M.-V. Le Lann

A neural software sensor for online prediction of coagulant dosage in a drinking water treatment plant

Artificial neural networks (ANNs) have been applied to an increasing number of real-world problems of considerable complexity. Considered good pattern recognition engines, they offer ideal solutions to a variety of problems such as prediction and modelling where the industrial processes are highly complex. The present paper reports on the elaboration and the validation of a ‘software sensor’ using ANNs for online prediction of optimal coagulant dosage from raw water quality measurements, in a drinking water treatment plant. In the first part, the main parameters affecting the coagulant dosage are determined using a Principal Component Analysis. A brief description of this statistical study is given and experimental results are included. The second part of this work is dedicated to the development of a neural software sensor and the generation of an uncertainty indicator attached to the prediction. Bootstrap sampling has been used to generate a confidence interval for the model outputs. The ANN model was developed using the Levenberg-Marquardt method in combination with ‘weight decay’ regularization to avoid over-fitting. A linear regression model has also been developed for comparison with the ANN model. Experimental and performance results obtained from real data are presented and discussed.

Robust iterative learning control of an exothermic semi-batch chemical reactor

This work focuses on the temperature control of a semi-batch chemical reactor used for flue chemicals production. Such reactor is equipped with a heating/cooling system composed of different thermal fluids. In order to ensure the tracking performance of the desired temperature profile, an iterative learning control (ILC) named batch model predictive control (BMPC) has been adopted. The synthesis of the considered strategy is illustrated and improvements of the algorithm scheme are proposed. Firstly, a guaranteed convergence of the algorithm is illustrated. Secondly, in presence of high frequency disturbance effects, an off-line filtering is adopted for enhancing the achieved performances. Third, a robust supervisory control procedure is employed to choose the right fluid and to reduce the superfluous fluid changeovers, mainly where fluids are of different nature. Finally, the incidence of repetitive disturbances, on line low frequency disturbances and model mismatch are investigated through simulation runs.

Robust fault detection for the hybrid dynamic systems

Abstract PrODHyS is a dynamic hybrid simulation environment which provides common and reusable object oriented components designed for the development and the management of dynamic simulation of industrial systems. In this work, we present a method for the fault detection based on the comparison between the real system and the reference model evolution generated by the extended Kalman filter. The reference model is simulated by PrODHyS.

Supervisory Temperature Control of Batch Reactors: Application to an Industrial Plant

Abstract This paper presents thermal control of a batch reactor equipped with a multifluid heating / cooling system with an intermediate utility: pressurised water. The appropriate utility is chosen by comparing the thermal flux that has to be transferred to the reaction mixture with the heat transfer capacities of the available utilities. The thermal flux to be transferred is calculated by a predictive control algorithm. Two different algorithms were considered: GPC and a predictive control algorithm that is based on the gradient of the control variable and of its reference value. Experimental performance is shown on a 160 litres industrial batch reactor of the company « Sanofi-Synthelabo » where this thermal controller, implemented in the DCS « DeltaV», is now being used.

Dynamics of reactive distillation for the production of ethyl acetate: experiments at a pilot plant and modelling

Abstract In order to understand the complex behaviour of the reactive distillation process and to be able to provide an accurate design of a reactive column, detailed analyses on both continuous and transient regime become necessary. The objective is the definition of a reliable simulation model, based on experimental data obtained from a real pilot-scale plant device for the heterogeneously catalysed esterification of acetic acid and ethanol to form ethyl acetate and water. The choice of the parameters for the continuous equilibrium model was discussed and the simulation results provided good agreement with experimental data, revealing an interesting sensitivity of the catalyst activity to the feed composition. Once column configuration and operational parameters were validated, dynamic experiments were realized so as to interpret the sensitivity of different disturbances. Feed flow rates, reflux ratio and heat duty were perturbed and the consequent open loop transient responses were identified. The assessment of hydrodynamic parameters and the validation of the transient data allow the definition of a reliable dynamic model that represents tendencies and behaviours of the process well. The resulting model is to be applied into a more complex controllability methodology.

