Abstract: The operation of flood gate is often affected by flow-induced vibration. Although some physical characteristics of specific gate types are known, it is still challenging to predict the dynamic behavior of the gate under unexpected conditions. A hybrid modeling system which combines machine learning with physics-based modeling is proposed to predict the vibration, to deal with possible emergencies. The gate response data is obtained by sensor and stored in database. For the gate under the condition of submerged flow, its safety state is determined by gate opening and equivalent velocity. The finite element model based on fluid-structure coupling is used to provide supplementary data to the system database. This system provides reference and suggestions for the gate control and flood management.

Key words: flow-induced vibration, machine learning, finite element model

摘要： 防洪闸门的运行常受到流激振动的影响，尽管对特定闸门类型的物理特性有所了解，但要判断非预期条件下闸门的动态行为仍然具有挑战性。笔者提出了一种混合建模系统，将机器学习与基于物理的建模相结合以预测振动，从而应对可能出现的紧急情况。闸门响应数据由传感器获取并存储在数据库中，对于淹没流条件下的闸门，根据闸门开度和折合流速划分安全和不安全情况。利用基于流固耦合的有限元模型向系统数据库提供补充数据。该系统为闸门控制和防洪管理提供了安全有效的参考建议。

关键词: 流激振动, 机器学习, 有限元模型