Armando B. Corripio

Dynamic neural networks partial least squares (DNNPLS) identification of multivariable processes

This paper presents the dynamic neural networks partial least squares (DNNPLS) as a strategy for open-loop identification of multivariable chemical processes that circumvent some of the difficulties associated with multivariable process control. The DNNPLS is an extension of the neural networks’ partial least squares (NNPLS) developed by Qin and McAvoy (Comp. Chem. Eng. 20 (1992) 379). Here, a dynamic extension to the NNPLS algorithm is proposed in which the static neural network models in the latent space (inner relationship) are replaced by dynamic neural network models. Though this approach has previously been dismissed as being sub-optimal (Am. Inst. Chem. Eng. J. 38 (1992) 1593; Chem. Eng. Sci. 48 (1993) 3447) in terms of the outer relationship (relationship between the residuals), Lakshminarayanan et al. (Am. Inst. Chem. Eng. J. 43 (1997) 2307) have shown that this sub-optimality problem comes into prominence only when no attention is placed on the design of the plant probing signals. As illustrations, the DNNPLS identification strategy is implemented on simulations of a model IV fluid catalytic cracking unit (FCCU) and of an isothermal reactor. In both cases, it is shown that the methodology is capable of modeling the dynamics of the chemical processes and an improved performance is achieved over that of the PLS-ARMA (Comp. Chem. Eng. 20 (1996) 147) for the isothermal reactor. # 2002 Elsevier Science Ltd. All rights reserved.

On-line optimization of a model IV fluid catalytic cracking unit

Abstract This paper presents the application of an on-line constrained optimization algorithm to the simulation of a Model IV Fluid Catalytic Cracking Unit (the well known Amoco challenge problem proposed by McFarlane et al. Dynamic simulator for a model IV fluid catalytic cracking unit. Comp. & Chem. Engng. 17 (3) (1993) 275–300.). The original model is modified to include reaction yields by incorporating the model of an FCC by Lee and Groves, Mathematical model of the fluidized bed catalytic cracking plant. Trans. Soc. Comput. Simulat. 2 (1985) 219. The optimization algorithm, Supervisory Multivariable Constrained Optimization (SMCO) developed by (Daniel et al., An algorithm for supervisory multivariable constrained optimization of industrial processes. Chem. Engng. Comm. 130 (1994) 203–223.). is applied to the simulation of the FCC Unit and tested for feed composition and ambient temperature disturbances. SMCO is designed for on-line implementation in distributed control systems and to minimize a real cost function.

A computer control algorithm for feedback composition control using off-line analysis

Abstract A tunable feedback control algorithm is developed for a first-order plus dead time process which has a variable sample interval and a variable dead time. This variable sample interval, variable dead time (VSIVDT) controller is applied to controlling the product composition of a blender using off-line composition analysis. The blender and controllers are simulated using the IBM Real-Time Process Management System/Advanced Control System (RTPMS/ACS). The VSIVDT controller performed better than an alternative controller based on the Dahlin algorithm which uses off-line analysis. When compared to controllers using on-line analysis, it performed as well as a PI controller but not as well as a controller using the Dahlin algorithm.

Implementation of a dynamically-compensated PID control algorithm

Abstract A PID feedback control algorithm with dynamic compensation for process dead time is proposed. The algorithm is formulated in the same form as the standard PID algorithm so that proportional and derivative kicks can be prevented on set point changes. Tuning formulas for the algorithm are presented which allw the user to directly set the sensitivity of the algorithm to set-point changes.

On-line optimization of distillation columns in series

Abstract This paper presents an on-line optimization algorithm for distillation columns which consists of a steepest descent technique based on a simple model of product recovery with on-line estimation of the critical model parameter. The algorithm is developed for a single column and then for a two-column train. A dynamic programming approach is used to reduce the optimization problem to a single parameter search for each column in the train. The resulting algorithm is simple and computer resource requirements are small. The algorithm has been successfully used in two industrial applications, one consisting of two columns in series and the other of three columns in series.

An Algorithm for Supervisory Multivariable Constrained Optimization of Industrial Processes.

Abstract An algorithm is presented for supervisory optimization of industrial processes that integrates the minimization of operating costs with process operating constraints. It is assumed that the supervisory algorithm manipulates the set points of a lower-level control system and that the set points are updated at long enough intervals of time so that the process reaches steady state between set point updates. This steady state assumption greatly simplifies the algorithm computations and, more importantly, significantly reduces the effort required for process identification. This article develops the algorithm and then presents results from its application to a simulated distillation train. The simulation parallels an application of the algorithm to an actual industrial train on a commercial distributed control system.

On-Line Model Identification and Control Using the Kalman Filter.

Abstract : For some systems a model determined off-line can be satisfactorily used to design control systems. However, for many systems, the best values for the model parameters will change as the system operation varies. For effective control these parameter changes must be identified on-line and incorporated into the digital control strategy. Equally as important as the identification of changing parameters is the identification of unmeasured disturbances which upset the system operation. Ideally, the identification of model parameters and unmeasured disturbances would be accomplished by measuring only the controlled variable. One method which holds promise is the extended Kalman filter. The Kalman filter, which was developed for rejecting noise from measurements, is applied in this paper to problems of parameters and disturbance identification and state estimation of first- and second-order processes. The primary objective is to evaluate its performance for estimating the unmeasured disturbances encountered so frequently in operating control systems. (Author)

Operating regime decomposition using neural networks

This article addresses the problem of identification of a nonlinear process operating over a wide range of conditions. The global space is divided into multiple local regimes, a nonlinear model is developed for each regime, and a quadratic programming-based algorithm is used to ensure smooth transition between the regimes on-line. The use of nonlinear models as opposed to linear models reduces the number of local regimes needed. Neural networks are used to model these regimes because of their strong ability to capture nonlinearity, and their combination with the switching algorithm improves transient performance. The performance of the method is demonstrated on an exothermic CSTR and a pH neutralization process.

Using Off-Line Analysis for Dynamic Feedback Composition control

A tunable feedback control algorithm is developed for a first-order plus dead time process which has a variable sample interval and a variable dead time. This variable sample interval, variable dead time (VSIVDT) controller is applied to controlling the product composition of a blender using off-line composition analysis. The blender and controllers are simulated using the IBM Advanced Control System. The VSIVDT controller performed better than an alternative controller based on the Dahlin algorithm which uses off-line analysis. When compared to controllers using on-line analysis, it performed as well as a PI controller but not as well as a controller using the Dahlin algorithm.

A Computer Program for Detailed Frequency Response Analysis of Distillation Column Dynamics

A model of a distillation column is developed that includes component and enthalpy balances, equilibrium relationships, variable pressure and complex tray hydraulic relationships. The model is incorporated into a computer program that produces Bode plots of the transfer functions between any input variable and any column variable. The paper includes examples of the use of the program to study column dynamic response and control loop interaction as a function of frequency.