Development of an artificial neural network for predicting energy absorption capability of thermoplastic commingled composites
Development of an artificial neural network for predicting energy absorption capability of thermoplastic commingled composites
复制标题
DOI：
10.1016/j.compstruct.2020.113131
复制
发表时间：
2021-02-01
影响因子：
6.3
通讯作者：
Gomes, G. F.
中科院分区：
文献类型：
--
作者：
Di Benedetto, R. M.;Botelho, E. C.;Gomes, G. F.
包括人工神经网络和机器学习在内的软计算技术为复合材料的行为预测模型提供了新的可能性。本研究在综合实验结果、多元回归分析模型和析因设计方法的基础上，以耐撞性为背景，建立了预测热塑性复合材料冲击吸能能力的人工神经网络。此外，该项目的科学方法包括(I)开发用于设计和制造新的复合材料部件的智能模型，(Ii)应用计算方法来预测材料性能和行为，以及(Iii)制造工艺的优化。该方案的创新之处在于开创了利用计算方法来描述热塑性复合材料的力学和结构性能，并开发了一种人工神经网络来预测这些材料的吸能能力，同时考虑了聚合物基质的一些性质、热降解动力学模型和固结参数。冲击试验结果表明，该方法能较好地预测冲击能量。使用分析模型数据库作为ANN的输入是一种提高ANN可靠性和准确性的创新方法。
Soft computing techniques including artificial neural networks (ANN) and machine learning reflect new possibilities to behavior prediction models of commingled composites. This study focuses on developing an artificial neural network capable of predicting the impact energy absorption capability of thermoplastic commingled composites, in the context of crashworthiness, based on a compilation of experimental results, multiple regression analytical model and factorial design method. Furthermore, the scientific approach of this project comprises the (i) development of intelligent models for designing and manufacturing of new composite components, (ii) application of computational methods to predict material performance and behavior, and (iii) optimization of manufacturing processes. The innovativeness of this proposal is to initiate the use of computational methods to describe mechanical and structural properties of thermoplastic commingled composite materials and the development of an artificial neural network able to predict the energy absorption capability of these materials, considering some properties of polymer matrix, thermal degradation kinetics model and consolidation parameters. The obtained results from impact testing indicate that the proposed approach can predict the impact energy with satisfactory accuracy. The use of an analytical model database as input for the ANN is an innovative methodology to increase the reliability and accuracy of the ANNs.