基于MC-XGBoost模型的航空发动机振动特性预测  

Prediction of aeroengine vibration characteristics based on an MC-XGBoost model

作　　者：梅潇[1] 池华山 岳聪 范建瑜 刘宗沁 MEI Xiao;CHI Huashan;YUE Cong;FAN Jianyu;LIU Zongqin(School of Logistics Engineering,Shanghai Maritime University,Shanghai 201306,China)

机构地区：[1]上海海事大学物流工程学院,上海201306

出　　处：《振动与冲击》2022年第16期271-277,共7页Journal of Vibration and Shock

基　　金：上海市科技人才计划项目(19YF1452400);上海科委创新行动计划重大专项(18DZ1100802)。

摘　　要：为探索航空发动机高压转子组件的装配参数对整机振动响应的影响,通过不平衡响应振动方程描述了装配阶段几何偏差和不平衡偏差与振型的机理关系,提出一种基于最大相关性(maximum correlation,MC)系数与极端梯度提升(XGBoost)的改进模型,MC-XGBoost预测模型。通过MC系数筛选影响振动性能的关键参数,即最相关的几何/不平衡装配参数;代入XGBoost模型进行振动特性预测。利用试验数据对预测模型进行验证,结果表明所提出的MC-XGBoost预测模型相比于RF、GBDT算法模型,具有更高的预测精度,能够为航空发动机面向装配质量的振动特性评估提供一种有效的评价方法。In order to explore the influence of assembly parameters of an aero-engine high-pressure rotor assembly on the vibration response of the whole machine,the mechanism relationship between geometric deviation,unbalance deviation and vibration mode in assembly stage was described by unbalanced response vibration equation,and an improved model,MC-XGBoost prediction model,based on maximum correlation(MC)coefficient and extreme gradient lifting(XGBoost)was proposed.The key parameters that affect the vibration performance,namely the most relevant geometric/unbalanced assembly parameters,were screened by the MC coefficient,and then the selected assembly parameters were brought into the XGBoost model to predict the vibration characteristics.The prediction model was verified by experimental data,and the results show that the proposed MC-XGBoost prediction model has higher prediction accuracy than the RF and the GBDT algorithm models,and can provide an effective evaluation method for aero-engine assembly quality-oriented vibration characteristic evaluation.

关 键 词：高压转子 装配质量 振动特性 最大相关性(MC)系数 XGBoost模型 

分 类 号：V264.2[航空宇航科学与技术—航空宇航制造工程]