为了解决轴流风扇叶型W65优化设计的问题，选取不同的优化方法，包括高斯过程回归方法，人工神经网络方法和序列二次规划方法。首先生成一定区间内的叶型样本集，通过CST参数化方法表示，使用B样条曲线进行光滑化处理，通过CFD模拟仿真方法获得升阻比数据，后进入优化阶段。优化方法分别采用高斯过程回归方法，人工神经网络方法和序列二次规划方法，前两者结合遗传算法和梯度下降法，后者不再结合其他优化方法，对带有面积约束的多攻角升阻比目标函数进行寻优优化，工况中攻角在0°至8°变化，马赫数为0.5。将优化后的叶型通过CFD方法进行验证，结果表明：优化叶型的综合升阻比分别提高了8.41%，8.49%，2.08%，优化方法中预估的相对误差分别为0.25%，-0.39%，6.31%。通过对优化结果的分析，发现高斯过程回归方法和人工神经网络方法优化效果较好，而序列二次规划的误差较大。

In order to solve the problem of optimizing the design of axial fan blade airfoil W65, different optimization methods were selected, including Gaussian process regression method, artificial neural network method, and sequential quadratic programming method. Firstly, generate a sample set of blade airfoils within a certain interval, represent them using CST parameterization method, smooth them using B-spline curves, obtain lift to drag ratio data through CFD simulation method, and then proceed to the optimization stage. The optimization methods include Gaussian process regression, artificial neural network, and sequential quadratic programming. The first two are combined with genetic algorithm and gradient descent method, while the latter no longer combines other optimization methods. The objective function of multi angle of attack lift drag ratio with area constraints is optimized, and the angle of attack varies from 0 ° to 8 ° under operating conditions, with a Mach number of 0.5. The optimized blade airfoil was validated using CFD method, and the results showed that the comprehensive lift to drag ratio of the optimized blade profile increased by 8.41%, 8.49%, and 2.08%, respectively. The estimated relative errors in the optimization method were 0.25%, -0.39%, and 6.31%, respectively. Through the analysis of optimization results, it was found that the Gaussian process regression method and artificial neural network method have better optimization effects, while the error of sequence quadratic programming is relatively large.