Abstract: In order to achieve the direct thrust control of aero engines, the thrust estimator with high precision and high reliability is needed.A design scheme for thrust estimator was proposed based on the kernel method, which combined global kernel k-means clustering and robust least squares support vector regression. The original input data was mapped to the feature space by the latent mapping of the kernel. And separability of the data is much better, because features of the samples are extracted and amplified. Then a sub-estimator was designed in each cluster using robust least squares support vector regression in which the least squares cost function was replaced by the robust cost function. So the robustness of the whole thrust estimator was enhanced. Finally, the simulation experiments on a certain type of turbofan engine show that the thrust estimator can satisfy the requirement of direct thrust control for aero engines.Compared with other scheme for thrust estimator, the proposed scheme has advantages on estimation accuracy and robustness. 

Key words: Aero engine; Thrust estimator;Kernel method; Clustering; Robust least squares support vector regression

摘要： 鉴于实现航空发动机的直接推力控制需要高精度及高可靠性的推力估计器,基于核方法,提出了结合全局核k-means聚类与鲁棒最小二乘支持向量回归机的推力估计器设计方案,通过核诱导的隐性映射将原始输入数据映射到特征空间,使数据样本特征信息被提取并放大,具有更好的可分性。在每个聚类内设计推力子估计器,用鲁棒代价函数代替最小二乘代价函数,增强了推力估计器的整体鲁棒性。通过对涡扇发动机的仿真试验表明,本推力估计器设计方法能够满足直接推力控制需要,与其它方法相比,在估计精度及鲁棒性上存在一定优势。 

关键词: 航空发动机;推力估计器;核方法;聚类;鲁棒最小二乘支持向量回归机 