L. Chen

A case study of adaptive nonlinear regulation of fed-batch biological reactors

Fed-batch biological reactors are good candidates for the application of adaptive nonlinear techniques because they involve kinetic parameters that are highly uncertain and slowly time-varying. In this paper, the issue is illustrated and discussed with the aid of an industrial application: the ethanol regulation in yeast production processes.

Hybrid modelling of biotechnological processes using neural networks

The hybrid modelling approach for bioprocesses combines a neural network representation of the reaction rates with a mass-balance description of the reactor. A procedure for the identification of hybrid models is proposed and illustrated with an experimental case-study. The key feature is a state transformation which allows to identify separately the kinetic models of the reaction rates even if they occur simultaneously in the reactor

Structural identifiability of the yield coefficients in bioprocess models when the reaction rates are unknown

The identification problem of bioprocesses in a stirred tank reactor essentially concerns the determination of the yield coefficients and the reaction rates involved in a general state-space model. This paper deals with the analysis of structural identifiability of the yield coefficients issuing from a two-step identification procedure. In fact, the identification of the yield coefficients can be completely decoupled from that of the reaction rates by means of an appropriate transformation of the dynamical model. It is shown that the identifiability properties of these coefficients can be drawn from the structure of the underlying reaction network. Necessary and sufficient identifiability conditions are given in the form of some simple algebraic tests.

Multivariable Linear-quadratic Control of a Cement Mill - An Industrial Application

The paper presents the design procedure of a model-based control algorithm for the regulation of tailings and product flowrates in a cement mill. The control variables are the feeding rate and the classifier speed. Experimental results of a full-scale industrial application are reported and discussed with a view to the control of the fineness of the cement.

An industrial application of multivariable linear quadratic control to a cement mill circuit

There are a variety of control strategies that one may consider for cement mill circuit control and often several methods for implementing each strategy. Ideally, the control strategy selected and methodology used in its implementation should not only optimize circuit performance in terms of productivity but also be able to optimize quality. A linear quadratic multivariable controller has been applied to milling circuits and these goals successfully accomplished. The results are a stable circuit with mill and separator optimized, producing a consistent quality product of a predictable fineness. This enables one to use cement quality as the single circuit control set point.

Can We Identify Biotechnological Processes

Abstract The identification problem of biotechnological processes is threefold : (i) determination of the number of biological reactions, (ii) identification of the underlying reaction network, (iii) identification of the kinetics. In most practical cases, these three parts of the identification problem can be completely decoupled from one another.

Identifiability of Scalar Linearly Parametrized Polynomial Systems

Abstract We present necessary and sufficient conditions for the identification of a vector θ of unknown parameters in scalar dynamical systems of the form x = f (x, θ) + g(x, u), where f(x, θ) is a polynomial function of the state x that is affine in θ, and g(x, u) is a polynomial in the state x and the input u. It is assumed that u and x are measured.

Adaptive nonlinear regulation of fed-batch biological reactors: an industrial application

The authors present a general methodology for the design of adaptive regulators for fed-batch biological reactors with the aid of an illustrative case study which has given rise to a genuine industrial application. The application under consideration is the feedback optimization of yeast fermentations, which are of particular interest because yeasts are among the most frequently used microorganisms in genetic engineering for the production of high added value metabolites. The case study is representative of a wide class of control problems in biotechnology where the regulation of some substrate or metabolite concentration helps to solve yield/productivity conflicts.<<ETX>>

An industrial application of mutlivariable linear quadratic control to a cement mill

The paper presents the design development of a model-based control algorithm for the regulation of tailings and product flow rates in a cement mill. The control variables are the feed rate and the classifier speed. Experimental results of a full-scale industrial application are reported and discussed with a view to the control of the fineness of the cement.

Model Based Control of a Reactive Distillation Column

The design and the performances of hybrid model based control of an industrial reactive distillation column are presented. The model structure is a combination of first principles with standard black-box techniques. The approach is based on easily established mass balance equations, the stoichiometry of the system and a model reduction from fast system dynamics. This model makes use of a combined physically meaningful variable from several physical inputs of the process. It is also a key variable for the optimization and the control of the underlying process. This model structure has been successfully used in an IMC scheme for the on-line control of this process.