The multilevel structure of the fabric and rubber in the fabric rubber composite seals makes it challenging to design the mechanical properties and the seal performance of the seal structure efficiently and accurately. In the present work, a machine learning framework is proposed to predict the sealing performance， as well as the mechanical properties， of fabric rubber composite seal combining the finite element and artificial intelligence methods. First, the input dataset is generated by establishing the finite element models considering the woven form of the fabric and the anisotropic hyperelastic behavior of the fabric rubber composites. Second, the dataset obtained by simulation is utilized to train, validate, and evaluate the artificial neural network. Finally, the trained network is well evaluated using the experimental and simulation results. The proposed machine learning framework provides the possibility to link the macroscopic performance with multilevel structures of composite materials, and demonstrates the applicability to predict the property-performance relationship of complex materials with multilevel structures.

中文翻译：

---

利用机器学习方法设计多层结构织物橡胶复合密封件

织物橡胶复合密封件中织物和橡胶的多级结构使得高效、准确地设计密封结构的机械性能和密封性能具有挑战性。在目前的工作中，提出了一种机器学习框架，结合有限元和人工智能方法来预测织物橡胶复合密封件的密封性能以及机械性能。首先，通过考虑织物的编织形式和织物橡胶复合材料的各向异性超弹性行为建立有限元模型来生成输入数据集。其次，利用模拟获得的数据集来训练、验证和评估人工神经网络。最后，使用实验和模拟结果对训练后的网络进行了很好的评估。所提出的机器学习框架提供了将复合材料的宏观性能与多级结构联系起来的可能性，并证明了预测具有多级结构的复杂材料的性能-性能关系的适用性。