schemes gives better performance. Applications to an industrial reactive polymer composite moulding process show that the developed hybrid model is more accurate than its mechanistic and neural network counterparts in predicting the degree of cure. The hybrid model is 7.7% and 17.1% more accurate than the neural network model and the mechanistic model respectively in terms of sum of absolute errors. The results also demonstrate the importance to have the modelling data covering the intended operating range of the process.

Publication metadata

Author(s): Sells S, Zhang J

Publication type: Article

Publication status: Published

Journal: SN Computer Science

Year: 2025

Volume: 6

Issue: 5

Online publication date: 03/06/2025

Acceptance date: 20/05/2025

Date deposited: 16/06/2025

ISSN (print): 2662-995X

ISSN (electronic): 2661-8907

Publisher: Springer

URL: https://doi.org/10.1007/s42979-025-04070-6

DOI: 10.1007/s42979-025-04070-6

Data Access Statement: Data will be made available on reasonable request and subject to industrial approval

Altmetrics

Share