Using a priori knowledge in fuzzy model identification

2000

In order to solve the problem of model based control arising from the process model has to be obtained by using small amount and different type of available information, a fuzzy modeling framework has been developed for the utilization of a priori knowledge. The proposed modeling approach transforms the different types of information into the structure of the model (fuzzy rule base), constraints defined on the parameters and variables, dynamic local model or data, and steady-state data or model. This modeling step is followed by an optimization procedure based on these transformed information. The paper describes one element of this framework that was developed to use prior knowledge in constrained adaptation of the rule consequences of Takagi-Sugeno fuzzy models. Experimental results have been obtained for a laboratory setup consisting of two cascaded tanks. It has been shown that by using constrained adaptation, good control performance can be achieved for a nonlinear, time-varying process. 1

1998

Fuzzy control systems have been extensively used on industrial applications. In the recent years Fuzzy Modeling had become the center of attention for researchers in the area. The theory has been oriented mainly to static function approximation and classi cation problems but very few publications have addressed the problem of dynamic function approximation. The problem of identi cation using nonlinear models is a problem of dynamic function approximation. This problem involves many subproblems: experiment design, structure selection and validation. This paper discuss these issues.

There are two approaches to extract a linear model from a Takagi-Sugeno fuzzy model for model based control. The first local approach obtains the linear model by interpolating the parameters of the local models in the TS model, while the second one is based on linearization by Taylor expansion. The locally interpreted interpolated model is not identical to the model obtained by the linearization of the fuzzy model. The paper analyzes the origin of this difference with regard to the applied identification method and the application of the resulted model in model predictive control. In order to keep the analysis simple and transparent, a fuzzy model of a Hammerstein system is studied.

Proceedings of the INTCOM’00 Tempus Symposium on Intelligent Systems in Control and Measurement, 2000

The identification of fuzzy c-regression models (FCRM) suffers from several problems characteristic of all calculus-based optimization methods, including good initialization, avoiding local minima and determining the number of clusters. This paper presents a grey-box approach that can solve the above-mentioned problems with the use of prior knowledge based constrained prototypes. The proposed approach has been applied to system identification, where the FCRM is used to initialize a Takagi-Sugeno fuzzy model of a nonlinear dynamic process. An example is shown to illustrate how knowledge about the type of nonlinearity can be incorporated into the FCRM used in the identification procedure.

Control Engineering Practice, 1996

In this article some aspects of fuzzy modeling are discussed in connection with nonlinear system identification and control design. Methods for constructing fuzzy models from process data are reviewed, and attention is paid to the choice of a suitable fuzzy model structure for the identification task. Some approaches to control design based on a fuzzy model are outlined.

Industrial & Engineering Chemistry Research, 2018

In the advanced process control area, model predictive control (MPC) implementations have been successful in many industrial applications. Despite being an optimization-based control technique, sometimes problems occur with the control algorithm when the dynamic model is not adequate. This work compares statistical techniques for model validation to quantify the quality of identified models used in multivariable MPC controllers. Additionally, a fuzzy validation system is proposed, showing the consistency between the model validation and the predictive controller performance. Multivariable identification, model validation, and predictive controller implementation are performed in an industrial-scale pH neutralization pilot plant.

IEEE Transactions on Fuzzy Systems, 2004

This paper addresses the optimization in fuzzy model predictive control. When the prediction model is a nonlinear fuzzy model, nonconvex, time-consuming optimization is necessary, with no guarantee of finding an optimal solution. A possible way around this problem is to linearize the fuzzy model at the current operating point and use linear predictive control (i.e., quadratic programming). For long-range predictive control, however, the influence of the linearization error may significantly deteriorate the performance. In our approach, this is remedied by linearizing the fuzzy model along the predicted input and output trajectories. One can further improve the model prediction by iteratively applying the optimized control sequence to the fuzzy model and linearizing along the so obtained simulated trajectories. Four different methods for the construction of the optimization problem are proposed, making difference between the cases when a single linear model or a set of linear models are used. By choosing an appropriate method, the user can achieve a desired tradeoff between the control performance and the computational load. The proposed techniques have been tested and evaluated using two simulated industrial benchmarks: pH control in a continuous stirred tank reactor and a high-purity distillation column.

2007 IEEE International Fuzzy Systems Conference, 2007

In this paper, a nonlinear fuzzy identification approach based on Genetic Algorithm (GA) and Takagi-Sugeno (TS) fuzzy system is presented for fuzzy modeling of a multi-input, multioutput (MIMO) dynamical system. In this approach, GA is used for tuning the parameters of the membership functions of the antecedent parts of IF-THEN rules and Recursive Least-Squares (RLS) algorithm is employed for parameter estimation of the consequent linear sub-model parts of the TS fuzzy rules. The presented method is implemented on a simulated nonlinear MIMO distillation column. The results show that the presented method gives a more accurate model in comparison with the conventional TS fuzzy identification approach.

1998

Fuzzy control and predictive control techniques are two modern control strategies that have been accepted by the industry to solve complex problems. Everyday the industry demands control strategies that can deliver better performance for several operating points and these requirements have motivated the development of the theory of Nonlinear Model Predictive Control. This type of controllers can be implemented using fuzzy models. The present paper presents 4 algorithms to construct the controllers. The comparison between the algorithms includes complexity, computational load, model representation, quality of the solution. The controllers are compared using a model of a chemical process (Continuous Stirred Tank Reactor).

2008 International Conference on Control, Automation and Systems, 2008

In this paper we present the nonlinear model predictive control based on the Takagi-Sugeno fuzzy model. The paper is divided into two parts. The first part focuses on the fuzzy model-identification, in which we employ the Takagi-Sugeno fuzzy model-a powerful structure for representing nonlinear dynamic systems. The second part emphasizes on the objective function optimization by using the Branch and Bound method (B&B) and Genetic Algorithm (GAs). Two methods were used as constrained optimizers to online plan optimal input policies over a defined prediction horizon basing on the identified fuzzy model. To reduce computational effort, we combine B&B with Dynamic Grid Size method and GAs with Fuzzy Adaptive Interval. Both developed methods are programmed and tested to control the liquid level of two tanks system which has hard nonlinearity and long delay time. Simulation results show that the proposed methods are successfully applied to nonlinear systems. Some comparisons about "optima" solutions and time executions are discussed.