Subspace model identification Part 1. The output-error state-space model identification class of algorithms

In this paper, we present two novel algorithms to realize a finite dimensional, linear time-invariant state-space model from input-output data. The algorithms have a number of common features. They are classified as one of the subspace model identification schemes, in that a major part of the identification problem consists of calculating specially structured subspaces of spaces defined by the input-output data. This structure is then exploited in the calculation of a realization. Another common feature is their algorithmic organization: an RQ factorization followed by a singular value decomposition and the solution of an overdetermined set (or sets) of equations. The schemes assume that the underlying system has an output-error structure and that a measurable input sequence is available. The latter characteristic indicates that both schemes are versions of the MIMO Output-Error State Space model identification (MOESP) approach. The first algorithm is denoted in particular as the (elementary MOESP scheme). The subspace approximation step requires, in addition to input-output data, knowledge of a restricted set of Markov parameters. The second algorithm, referred to as the (ordinary MOESP scheme), solely relies on input-output data. A compact implementation is presented of both schemes. Although we restrict our presentation here to error-free input-output data, a framework is set up in an identification context. The identification aspects of the presented realization schemes are treated in the forthcoming Parts 2 and 3.     

Convergence analysis of instrumental variable recursive subspace identification algorithms

The convergence properties of recently developed recursive subspace identification methods are investigated in this paper. The algorithms operate on the basis of instrumental variable (IV) versions of the propagator method for signal subspace estimation. It is proved that, under suitable conditions on the input signal and the system, the considered recursive subspace identification algorithms converge to a consistent estimate of the propagator and, by extension, to the state-space system matrices.     

Identification of the Tennessee Eastman challenge process with subspace methods

The Tennessee Eastman challenge process is a realistic simulation of a chemical process that has been widely used in process control studies. In this case study, several identification methods are examined and used to develop MIMO models that contain seven inputs and ten outputs. ARX and finite impulse response models are identified using reduced-rank regression techniques (PLS and CCR) and state-space models identified with prediction error methods and subspace algorithms. For a variety of reasons, the only successful models are the state-space models produced by two popular subspace algorithms, N4SID and canonical variate analysis (CVA). The CVA model is the most accurate. Important issues for identifying the Tennessee Eastman challenge process and comparisons between the subspace algorithms are also discussed.     

Identification of MIMO time-varying stochastic systems with unknown dead times

This paper presents a new on-line procedure for identifying MIMO time-varying stochastic dynamic systems. It is based on a decomposition model derived from the state-space representation. This scheme gives an estimated model with a minimal number of parameters. The parameters of the MIMO system are identified from short records with unknown dead times. Extended recursive least-squares and fixed-lag smoothing identification algorithms are applied to identify the validity of the model structure. Artificial and hydrological records are used to illustrate the efficiency of the proposed procedure.     

Subspace state space system identification for industrial processes

We give a general overview of the state-of-the-art in subspace system identification methods. We have restricted ourselves to the most important ideas and developments since the methods appeared in the late eighties. First, the basics of linear subspace identification are summarized. Different algorithms one finds in literature (such as N4SID, IV-4SID, MOESP, CVA) are discussed and put into a unifying framework. Further, a comparison between subspace identification and prediction error methods is made on the basis of computational complexity and precision of the methods by applying them on 10 industrial data sets.     

Subspace identification of Hammerstein systems using least squares support vector machines

This paper presents a method for the identification of multiple-input-multiple-output (MIMO) Hammerstein systems for the goal of prediction. The method extends the numerical algorithms for subspace state space system identification (N4SID), mainly by rewriting the oblique projection in the N4SID algorithm as a set of componentwise least squares support vector machines (LS-SVMs) regression problems. The linear model and static nonlinearities follow from a low-rank approximation of a matrix obtained from this regression problem.     

状态空间模型下的Hammerstein系统的递推子空间辨识方法

一般来说, 单输入单输出情形下的Hammerstein系统可以由转移函数来表示, 而对于多输入多输出情形下的系统其输入与输出间的关系难以表示. 本文基于Hammerstein系统的状态空间模型, 研究了其子空间辨识方法. 在开环情形, 对Hammerstein系统给出了在其非线性函数可以由有限基函数线性表示的情形的子空间辨识方法, 及其递推实现. 并且初步分析了这些方法的渐近性质. 针对这一方法, 我们给出了一个数据模拟实例分析方法的优劣.     

Identification of quadratic systems using higher order cumulants and neural networks: Application to model the delay of video-packets transmission

This work concerns the development of two approaches for the identification of diagonal parameters of quadratic systems from only the output observation. The systems considered are excited by an unobservable independent identically distributed (i.i.d), stationary zero mean, non-Gaussian process and corrupted by an additive Gaussian noise. The proposed approaches exploit higher order cumulants (HOC) (fourth order cumulants) and are the extension of the algorithms developed in the linear version 1D, which uses a non-Gaussian signal input. For test and validity purpose, these approaches are compared to recursive least square (RLS), least mean square (LMS) and neural network identification algorithms using non-linear model in noisy environment. To demonstrate the applicability of the theoretical methods on real processes, we applied the developed approaches to search for models able to describe the delay of the video-packets transmission over IP networks from video server. The simulation results show the correctness and the efficiency of the developed approaches.     

