Wu Chenchen, Zhang Ruming, Zhao Pengyuan, Li Liang, Zhang Dingguo
School of Physics, Nanjing University of Science and Technology, Nanjing, 210094, China.
Shenyang Institute of Automation, Chinese Academy of Science, Shenyang, 110016, China.
Sci Rep. 2024 Dec 30;14(1):31860. doi: 10.1038/s41598-024-83379-3.
Molding has been widely used to manufacture thermoset composite structures in the aerospace and automotive industries owing to its efficiency in reducing the number of parts and the manufacturing cost. For such molded composite parts, the degree-of-cure curve is generally used to evaluate the solidification of the resin. Nevertheless, in simulation of cure is not the cure model itself, but rather knowing the initial conditions such as fiber volume fraction, initial curing degree, convective boundary conditions etc. Additionally, solving the heat transfer coupled with the cure kinetics presents additional requirements for time, making artificial intelligence tools promising for these problems. This paper focuses on developing a data-driven approach for predicting the degree-of-cure curve. The simulated degree-of-cure curve for the model corresponds to a specific temperature-time curve was verified by the published value. Then, the temperature-time and the resulting degree-of-cure-time curves obtained from finite element simulations were created for training the prediction models using machine learning approaches of support vector regression (SVR), back propagation (BP) neural network and BP neural network optimized by genetic algorithm (GA-BP). The validation and evaluation indices illustrate that the degree-of-cure curve prediction model trained by the GA-BP neural network yields the highest accuracy.
由于模塑在减少零部件数量和制造成本方面的效率，它已在航空航天和汽车工业中广泛用于制造热固性复合材料结构。对于此类模塑复合材料部件，固化度曲线通常用于评估树脂的固化情况。然而，在固化模拟中，关键不在于固化模型本身，而在于了解诸如纤维体积分数、初始固化度、对流边界条件等初始条件。此外，求解与固化动力学耦合的传热问题对时间有额外要求，这使得人工智能工具在解决这些问题方面具有潜力。本文重点开发一种数据驱动的方法来预测固化度曲线。通过已发表的值验证了对应于特定温度 - 时间曲线的模型模拟固化度曲线。然后，利用支持向量回归（SVR）、反向传播（BP）神经网络和遗传算法优化的BP神经网络（GA - BP）等机器学习方法，创建从有限元模拟获得的温度 - 时间曲线以及相应的固化度 - 时间曲线，用于训练预测模型。验证和评估指标表明，由GA - BP神经网络训练的固化度曲线预测模型具有最高的准确性。
