Abstract: In order to study the problem of the lack of gas path fault samples for a neo-aero-engine，engine nonlinear model and gas path components fault simulation method are utilized to gain fault samples. Then，least squares support vector regression(LSSVR)machine is used to establish the gas path health parameter estimator. Based on LSSVR estimator，the main work of this paper is to design a correction system based on inverse model track control，which is composed of inverse model online identification and variable structure controller. The purpose of correction system is to eliminate the effects of measurement noise and systematic noise on estimation accuracy between ideal estimator and real engine. Through experiment，with offline bias correction，the maximum of relative error reduced from 69% to 5%，with online bias correction，the maximum of relative error reduced from 69% to 13%. Estimation accuracy is improved by correction system effectively，and the probability of misjudgment and the misdiagnosis is reduced.

Key words: Gas path component health parameter estimation；Inverse track control；Online identification；Sliding mode variable structure control；Correction

摘要： 针对新型航空发动机气路部件状态监控缺乏故障样本的不足，利用发动机非线性模型和气路故障模拟方法，获取故障样本，运用最小二乘支持向量回归机(LSSVR)建立气路健康参数的(理想)估计器。主要工作是在LSSVR的基础上，从逆跟踪控制的角度，设计了基于逆模型在线辨识与变结构控制相结合的修正系统，减小理想估计器与真实发动机之间由于测量噪声与系统噪声对估计精度造成的影响。通过仿真实验，离线修正后，估计值的相对误差最大值从69%缩小5%，在线修正后，估计值的相对误差最大值从69%缩小13%，修正系统有效地提高了性能估计的准确度，减小了气路部件状态监控的误判和误诊的概率。

关键词: 气路健康参数估计；逆跟踪控制；在线辨识；滑模变结构控制；修正