考虑起飞工况的航空发动机性能退化预测研究

Prediction of aero engine performance degradation considering takeoff condition

关键词:  航空发动机  性能退化  工况修正  非线性维纳  局部线性嵌入算法

Key words:  Aero-engine  Performance degradation  Working condition correction  Nonlinear Wiener  Locally linear embedding algorithm

摘要:

针对现阶段航空发动机性能退化建模研究没有考虑起飞工况的影响问题，提出了基于修正的非线性维纳过程发动机性能退化建模方法。该方法结合了同类型号发动机的历史性能退化数据与个体发动机的实时退化和工况数据。首先，考虑发动机每次起飞的工况不同，把工况修正引入非线性维纳过程建立发动机的性能退化模型。然后利用极大似然估计方法求得退化模型离线估计值，基于贝叶斯理论对退化参数进行在线更新，最后基于局部线性嵌入算法，对工况参数进行融合构建工况因子，修正退化参数，实现了基于起飞工况的单台发动机性能退化预测。结果表明，采用融合工况因子修正模型，与未修正和压比修正模型相比，平均绝对百分比误差分别降低1.50%和1.01%。证明融合工况因子修正模型能降低发动机起飞工况差异和仅用单工况参数修正所造成的预测误差，可以用来辅助指导下发决策。

Abstract:

With regard to the problem that the current study of aero-engine performance degradation modeling does not take into account the takeoff condition influence， a modified nonlinear Wiener process for engine performance degradation modeling was proposed. This method combined the historical performance degradation data of the same engine type with the real-time degradation and operational condition data of individual engines. Firstly， regarding the different take-off condition for each takeoff， a nonlinear Wiener process with takeoff condition modification was used to establish the engine performance degradation model. Then， the off-line estimation of the degradation model was obtained by utilizing the maximum likelihood estimation. The degradation parameters were updated online based on Bayesian theory. Lastly， based on the locally linear embedding algorithm， the fusion of operating conditions parameters was used to construct the fused operating condition factor， which was used to modify the degradation parameters， and the individual engine performance degradation prediction under different takeoff conditions was achieved. The results show that the mean absolute percentage errors of the model corrected with fused condition factor are reduced by 1.50% and 1.01% respectively compared with the uncorrected model and the model corrected only with engine pressure ratio. It proves that the proposed model can reduce the prediction error caused by the variation of engine takeoff conditions， or the model corrected only by a single operational condition parameter， and it can be used to assist in guiding the determination of engine removal.