Open Collections

Batouei, Mohammad Amin

University of British Columbia

Composite materials are essential in high-performance industries due to their superior strength-to-weight ratios and design versatility. However, ensuring consistent quality during their manufacturing processes, particularly in the curing stage, remains challenging. The autoclave curing process, involving heat and pressure application to prepreg materials, significantly impacts final mechanical properties. Unoptimized temperature, pressure, and heat transfer coefficient (HTC) selection can lead to considerable product quality deviations. This research explores using synthetic data and Artificial Neural Networks (ANNs), combined with thermal imaging and Haralick texture features, to predict and optimize autoclave curing outcomes. Unlike traditional off-line, simulation-based optimization techniques or experimental trial-and-error methods, the proposed approach enables near real-time control. Its application is demonstrated in three scenarios: (1) ideal conditions with no uncertainties in thermal camera measurements (prediction mode), (2) practical conditions accounting for measurement uncertainties (prediction mode), and (3) multi-objective design optimization using genetic algorithms (reverse engineering mode). SHapley Additive exPlanations (SHAP) sensitivity analysis was employed to quantify the contributions of input factors (Haralick features) to predictions. The results of the first scenario demonstrated that ANN could accurately predict the key process outcomes, such as degree of cure (DOC) and cure time, with an accuracy of > 94%. In the second scenario, the DOC prediction accuracy exceeded 96%, along with a prediction error of a maximum of 1% for the minimum degree of cure (DOC) outcome. Additionally, in both scenarios, the models successfully predicted the gel time and cure time with a maximum error of 6%. A sensitivity analysis, in the second scenario, further confirmed the model’s robustness, with performance remaining stable despite simulated temperature measurement uncertainties. Particularly, when the simulated measured temperature data was largely perturbed, the accuracy remained above 0.84 across all outcomes. In the third scenario, the ANN-guided approach, integrated with genetic algorithms, identified an optimal solution, estimating HTC values and balancing trade-offs between minimum DOC and cure uniformity. Aligned with Industry 4.0, this study highlights AI’s potential for improving on-site monitoring and process optimization in composite autoclave manufacturing through thermal imaging and feature extraction.

Text

2024-12-09

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/89805

Mechanical Engineering

University of British Columbia

UBCO

http://creativecommons.org/licenses/by-nc-nd/4.0/