为了提高圆柱流致振动的研究效率，以计算流体力学数值模拟和动网格自定义函数（user defined function, UDF）技术为基础，结合深度神经网络（deep neural network, DNN）分析法，构建圆柱流致振动升力系数预测模型，并探索圆柱升力系数的规律。结果表明：无论质量比为何值，升力系数随着流速的增大呈现先增后减再稳定的趋势，其大值的集中区间在小质量比和较小速度下，最大值对应的流速约为0.4 m/s。综上所述，DNN升力系数预测模型简单高效，实现快速又准确的数据预测，为研究流致振动（flow-induced vibration, FIV）受力特性提供一种新的思路。

To improve the research efficiency of cylindrical flow-induced vibration and based on computational fluid dynamics numerical simulation and dynamic mesh UDF technology, combined with the analysis method of deep neural network (DNN), a prediction model for the lift coefficient of flow-induced vibration of a cylinder was constructed, and the regularity of the lift coefficient of the cylinder was explored. The results show that regardless of the mass ratio, the lift coefficient exhibits a trend of increasing first, then decreasing, and finally stabilizing with the increase in flow velocity. The range with large values is concentrated under small mass ratios and lower velocities, with the maximum value corresponding to a flow velocity of approximately 0.4 m/s. In summary, the DNN lift coefficient prediction model is simple and efficient, achieving fast and accurate data prediction, providing a new approach for studying the force characteristics of flow-induced vibration (FIV).