Identification for control of multivariable systems: Controller validation and experiment design via LMIs

This paper presents a new controller validation method for linear multivariable time-invariant models. Classical prediction error system identification methods deliver uncertainty regions which are nonstandard in the robust control literature. Our controller validation criterion computes an upper bound for the worst case performance, measured in terms of the -norm of a weighted closed loop transfer matrix, achieved by a given controller over all plants in such uncertainty sets. This upper bound on the worst case performance is computed via an LMI-based optimization problem and is deduced via the separation of graph framework. Our main technical contribution is to derive, within that framework, a very general parametrization for the set of multipliers corresponding to the nonstandard uncertainty regions resulting from PE identification of MIMO systems. The proposed approach also allows for iterative experiment design. The results of this paper are asymptotic in the data length and it is assumed that the model structure is flexible enough to capture the true system.     

Optimal experiment design for linear systems with input-output constraints

The problem of optimal experiment design for parameter estimation in linear dynamic systems is studied. Results relating to both constrained input and output variances are established. For the case of constrained input variance, it is shown that a D-optimal experiment exists in which the system input is generated externally provided the system and noise transfer functions have no common parameters. For the case of constrained output variance, it is shown that an experiment in which the system input is generated by a combination of a minimum variance control law together with an external set point perturbation is D-optimal for certain classes of systems. Other related results are also presented which illustrate the role of feedback in optimal experiment design.     

On the equivalence of least costly and traditional experiment design for control

In this paper we establish the equivalence between least costly and traditional experiment design for control. We consider experiment design problems for both open and closed loop systems. In open loop, equivalence is established for three specific cases, relating to different parametrisations of the covariance expression (i.e. finite and high order approximations) and model structure (i.e. dependent and independently parameterised plant and noise models). In the closed loop setting, we consider only finite order covariance expressions. H_∞ performance specifications for control are used to determine the bounds on the covariance expression for both the open and closed loop cases.     

Least costly identification experiment for control

All approaches to optimal experiment design for control have so far focused on deriving an input signal (or input signal spectrum) that minimizes some control-oriented measure of plant/model mismatch between the nominal closed-loop system and the actual closed-loop system, typically under a constraint on the total input power. In practical terms, this amounts to finding the (constrained) input signal that minimizes a measure of a control-oriented model uncertainty set. Here we address the experiment design problem from a “dual” point of view and in a closed-loop setting: given a maximum allowable control-oriented model uncertainty measure compatible with our robust control specifications, what is the cheapest identification experiment that will give us an uncertainty set that is within the required bounds? The identification cost can be measured by either the experiment time, the performance degradation during experimentation due to the added excitation signal, or a combination of both. Our results are presented for the situation where the control objective is disturbance rejection only.     

From experiment design to closed-loop control

The links between identification and control are examined. The main trends in this research area are summarized, with particular focus on the design of low complexity controllers from a statistical perspective. It is argued that a guiding principle should be to model as well as possible before any model or controller simplifications are made as this ensures the best statistical accuracy. This does not necessarily mean that a full-order model always is necessary as well designed experiments allow for restricted complexity models to be near-optimal. Experiment design can therefore be seen as the key to successful applications. For this reason, particular attention is given to the interaction between experimental constraints and performance specifications.     

Closed loop identification and control loop reconfiguration: an industrial case study

The results of a joint university–industry collaborative project for control loop reconfiguration using closed loop experimental data from a fuel gas pressure control system are described in this paper. The fuel gas pressure was being regulated using a butane stream. For economic reasons, it was necessary to switch control to the ethane stream. Previous attempts at effecting this changeover had proved unsuccessful. In this study, a powerful system identification technique namely Canonical Variate Analysis (CVA) was employed to obtain the empirical plant models. A PI controller was then designed using the direct synthesis method. Acceptable closed loop behavior was obtained with little online tuning.     

An effective direct closed loop identification method for linear multivariable systems with colored noise

Multivariable control systems with colored noise widely exist in the most industrial fields, while the system identification under the closed loop conditions is needed in many cases. In view of the above two situations, it needs to find a convenient and effective method to solve the problem. Firstly, the design of the external input signals ensures the identifiability of closed loop system. Secondly, to make the direct method feasible for closed loop identification, the noise model selected is reasonably flexible and independently parameterized. On this basis, this paper proposes an improved method combining the direct closed loop identification approach with the iterative least squares parameter estimation algorithm, which can be an practical solution to the closed loop identification of multivariable systems with colored noise. The presented algorithm based hierarchical identification principle has a strong anti-jamming capability to effectively deal with colored noise existed in the system. Finally, the illustrative examples are given to demonstrate the effectiveness and accuracy of the proposed algorithm.     

