为实现复杂汽轮机叶片流域的快速计算,本文针对汽轮机高温段变反动度叶片流域,提出一种非侵入式POD-ROM(Proper Orthogonal Decomposition–Reduced Order Model)方法。首先,利用ANSYS CFX对叶片流场进行高保真全阶仿真,获取全阶快照数据；随后,采用POD对全阶数据进行分解,根据各阶模态的能量占比选取适当数量的模态,将全阶数据正交投影至模态空间以获得模态系数；进一步,基于符号回归方法构建边界参数与模态系数之间的映射关系,从而实现叶片流场的快速预测。预测结果表明,该方法的L2相对误差低于1%,并在计算效率上实现了约5.18×106倍的加速,所构建的快速计算模型在准确性与效率方面表现出一定优势。

To enable rapid computation of complex turbine blade flow fields, this study proposes a non-intrusive POD-ROM (Proper Orthogonal Decomposition–Reduced Order Model) method for the flow field of variable-reaction blades in the high-temperature section of a steam turbine. First, high-fidelity full-order simulations of the blade flow field are performed using ANSYS CFX to obtain full-order snapshot data. Subsequently, proper orthogonal decomposition (POD) is applied to the full-order data, and an appropriate number of modes are selected based on their energy contributions. The full-order data are then projected onto the modal space to obtain the corresponding modal coefficients. Furthermore, symbolic regression is employed to establish the mapping between boundary parameters and modal coefficients, enabling rapid prediction of the blade flow field. The prediction results indicate that the proposed method achieves an L2 relative error below 1% and accelerates the computation by approximately 5.18×10? times, with the developed rapid computational model demonstrating certain advantages in both accuracy and computational efficiency.