Study A Subspace Identification Method Based On The Orthogonal Decomposition And PCA And Its Application

In recent20years, subspace identification method has been paid great attentionas a new identification approach. Compared with traditional identification methods,subspace identification method has many advantages and is more suitable formultivariable systems. Classic subspace identification methods are based on theopen-loop system. Subsequently, some subspace identification methods forclosed-loop and nonlinear systems have also been proposed. However, On the basis oftraditional subspace identification algorithms, this paper researches a new approachwhich is based on orthogonal decomposition and principal component analysis. At last,a case simulation research based kiln is carried out.The main contents of this paper are as follows.At the beginning, this paper presents the research background and current statusof this subject, and then elaborates some basic theory and methods to be used, such assystems theory, linear algebra, numerical calculation and other related knowledge.Besides, the implementation of the proposed subspace identification algorithm ispresented and several kinds of classical methods are also described.The classical subspace identification algorithm is only suitable for open-loopsystem, but the rotary kiln production process is a closed-loop system. For this case,this paper researches a subspace identification algorithm based on orthogonaldecomposition and PCA. The proposed method decomposes the data to determiningpart and random part by the way of orthogonal decomposition, and then identifies thedetermining part to obtain the system matrix. What’s more, the AIC(Akaikeinformation criterion) criterion is introduced for the identification of the system’sorder. Simulation results show the great performance of the AIC criterion on system’sorder identification and the effectiveness of the proposed method on closed-loopidentification systems.Lastly, the model of rotary kiln is established using the proposed subspaceidentification algorithm. This paper also describes the way of selecting sample dataand illustrates a series of operations for the obtained data including denoising,normalization, smoothing and so on. After several simulations to the processed rotarykiln data, we obtain a more accurate rotary kiln model whose input variable is the hostcooler currents and the output value is temperature from the kiln header and trailer.Simulation results have shown the effectiveness of the proposed method in the model of rotary kiln temperature.