Abstract: It is difficult to improve the prediction accuracy of the engine Exhaust Gas Temperature（EGT） effectively, aiming at this problem, a regularization chaotic prediction algorithm based on the nonlinear noise-reduction is proposed. To eliminate the impacts of other factors and outliers, data pretreatment for the actual measurement data is studied. As the prediction accuracy of EGT is affected by the noise in the measurement data and chaotic phase space reconstruction parameters, the method of using search algorithm is proposed to find the nonlinear noise-reduction parameters and the delay time and embedding dimensional of phase space which obtain better prediction accuracy of EGT. The simulation results show that the proposed prediction method is applied to predict the EGT of aeroengine, and the test results are satisfactory.

Key words: Exhaust Gas Temperature（EGT）, eliminate gross errors, determination of the noise level, simple nonlinear noise-reduction method, regularization, principal component regression estimation

摘要： 针对传统方法难以进一步提高发动机排气温度（EGT）预测精度的问题，提出一种基于非线性去除噪声的正则化混沌预测算法。为了消除其他因素和离群点对EGT预测精度的影响，研究了实际测量数据的预处理方法。由于观测数据中的噪声和混沌相空间重构参数对EGT预测精度有较大影响，提出利用搜索算法寻找使EGT预测精度较优的非线性去噪参数和重构相空间的延迟时间和嵌入维变量。仿真结果表明，该算法用于某型航空发动机EGT序列的预测，获得了满意的结果。

关键词: 排气温度, 粗大误差剔除, 噪声级估计, 简单非线性去噪, 正则化, 主成分回归估计