IMPLEMENTATION PROCEDURE OF AN ADVANCED SUPERVISORY AND CONTROL STRATEGY IN THE PHARMACEUTICAL INDUSTRY

Abstract Most batch pharmaceutical fermentations have successive phases of operation. Detection of process phase landmarks is important for improving the performance of process monitoring and control. This study considers techniques for adjustment of batch data lengths to match landmarks of phases during the progress of the batch. Time synchronization and landmark alignment techniques are presented and their integration with on-line monitoring is discussed through illustrative examples.

EXPERIMENTAL APPLICATION OF THE ITERATIVE LEARNING CONTROL TO THE TEMPERATURE CONTROL OF BATCH REACTOR

Abstract A learning control solution to the problem of finding a finite-time optimal control history that minimizes a quadratic cost is presented. Learning is achieved without requiring a detailed knowledge of the system, which may be affected by unknown but repetitive disturbances. The objective of this work is considered as a validation of the results obtained in (Mezghani, et al. , 2001). It focuses on the temperature control of a semi-batch chemical reactor used for fine chemicals production. Such reactor is equipped with a heating/cooling system composed of different thermal fluids. With less modelling investigation, a feedback-feedforward control structure is proposed for ensuring the tracking performance of the desired temperature profile. Such strategy is derived from the family of the Iterative Learning Control named Batch Model Predictive Control ( BMPC ). The synthesis of the considered strategy is studied and improvements of the algorithm features are proposed. A robust supervisory control procedure is employed to choose the right fluid and to reduce the superfluous fluid changeovers. Experimental results are presented to illustrate the practical appeal and effectiveness of the proposed scheme.

A model-based supervisory control routine for temperature control of batch reactors: experimental results

Publisher Summary This chapter describes the application of a model-based strategy for the supervisory temperature control of a 12 liters industrial pilot-batch reactor. The reactor is equipped with a multifluid heating/cooling system. The automation of this type of heating/cooling system requires a supervisory control to handle automatically the utility fluids fed into the reactor jacket as well as the intermediate air purge and the refilling after the changeover of the fluid. The objective is to obtain good control of the process even during transient operating conditions, such as refilling of the jacket, which cannot be neglected on industrial scale. The supervisory control is based on the dynamic model of the reactor pilot plant. Once the utility fluid being chosen by the supervisory routine, its flowrate is computed by a controller, which is a nonlinear model predictive control algorithm. Different experiments are presented to demonstrate the performances of the supervisory control of the pilot-plant reactor.

Inferential control of microbial batch culture

Publisher Summary Batch culture of Brevibacterium linens is grown on lactic acid and amino acids as substrate in a 20 liters bioreactor. During the culture growth, rate is controlled according to a time-varying set point profile. Growth rate is inferred on from exhaust-gas measurements. In the present case, carbon exhaust rate is used. As this variable is corrupted with noise and disturbances, a model based filter, including a simple kinetic model, is implemented. Experimental evidence shows that the filter rejects noise and disturbances to a large extend and renders the growth-rate estimation suitable for control purposes. The set point for growth rate is the input to a supervisory control routine based on the simple kinetic model. This routine determines the values of the manipulated operation conditions so that the growth rate matches the desired set point. As shown, this routine is inherently suitable for multi-input single output (MISO)-control, although culture temperature is the only manipulated operation condition used in the present case. Experimental results prove that the presented approach allows for precisely tracking time-varying growth rate set point profiles.

Simulation-aided implementation of supervisory control of industrial batch reactors

Temperature control of batch reactors equipped with a multifluid type heating/cooling system containing an intermediate thermal fluid (pressurised water) is presented. By means of this intermediate fluid the heating/cooling system gains monofluid type behaviour over a wide temperature range. Hereby operation of the reactor is improved. The control strategy consists in using a GPC algorithm to calculate the thermal flux that has to be transferred between the heating/cooling system and the reaction mixture. The thermal flux is then introduced into a model based supervisory algorithm that automatically chooses on-line the appropriate utility. Testing the intervening software is partially carried out by connecting the original plant control software to a simulator of the process. Hereby the software validation procedure is accelerated and utility consumption is reduced, too. Experimental validation of the presented approach on an industrial 160 litres batch reactor gives satisfying performances.