Least squares based iterative identification for a class of multirate systems

This paper studies modeling and identification problems for multi-input multirate systems with colored noises. The state-space models are derived for the systems with different input updating periods and furthermore the corresponding transfer functions are obtained. To solve the difficulty of identification models with unmeasurable noises terms, the least squares based iterative algorithm is presented by replacing the unmeasurable variables with their iterative estimates. Finally, the simulation results indicate that the proposed iterative algorithm has advantages over the recursive algorithms.     

工业过程的子空间模型辨识

子空间模型辨识方法(SMI)是一类新兴的直接估计线性状态空间模型的黑箱建模方法,近年来获得了广泛关注.和传统的线性建模方法相比,SMI的优势不仅在于算法本身的简单可靠,也在于它的状态空间表达.本文首先简要介绍了SMI的基本思想以及3种基本算法(N4SID,MOESP,CVA).然后将这类方法应用于一个实际的工业过程建模,同时对3种SMI基本算法和一种传统辨识算法—预测误差方法(PEM)进行了研究对比.     

Subspace model identification Part 2. Analysis of the elementary output-error state-space model identification algorithm

The elementary MOESP algorithm presented in the first part of this series of papers is analysed in this paper. This is done in three different ways. First, we study the asymptotic properties of the estimated state-space model when only considering zero-mean white noise perturbations on the output sequence. It is shown that, in this case, the MOESPl implementation yields asymptotically unbiased estimates. An important constraint to this result is that the underlying system must have a finite impulse response and subsequently the size of the Hankel matrices, constructed from the input and output data at the beginning of the computations, depends on the number of non-zero Markov parameters. This analysis, however, leads to a second implementation of the elementary MOESP scheme, namely MOESP2. The latter implementation has the same asymptotic properties without the finite impulse response constraint. Secondly, we compare the MOESP2 algorithm with a classical state space model identification scheme. The latter scheme, referred to as the CLASSIC algorithm, is based on the Ho and Kalman realization scheme and estimated Markov parameters. The comparison is done by a sensitivity study, where the effect is studied of the errors on the data on the calculated column space of the shift-invariant subspace. This study demonstrates that the elementary MOESP2 scheme is more robust with respect to the errors considered than the CLASSIC algorithm. In the third part, the model reduction capabilities of the elementary MOESP schemes are analysed when the observations are error-free. We demonstrate in which sense the reduced order model is optimal when acquired with the MOESP schemes. The optimality is expressed by the difference between the 2-norm of the errors on the state (or output) sequence of the reduced-order model and the 2-norm of the matrix containing the rejected singular values being as small as possible. The insights obtained in these three parts are evaluated in a simulation study, and validated in this paper. They lead to the assertion that the MOESP2 implementation allows identification of a compact, low-dimensional, state-space model accurately describing the input -output behaviour of the system to be identified, while making use of ‘perturbed’ input-output data. This can be done efficiently.     

Identification of step response estimates utilizing continuous time subspace approach

Direct identification procedures using raw data seem to face difficulties especially when the data is corrupted with noise or the data acquisition leads to huge amount of data to be processed. This will lead to complexity in obtaining the accurate model of the system and the increase of computational load and time may also arise. In this paper, we present 2-stage identification, in which, the first stage involves a process to obtain step response estimates. A multi input multi output frequency sampling filter model is used to simulate the estimates. With the aid of finite impulse response model, maximum likelihood method and the predicted sum of square statistics, this procedure able to clean the noise that occurred at high frequency region, compressed the data into the reduced amount and obtained only meaningful parameter that describes the system. Next, at the second stage the continuous time subspace model identification is conducted using the step response estimates obtained from the first stage. Here, three continuous time subspace methods will be observed to develop a state space mathematical model; those are the MOESP, CCA and ORT methods. A Monte Carlo simulation is performed as to see the efficacy and robustness of those models in identifying the step response estimates of the observed system. Comparative analysis with respect to two-stage identification and direct identification procedure is also conducted. This is to show the significant contribution of having MIMO FSF in the overall identification procedure. From results, the developed MIMO FSF is able to compress raw MIMO data into fewer numbers, and produce cleaned and unbiased step response estimates. When it is implemented to MIMO continuous-time subspace identification, MOESP method has demonstrated good performance based on the accuracy and robustness of the developed model.     

