叶片剖面型线的配准与轮廓度评定方法研究

Study on Registration and Profile Error Evaluation of Blade Section

针对叶片的叶型轮廓度计算难题,本文将叶身剖面型线的配准过程分为粗配准和精配准。前者通过变换矩阵H1实现,为实测数据与理论数据之间的刚性变换;后者通过变换矩阵H2实现,以距离平方和作为目标函数,采用随机梯度下降法计算H2中的旋转角和平移量。完成叶型配准后,应用实测数据点到最近邻的两个理论数据点所在直线的距离来计算叶型轮廓度误差,并进行符号判断。试验选取10组叶型数据点进行比对分析,各组叶型轮廓度误差的评定结果偏差均≤±7μm,从而验证了本文算法的有效性和精度水平。

Aiming at the difficulties in profile error calculation of blades,the registration process of blade profile lines was divided into coarse registration and fine registration.The former was achieved through the transformation matrix H1,which is a rigid transformation between measured data and theoretical data.The latter was achieved by transforming the matrix H2,using the sum of squared distances as the objective function,and using the random gradient descent method to calculate the rotation angle and displacement in H2.After completing the registration of the blade profile,the distance between the measured data points and the two nearest theoretical data points on the straight line was applied to calculate the blade profile error and perform symbol judgment.10 sets of blade profile data points were selected for comparative analysis in the experiment.The deviation of the evaluation results for the profile error of each group of blades was smaller than±7μm.The effectiveness and accuracy level of the methods have been verified.