用改进子空间辨识法建立航空发动机模型

Establishment of Aeroengine Model with Improved Subspace Model Identification Method

研究一种采用改进子空间辨识法建立用于航空发动机故障诊断与控制系统设计的小偏差状态变量模型方法。首先,利用发动机非线性模型的输入输出数据序列,在离散域下基于子空间辨识法建立指定阶数、无噪声干扰的状态变量模型,然后将其转化到连续域下进行相似变换,从而获得具有明确物理意义的发动机状态变量模型。这样,不仅避免了那些基于最优化思想方法所带来的一系列问题,即非线性迭代优化、对初始值敏感、计算时间长、系统矩阵参数规律性差等,而且不受模型阶次影响,并具有实现简单等特点。应用于建立某型涡扇发动机的小偏差状态变量模型,并与改进拟合法在拟合精度、计算时间、参数变化三个方面进行比较,从而验证了改进子空间辨识方法的优点与有效性。

An improved subspace model identification method was proposed to establish small perturbation state variable model（SVM） for aeroengine fault diagnose and control system design. According to the input -output data sequences from aeroengine nonlinear model, the SVM with n - order and no noise was established via subspace model identification method in the discrete domain. Then it was transformed to the continue domain and finally the physical - based aeroengine SVM can be obtained via similarity transformation. This method not only avoids a series of issues due to nonlinear optimization techniques, such as nonlinear iteration, sensitive to the initial conditions and long com- puting time, but also takes no influence on the order of model and can be realized simply. One small perturbation SVM of a certain turbofan engine has been established by this method, and compared with improved fitting method in fitting precision, computing time and parameter variety, proving the merits and validity of this method.