基于神经网络的飞机关键结构载荷预测方法研究  

Research on loads prediction method of key structure of aircraft based on neural network

作　　者：薛海峰 张彦军 宁宇 XUE Haifeng;ZHANG Yanjun;NING Yu(Department of Strength Design,AVIC The First Aircraft Institute,Xi’an 710089,China)

机构地区：[1]航空工业第一飞机设计研究院强度设计研究所,西安710089

出　　处：《航空工程进展》2025年第1期151-157,168,共8页Advances in Aeronautical Science and Engineering

摘　　要：飞机使用数据的关键结构载荷预测对飞机的损伤分析和寿命预测具有重要的作用,可以为实现飞机的主动视情维修提供技术支撑。采用机器学习方法中的前馈神经网络建立大型运输机尾翼根部剪力、弯矩、扭矩载荷模型,与有限元模型计算的载荷进行对比,并且将前馈神经网络的预测结果与随机森林方法和多元线性回归方法的预测结果进行对比和分析。结果表明:前馈神经网络的预测结果的相对误差满足工程要求,前馈神经网络的预测精度优于随机森林方法和多元线性回归方法,为构建基于数据驱动的飞机结构损伤分析和寿命评估体系提供了重要的技术支持。The load prediction based on data used in aircraft plays an important role in the damage analysis and life prediction of aircraft,which can provide the technical support for the active on-condition maintenance of aircraft.The forward neural network of machine learning method is used to establish the load model of shear force,bending moment and torque of the tail wing root of a large transport aircraft.The load calculated by finite element model is compared.The prediction result of forward neural network is compared and analyzed with that of random forest method and multivariate linear regression method.The results show that the prediction errors of the neural network method can satisfy the engineering requirements.The prediction result of forward neural network is superior to that of the random forest method and multivariate linear regression method,which can provide important technique sup⁃port for aircraft structure damage analysis and life evaluation system based on data driven.

关 键 词：神经网络 飞行参数 载荷预测 多元线性回归 视情维修 

分 类 号：V215.5[航空宇航科学与技术—航空宇航推进理论与工程]