基于深度神经网络的复合材料设计预测生成框架
Deep neural networks based predictive-generative framework for designing composite materials
摘要:
根据应用要求设计复合材料基本上是一项具有挑战性且耗时的任务。在这里,我们报告了一个基于深度神经网络的计算框架的开发,该框架能够解决正向(预测)和反向(生成)设计问题。预测模型基于流行的卷积神经网络架构,并在有限元模拟的帮助下进行训练。此外,开发的属性预测器模型被用作基于神经网络的发电机模型中的反馈机制。建议的预测生成模型可以用来获得特定的弹性性能的最大化,以及指定的弹性常数的微观结构。在复合材料设计中部署深度学习技术的主要障碍之一是生成训练数据集所需的密集计算资源。为此,提出了一种新的数据增强方案。数据增强方案的应用结果在训练阶段的计算资源的显着节省。所提出的数据增强方法是通用的,可以用于任何涉及周期性微结构的设置。预测生成模型的有效性通过各种例子证明。据设想,开发的模型将显着降低成本和时间与先进应用的复合材料设计过程。
英文摘要:
Designing composite materials as per the application requirements is fundamentally a challenging and time consuming task. Here we report the development of a deep neural network based computational framework capable of solving the forward (predictive) as well as inverse (generative) design problem. The predictor model is based on the popular convolution neural network architecture and trained with the help of finite element simulations. Further, the developed property predictor model is used as a feedback mechanism in the neural network based generator model. The proposed predictive-generative model can be used to obtain the micro-structure for maximization of particular elastic properties as well as for specified elastic constants. One of the major hurdle for deployment of the deep learning techniques in composite material design is the intensive computational resources required to generate the training data sets. To this end, a novel data augmentation scheme is presented. The application of data augmentation scheme results in significant saving of computational resources in the training phase. The proposed data augmentation approach is general and can be used in any setting involving the periodic micro-structures. The efficacy of the predictive-generative model is demonstrated through various examples. It is envisaged that the developed model will significantly reduce the cost and time associated with the composite material designing process for advanced applications.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。