Identifiability of MIMO linear dynamic systems operating in closed loop

Identification of multi-input multi-output (MIMO) linear dynamic systems is considered for the case when the measurements are obtained during closed loop operation. Both noise free and noisy feedback situations are analysed. For the case where the disturbances in the feedback path are a full rank stochastic process it is shown that, under certain mild conditions, physically meaningful models for the forward and reverse paths can be uniquely determined. For the case where the feedback path is noise free it is shown that the forward path model can be uniquely determined provided the regulator satisfies certain minimal complexity requirements.     

Model reduction via time-interval balanced stochastic truncation for linear time invariant systems

In this article, a new method for model reduction of linear dynamical systems is presented. The proposed technique is from the family of gramian-based relative error model reduction methods. The method uses time-interval gramians in the reduction procedure rather than ordinary gramians and in such a way it improves the accuracy of the approximation within the time interval which is applied. It is proven that the reduced order model is stable when the proposed method applies to a stable system. The method uses a recently proposed inner–outer factorisation algorithm which enhances the numerical accuracy and efficiency. In order to avoid numerical instability and also to further increase the numerical efficiency, projector matrices are constructed instead of the similarity transform approach for reduction. The method is illustrated by a numerical example and finally it is applied to a practical CD player example. The numerical results show that the method is more accurate than ordinary balanced stochastic truncation.     

Designing parametric linear quadratic regulators for parametric LTI systems via LMIs

This paper addresses the problem of determining parametric linear quadratic regulators (LQRs) for continuous-time linear-time invariant systems affected by parameters through rational functions. Three situations are considered, where the sought controller has to minimise the best cost, average cost, and worst cost, respectively, over the set of admissible parameters. It is shown that candidates for such controllers can be obtained by solving convex optimisation problems with linear matrix inequality (LMI) constraints. These candidates are guaranteed to approximate arbitrarily well the sought controllers by sufficiently increasing the size of the LMIs. In particular, the candidate that minimises the average cost approximates arbitrarily well the true LQR over the set of admissible parameters. Moreover, conditions for establishing the optimality of the found candidates are provided. Some numerical examples illustrate the proposed methodology.     

A virtual closed loop method for closed loop identification

Indirect methods for the identification of linear plant models on the basis of closed loop data are based on the use of (reconstructed) input signals that are uncorrelated with the noise. This generally requires exact (linear) controller knowledge. On the other hand, direct identification requires exact plant and noise modelling (system in the model set) in order to achieve accurate results, although the controller can be non-linear. In this paper, a generalized approach to closed loop identification is presented that includes both methods as special cases and which allows novel combined methods to be generated. Besides providing robustness with respect to inexact controller knowledge, the method does not rely on linearity of the controller nor on exact noise modelling. The generalization is obtained by balancing input-noise decorrelation against noise whitening in a user-chosen flexible fashion. To this end, a user-chosen virtual controller is used to parametrize the plant model, thereby generalizing the dual-Youla method to cases where knowledge of the controller is inexact. Asymptotic bias and variance results are presented for the method. Also, the benefits of the approach are demonstrated via simulation studies.     

Minimum variance lower bound ratio based nonlinearity measure for closed loop systems

Uninterrupted wide range operations are salient features of the process industries. Characteristics including nonlinearity, time-delay, and inertia, of industrial control loops are always significantly changed with the working conditions. To qualify the nonlinear degree, a nonlinearity measure based on the minimum variance lower bound ratio is presented in this paper. This measure belongs to the data-driven class, and can be applied to Hammerstein structure, Wiener structure and Wiener–Hammerstein structure, whether the controller is linear or nonlinear. The effectiveness and consistency of this measure are illustrated through five simulation tests. An industrial case in the fossil fuel power generation process is studied to confirm the practicability of the proposed measure.     

Input design as a tool to improve the convergence of PEM

The Prediction Error Method (PEM) is related to an optimization problem built on input/output data collected from the system to be identified. It is often hard to find the global solution of this optimization problem because the corresponding objective function presents local minima and/or the search space is constrained to a nonconvex set. The shape of the cost function, and hence the difficulty in solving the optimization problem, depends directly on the experimental conditions, more specifically on the spectrum of the input/output data collected from the system. Therefore, it seems plausible to improve the convergence to the global minimum by properly choosing the spectrum of the input; in this paper, we address this problem. We present a condition for convergence to the global minimum of the cost function and propose its inclusion in the input design. We present the application of the proposed approach to case studies where the algorithms tend to get trapped in nonglobal minima.     

