This paper describes a new surface-to-surface contact search method for contact problems between curved surfaces defined by the non-uniform rational B-spline basis functions. The method consists of a global search method based on the hierarchical virtual axis aligned bounding boxes and a local contact search method enhanced by the deep learning. The artificial neural network is employed to simulate the mapping from the parameters of contacting surface to the contact state: whether contact occurs or not, where the contact occurs and how far the slave surface penetrates to the master surface. With the proposed method, the time-consuming local contact search is successfully performed independently from the dynamic contact analysis part, where its computational efficiency is quantitatively discussed through some numerical examples.

中文翻译：

---

一种深度学习增强的面对面接触搜索方法

本文描述了一种新的面对面接触搜索方法，用于由非均匀有理 B 样条基函数定义的曲面之间的接触问题。该方法由基于分层虚拟轴对齐边界框的全局搜索方法和通过深度学习增强的局部接触搜索方法组成。采用人工神经网络模拟从接触面参数到接触状态的映射：是否发生接触、接触发生的位置以及从面穿透到主面的距离。使用所提出的方法，独立于动态接触分析部分成功地执行了耗时的局部接触搜索，其中通过一些数值例子定量讨论了其计算效率。