Dale E. Seborg

Nonlinear internal model control strategy for neural network models

Abstract A nonlinear internal model control (NIMC) strategy based on neural network models is proposed for SISO processes. The neural network model is identified from input—output data using a three-layer feedforward network trained with a conjugate gradient algorithm. The NIMC controller consists of a model inverse controller and a robustness filter with a single tuning parameter. The proposed strategy includes time delay compensation in the form of a Smith predictor and ensures offset-free performance. Extensions for measured disturbances are also presented. The NIMC approach is currently restricted to processes with stable inverses. Two alternative implementations of the control law are discussed and simulations results for a continuous stirred tank reactor and pH neutralization process are presented. The results for these two highly-nonlinear processes demonstrate the ability of the new strategy to outperform conventional PID control.

Adaptive nonlinear control of a pH neutralization process

An adaptive nonlinear control strategy for a bench-scale pH neutralization system is developed and experimentally evaluated. The pH process exhibits severe nonlinear and time-varying behavior and therefore cannot be adequately controlled with a conventional PI controller. The nonlinear controller design is based on a modified input-output linearization approach which accounts for the implicit output equation in the reaction invariant model. Because the reaction invariants cannot be measured online and the linearized system is unobservable, a nonlinear output feedback controller is developed by combining the input-output linearizing controller with a reduced-order, open-loop observer. The adaptive nonlinear control strategy is obtained by augmenting the non-adaptive controller with an indirect parameter estimation scheme which accounts for unmeasured buffering changes. Experimental tests demonstrate the superior performance of the adaptive nonlinear controller as compared to a non-adaptive nonlinear controller and conventional PI controller. >

A self-tuning controller with a PID structure

Some self-tuning controllers are proposed which have the same structure as conventional PID controllers. The new self-tuning PID controllers are based on a modified version of the design method of Clarke and Gawthrop (1975, 1979). Experimental and simulation studies for a stirred-tank heating system indicate that the new self-tuners perform well and can be easily adjusted on-line.

Nonlinear control strategies for continuous fermenters

Abstract Nonlinear controllers based on exact linearization are designed and evaluated for continuous fermenters. The dilution rate and feed substrate concentration are considered as manipulated inputs in single-input/single-output strategies for productivity control. The resulting controllers are compared theoretically and via simulation. It is shown that state—space linearization techniques are not appropriate for this class of fermenters. Exact input—output-linearizing control employing the dilution rate as the manipulated input is shown to provide excellent regulatory behavior. Conversely, input—output linearization with the feed substrate concentration as the manipulated input is problematic. The exact approach yields unreasonably large control moves while an approximate technique recently proposed results in very sluggish responses. A modified approach to exact input—output linearization is proposed that results in satisfactory control.

Critique of exact linearization strategies for process control

Abstract During the last decade significant progress has been made in the control of non-linear systems. Some of the most promising controller design techniques are based on exact linearization theory. In this paper, exact linearization strategies for process control are critically evaluated. It is shown that many design techniques recently developed for non-linear process control are based, either implicitly or explicitly, on exact linearization of the input-output map. Extensions of basic techniques that are especially pertinent for process control problems are reviewed and evaluated. A thorough survey of applications to process control problems is included. Finally, an input-output linearizing controller is designed for a different process control problem, pH neutralization, and simulation results are presented. This example demonstrates that input-output linearization can be extended to the class of problems where the output equation is an implicit rather than explicit function of the state variables.

Control-Relevant Models for Glucose Control Using A Priori Patient Characteristics

One of the difficulties in the development of a reliable artificial pancreas for people with type 1 diabetes mellitus (T1DM) is the lack of accurate models of an individuals response to insulin. Most control algorithms proposed to control the glucose level in subjects with T1DM are model-based. Avoiding postprandial hypoglycemia (<;60 mg/dl) while minimizing prandial hyperglycemia (>;180 mg/dl) has shown to be difficult in a closed-loop setting due to the patient-model mismatch. In this paper, control-relevant models are developed for T1DM, as opposed to models that minimize a prediction error. The parameters of these models are chosen conservatively to minimize the likelihood of hypoglycemia events. To limit the conservatism due to large intersubject variability, the models are personalized using a priori patient characteristics. The models are implemented in a zone model predictive control algorithm. The robustness of these controllers is evaluated in silico, where hypoglycemia is completely avoided even after large meal disturbances. The proposed control approach is simple and the controller can be set up by a physician without the need for control expertise.

Automatic detection of excessively oscillatory feedback control loops

A statistically-based approach is proposed to detect excessively oscillatory feedback control loops. The technique is simple and requires only normal operating data. The effectiveness and widespread applicability of the new approach are demonstrated in several experimental applications, including an industrial distillation column.

An extension of the Smith Predictor method to multivariable linear systems containing time delays

The classical Smith Predictor method for single variable systems is extended to a class of linear multivariable systems. Derivations of the multivariable Smith Predictor are presented for both continuous-time and discrete-time systems which contain time delays in the control variables and/or output variables. As in the classical method, use of the multivariable Smith Predictor eliminates the time delays from the characteristic equation of the closed-loop system

A nonlinear predictive control strategy based on radial basis function models

A predictive control strategy for nonlinear processes based on radial basis function models is proposed. First, a radial basis function model of the process is developed using stepwise regression and least squares estimation. This model is then used to train a nonlinear predictive controller, which is also implemented as a radial basis function network. Since no optimization problems have to be solved on-line, this control strategy can be implemented easily. The proposed strategy is applied to an experimental pH neutralization process; it provides both excellent setpoint tracking and disturbance rejection when compared to conventional PI control.

Identification of non-linear processes using reciprocal multiquadric functions

Abstract In this paper radial basis function (RBF) networks are used to model general non-linear discrete-time systems. In particular, reciprocal multiquadric functions are used as activation functions for the RBF networks. A stepwise regression algorithm based on orthogonalization and a series of statistical tests is employed for designing and training of the network. The identification method yields non-linear models, which are stable and linear in the model parameters. The advantages of the proposed method compared to other radial basis function methods and backpropagation neural networks are described. Finally, the effectiveness of the identification method is demonstrated by the identification of two non-linear chemical processes, a simulated continuous stirred tank reactor and an experimental pH neutralization process.