【摘要】 薄板结构广泛应用于航空航天、各种机械设备制造等领域,而其热环境下的固有频率检测是其重要的研究内容之一。文中提出了1种检测热环境下薄板固有频率的物理信息神经网络(PINN)方法。首先通过经典小挠度理论和汉密尔顿原理获取控制方程,其次搭建前向神经网络预测挠度,最后通过自动微分模块计算控制方程偏差和边界条件偏差对网络参数进行迭代修正,得到薄板固有频率检测模型。最后验证PINN在检测热环境下薄板固有频率问题的有效性,结果表明热环境下薄板固有频率检测误差小于2.4%。更多还原

【Abstract】 Thin plate structures are widely used in aerospace,machinery manufacturing,and various engineering fields.The detection of their natural frequencies under thermal environments is a crucial research topic.This paper proposes a PhysicsInformed Neural Network(PINN) approach for detecting the natural frequency of thin plates under thermal conditions.First,the governing equation is derived based on the classical small deflection theory and Hamilton’s principle.Then,a forward neural network is constructed to predict deflections.Finally,the automatic differentiation module is utilized to calculate the deviation of the governing equation and boundary conditions,iteratively adjusting network parameters to develop a detection model for the natural frequency of thin plates.The effectiveness of the proposed PINN method is validated,and it is shown that the detection error of the natural frequency for thin plates under thermal environment is less than 2.4%.更多还原