基于降维和随机森林的航空发动机涡轮盘应力预测

Stress Prediction of Aero-Engine Turbine Disk based on Dimension Reduction and Random Forest

关键词:  涡轮盘  降维  主成分分析  随机森林  寿命管理

Key words:  Turbine disk  Dimension reduction  Principal component analysis  Random forest  Life management

摘要:

为了给发动机涡轮盘寿命管理提供有效的数据输入及后续工程应用提供依据，本文基于统计学习和机器学习方法，提出基于降维和随机森林的航空发动机涡轮盘应力预测模型，以发动机可测参数作为初始特征，通过相关性分析、主成分分析与聚类分析，实现了对总体参数样本的降维，并提取出主控因素，再利用随机森林算法建立航空发动机涡轮盘应力预测模型。结果表明：该方法预测精度比未降维的随机森林模型更高，判定系数R²达到0.985以上，证明该方法对航空发动机涡轮盘应力预测是有效的。

Abstract:

Based on statistical learning and machine learning methods， an aero-engine turbine disk stress prediction model based on dimension reduction and random forest was proposed to provide effective data input for engine turbine disk life management and provide a basis for subsequent engineering applications. Taking the measurable engine parameters as the initial characteristics， the dimensionality of the overall parameter samples was reduced through correlation analysis， principle component analysis and cluster analysis， and the main control factors were extracted. The random forest algorithm was used to establish the aero-engine turbine disk stress prediction model. The results show that the prediction accuracy of this method is higher than that of the random forest model without dimensional-reduction， and the determination coefficient is above 0.985， which proves that this method is effective for the prediction of turbine disk stress of aero-engine， and has great significance for the technical support of aero-engine life management.