N4SID and MOESP algorithms to highlight the ill-conditioning into subspace identification

In this paper,an analysis for ill conditioning problem in subspace identifcation method is provided.The subspace identifcation technique presents a satisfactory robustness in the parameter estimation of process model which performs control.As a frst step,the main geometric and mathematical tools used in subspace identifcation are briefly presented.In the second step,the problem of analyzing ill-conditioning matrices in the subspace identifcation method is considered.To illustrate this situation,a simulation study of an example is introduced to show the ill-conditioning in subspace identifcation.Algorithms numerical subspace state space system identifcation(N4SID)and multivariable output error state space model identifcation(MOESP)are considered to study,the parameters estimation while using the induction motor model,in simulation(Matlab environment).Finally,we show the inadequacy of the oblique projection and validate the efectiveness of the orthogonal projection approach which is needed in ill-conditioning;a real application dealing with induction motor parameters estimation has been experimented.The obtained results proved that the algorithm based on orthogonal projection MOESP,overcomes the situation of ill-conditioning in the Hankel s block,and thereby improving the estimation of parameters.     

一种基于ＳＩＭＯ系统的递推盲辨识算法

传统的盲辨识算潮基于通道输入的统计模型和通道输出,然而在对输入的统计特性进行准确估计时需要大量的数据,为了避免这种缺点,本文提出了一种单输入多输（SIMO)线性时不变有限单位冲击响应（ＦＩＲ）系统的盲辨识最小二乘算法,利用递推最小二乘算法求解这类算法中的ＸＬＴＫ方程,大大降低了算法对计算存储量的要求,在盲辨识的基础上,利用多项式互质的一个判别定理,通过解卷积求出ＳＩＭＯ－ＦＩＲ系统输入,最后通过实验验证了算法的有效性。     

A new subspace identification approach based on principal component analysis

Principal component analysis (PCA) has been widely used for monitoring complex industrial processes with multiple variables and diagnosing process and sensor faults. The objective of this paper is to develop a new subspace identification algorithm that gives consistent model estimates under the errors-in-variables (EIV) situation. In this paper, we propose a new subspace identification approach using principal component analysis. PCA naturally falls into the category of EIV formulation, which resembles total least squares and allows for errors in both process input and output. We propose to use PCA to determine the system observability subspace, the A, B, C, and D matrices and the system order for an EIV formulation. Standard PCA is modified with instrumental variables in order to achieve consistent estimates of the system matrices. The proposed subspace identification method is demonstrated using a simulated process and a real industrial process for model identification and order determination. For comparison the MOESP algorithm and N4SID algorithm are used as benchmarks to demonstrate the advantages of the proposed PCA based subspace model identification (SMI) algorithm.     

基于多维泰勒网的超前d步预测模型

针对单入单出(SISO)与多入多出(MIMO)非线性时滞系统构建预测模型准确性问题,分别提出基于多维泰勒网(MTN)的预测模型构建方案.首先,分别依靠非递推技术与递推技术来设计非递推d步与递推d步超前MTN预测模型,给出二者表达式,二者皆可对未来d步范围进行预测,并有效弥补时滞带来的影响;然后,利用阻尼递推最小二乘(D...     

Subspace identification using instrumental variable techniques

Subspace-based algorithms for system identification have lately been suggested as alternatives to more traditional techniques. Variants of the MOESP type of subspace algorithms are in addition to open-loop identification applicable to closed-loop and errors-in-variables identification. In this paper, a new instrumental variable approach to subspace identification is presented. It is shown how existing MOESP-algorithms can be derived within the proposed framework, simply by changing instruments and weighting matrices. A noteworthy outcome of the analysis is that an improvement of an existing MOESP method for errors-in-variables identification can be proposed.     

Recursive 4SID algorithms using gradient type subspace tracking

Sometimes we obtain some prior information about a system to be identified, e.g., the order, model structure etc. In this paper, we consider the case where the order of a MIMO system to be identified is a priori known. Recursive subspace state-space system identification algorithms presented here are based on the gradient type subspace tracking method used in the array signal processing. The algorithms enable us to estimate directly the subspace spanned by the column vectors of the extended observability matrix of the system to be identified without performing the singular value decomposition. Also, a new convergence proof of the gradient type subspace tracking is given in this paper. Under the condition of a step size between 0 and 1, we prove the convergence property of the recursive equation of the gradient type subspace tracking. A numerical example illustrates that our algorithm is more robust with respect to the choice of the initial values than the corresponding PAST one.     

Block pulse function approach for the identification of Hammerstein model non-linear systems

Two recursive algorithms based on block pulse functions are presented for identifying continuous Hammerstein models of non-linear systems with (i) a state space model and (ii) an input–output model. Since the continuous non-linear systems are transformed approximately into the corresponding difference equations via block pulse functions, these recursive estimation algorithms can easily be obtained using a derivation similar to that of the discrete-time models expressed by difference equations. Both algorithms derived here are simple and straightforward, and can easily be implemented on-line. As discussed in this paper, these algorithms can also be extended to the identification of certain continuous non-linear systems with a feedback loop or with time delays. The illustrative examples show that these recursive algorithms give satisfactory results for the identification problems of certain continuous non-linear systems.