Despite its great scientific and technological importance, wall-bounded turbulence is an unresolved problem that requires new perspectives to be tackled. One of the key strategies has been to study interactions among the coherent structures in the flow. Such interactions are explored in this study for the first time using an explainable deep-learning method. The instantaneous velocity field in a turbulent channel is used to predict the velocity field in time through a convolutional neural network. The predicted flow is used to assess the importance of each structure for this prediction using a game-theoretic algorithm (SHapley Additive exPlanations). This work provides results in agreement with previous observations in the literature and extends them by quantifying the importance of the Reynolds-stress structures, finding a causal connection between these structures and the dynamics of the flow. The process, based on deep-learning explainability, has the potential to shed light on numerous fundamental phenomena of wall-bounded turbulence, including the objective definition of new types of flow structures.
翻译:尽管具有巨大的科学和技术重要性,但墙壁上的动荡是一个需要从新的角度处理的未决问题。关键战略之一是研究流动中连贯结构之间的相互作用。本研究报告首次使用可解释的深层学习方法探讨了这种相互作用。动荡通道的瞬时速度场被用来通过一个卷发神经网络预测速度场。预测流量被用来利用游戏理论算法(SHapley Additive Explectations)评估每一结构对这一预测的重要性。这项工作的结果与文献中先前的观察一致,并通过量化雷诺斯压力结构的重要性、找出这些结构与流动动态之间的因果关系来扩展这些结构。这一进程基于深层学习解释,有可能揭示许多墙壁上动荡的基本现象,包括新类型流动结构的客观定义。