非线性回归算法在飞机载荷标定试验中的应用  

Application of Nonlinear Regression Algorithm to Aircraft Load Calibration Test

作　　者：韩聪玲 刘克格 刘彦鹏 Han Congling;Liu Kege;Liu Yanpeng(Beijing Aircraft Strength Research Institute Co.,Ltd.,Beijing100083,China)

机构地区：[1]国机集团北京飞机强度研究所有限公司,北京100083

出　　处：《工程与试验》2022年第4期31-35,共5页Engineering and Test

基　　金：国家自然科学基金(50135010)。

摘　　要：飞机载荷标定试验可以在载前测试中确定典型结构剖面承受载荷与应变之间的传递关系,多元线性回归算法是当前该试验最常用的应变-载荷求解方法。针对以往一些试验数据线性回归计算结果不够理想的情况,采用非线性回归算法进行计算,进行了基于决策树的随机森林、极端随机树、XGBoost模型、GBDT等4种非线性回归模型的研究。其中,GBDT计算模型可以将预测误差控制到3%左右,验证了基于机器学习的非线性回归算法在标定试验方程求解过程中的有效性,为标定试验方程求解提供了一种思路。Aircraft load calibration test can determine the load bearing capacity and load transmission of typical structural sections.Multiple linear regression algorithm is the most commonly used strain load solution method in the test.In view of the unsatisfactory linear regression calculation results of some previous test data,this paper adopts the nonlinear regression algorithm for calculation,and carries out the research and calculation of four nonlinear regression models based on decision tree,such as random forest,extreme random tree,XGBoost model and GBDT.GBDT can control the prediction accuracy to about 3%,and the effectiveness of the nonlinear regression algorithm based on machine learning in solving the calibration test equation is verified,which provides an idea for solving the calibration test equation.

关 键 词：回归分析 非线性回归 标定试验 

分 类 号：V217[航空宇航科学与技术—航空宇航推进理论与工程]