Modal control of linear time invariant systems

A procedure is presented for the determination of a linear control law (for a system with one input) such that the resultant closed–loop System has specified eigenvalues. The design procedure is simpler than existing techniques.     

Estimators for linear dynamical SISO systems with disturbed input-output measurements

Several types of instrumental variable estimators are proposed for estimating parameters of linear dynamical single-input single-ouput systems with disturbed input-output measurements.No restrictions on the colouring of the distributions are imposed. The availability of extra (disturbed) measurements of input or output data simpifies the choice of the instrumental variable.     

Input design for structured nonlinear system identification

This paper is concerned with the input design problem for a class of structured nonlinear models. This class contains models described by an interconnection of known linear dynamic systems and unknown static nonlinearities. Many widely used model structures are included in this class. The model class considered naturally accommodates a priori knowledge in terms of signal interconnections. Under certain structural conditions, the identification problem for this model class reduces to standard least squares. We treat the input design problem in this situation.An expression for the expected estimate variance is derived. A method for synthesizing an informative input sequence that minimizes an upper bound on this variance is developed. This reduces to a convex optimization problem. Features of the solution include parameterization of the expected estimate variance by the input distribution, and a graph-based method for input generation.     

Optimal-observer design for linear dynamical systems with uncertain parameters

This paper presents a design technique for the synthesis of robust observers for linear dynamical systems with uncertain parameters. The perturbations under consideration are modelled as unknown but bounded disturbances and an ellipsoidal set-theoretic approach is used to formulate the optimal-observer design problem. The optimal criterion introduced here is the minimization of the ‘size’ of the bounding ellipsoid of the estimation error. A necessary and sufficient condition for this optimal design problem is presented. The results are stated in terms of a reduced-order observer with constant gain matrix, which is then determined by solving a matrix Riccati-type equation. Furthermore, a gradient-search algorithm is presented to find the optimal solution when the free parameter that enters in the construction of the bounding ellipsoids of the estimation error is considered as a design parameter. The effectiveness of the proposed approach is illustrated through a numerical example.     

Constrained piecewise linear systems with disturbances: Controller design via convex invariant sets

This paper presents two control strategies under the time optimal control and model predictive control frameworks for constrained piecewise linear systems with bounded disturbances (PWLBD systems). Each of the proposed approaches uses an inner convex polytopal approximation of the non‐convex domains of attraction and results in simplified control laws that can be determined off‐line via multi‐parametric programming. These control strategies rely on invariant sets of PWLBD systems. Thereby, approaches for the computation of the disturbance invariant outer bounds of the minimal disturbance invariant set, F_∞, and convex polytopal disturbance invariant sets are presented. The effectiveness of the approaches is assessed through numerical examples. Copyright © 2007 John Wiley & Sons, Ltd.     

Unprejudiced optimal open loop input design for identification of transfer functions

The problem to estimate transfer functions of linear systems is considered. The quality of the resulting estimate depends, among other things, on the input used during the identification experiment. We measure the quality using a quadratic norm in the frequency domain. The problem to determine optimal inputs, i.e. inputs that minimize the chosen norm, subject to constrained input variance, has long been studied. We point out that such procedures may involve a prejudice (that the system is to be found in a certain model set) that may have some surprising effects. We discuss how such a prejudice can be reduced by allowing the possibility that the true system cannot be exactly described in the chosen model set. We also calculate explicit expressions for the resulting “unprejudiced” optimal inputs. These expressions relate the signal-to-noise ratio (as a function of frequency) to the chosen weighting function in the quadratic norm. We also point out the role of the employed noise model for the design.     

Structural identification and software package for linear multivariable systems

This paper is concerned with the structural identification of linear multivariable systems and an interactive identification package. The structural identification is done by taking the time-invariant subsystem from the realizations of the input-output relations identified using data of disjoint time intervals, and the statistical hypothesis test is employed to determine the order, where the input-output relation is identified based on the generalized least squares method using the possibly larger model for the plant. The identification package is for the identification of the input-output relation of a linear multivariable system, for the structural identification based on the realization and for data management.