Dynamic Parameter Estimation and Optimization for Batch Distillation (original) (raw)

Nonlinear dynamic modeling of multicomponent batch distillation: a case study

Brazilian Journal of Chemical Engineering, 2002

The aim of this work is to compare several of the commercial dynamic models for batch distillation available worldwide. In this context, BATCHFRAC™, CHEMCAD™ BATCH, and HYSYS.Plant ® software performances are compared to experimental data. The software can be used as soft sensors, playing the roll of ad-hoc observers or estimators for control objectives. Rigorous models were used as an alternative to predict the concentration profile and to specify the optimal switching time from products to slop cuts. The performance of a nonlinear model obtained using a novel identification algorithm was also studied. In addition, the strategy for continuous separation was revised with residue curve map analysis using Aspen SPLIT™.

Dynamic modeling of batch distillation: comparison between commercial software

The current increasing production of small volume, high added value products has called attention to batch production technologies. Although batch distillation typically consumes more energy than continuous distillation, it provides more flexibility and involves lower capital investment. Thus, since energy costs are not too significant in the separation of small volume of high profit products, batch distillation is often attractive. Available batch distillation models have appeared in the literature, ranging from very simple approaches to more detailed. Commercial software packages such as BATCHFRAC™ (Boston et al. 1981), BASIS (Simulation Sciences Inc., 1989) and ProSimBatch, (1992) are available only to model batch distillation operation. Luyben W. (1992) implemented an inferential control using a rigorous quasi-dynamic model of a batch distillation column. Recent progress in nonlinear model-based control techniques have made the practical applicability of nonlinear controller a true. Many of these techniques use a nonlinear dynamic process model directly in the control law development phase (which is performed off-line). These features are very attractive for practical implementation. Henson and Seborg (1997) list a number of simulation and experimental studies in which this feedback controllers were used. In this context, a nonlinear model–based strategy implemented with commercial software is analyzed here. Therefore BATCHFRAC™, HYSYS® and CHEMCAD® performances were tested to evaluate whether their implementation in an inferential control structure and /or soft sensor could be attractive. Hence, using the worldwide commercial available tools, the dynamic behavior of a batch distillation column was modeled.

Validation of a solution model for the optimization of a binary batch distillation column

2005

For the optimization of dynamic systems, it is customary to use measurements to combat the effect of uncertainty. In this context, an approach that consists of tracking the necessary conditions of optimality is gaining in popularity. The approach relies strongly on the ability to formulate an appropriate solution model, i.e. an approximate parameterization of the optimal inputs with a precise link to the necessary conditions of optimality. Hence, the capability of a solution model to optimize an uncertain process needs to be assessed. This paper introduces an optimality measure that can be used to verify the conjecture that the solution model derived from a simplified process model can be applied to a more rigorous process model with negligible performance penalty. This conjecture is tested in a simulation of the dynamic optimization of a batch distillation column.

Dynamic composition estimation for a ternary batch distillation

Chemical Engineering Science, 2008

A novel Kalman estimator has been proposed to provide the estimates of dynamic composition in a ternary batch distillation process operated in an optimal-reflux policy. The estimator is formulated based on a sequence of reduced-order process models representing a whole batch behavior. Therefore, the full-order models are first developed around different pseudo-steady-state operating conditions along batch optimal profiles. Then they reduce their orders to achieve all state observability and controllability by a balanced truncation method. In the estimator scheme, the reduced models as well as relevant covariance matrices of process noise are pre-scheduled and switched according to any desired periods. Four important issues have been studied including selection of a sensor frequency, effects of an integrating step size, a state initialization and a measurement noise. The performances of the reduced estimator have been investigated and compared with those of a conventional nonlinear estimator. Simulation results have demonstrated that the performances of the novel linear estimator are reasonably good and almost identical to the nonlinear estimator in all cases, though the linear estimator performs rather sensitively to the effect of high measurement noise. Nevertheless, it has been found to be applicable to implement in real plants with much lower computation effort, easier state initialization and unrequired a priori knowledge of thermodynamics.

Inferential Estimation for a Ternary Batch Distillation

2008

A Kalman filter (KF) estimator has been formulated using a sequence of reduced-order models representing a whole batch behavior for providing the estimates of dynamic composition in a ternary batch distillation process operated in an optimal-reflux policy. A set of full-order models is firstly obtained by linearizing around different pseudo-steady state operating conditions along batch optimal profiles. They are further reduced their orders to achieve their observability and controllability individually by using a model reduction method. The performances of the reduced-estimator have been investigated and compared with those of a conventional nonlinear estimator. Simulation results have demonstrated that the performances of the proposed estimator are reasonably good and almost identical to the conventional one in all cases.

Optimization of Multiple-Fraction Batch Distillation by Nonlinear-Programming

AICHE JOURNAL, 1990

Batch distillation processes are very attractive for the recent development of the chemical industry: multipurpose, flexible plants and fine chemistry. For many separations of high-added value products, even a modest change in operating conditions has a significant economic impact-there is an important challenge for optimizing such processes.

Batch distillation: simulation and experimental validation

Chemical Engineering and Processing: Process Intensification, 2004

Batch industries, specially the fine chemical industry, need rapid and robust mathematical models that can easily predict the degree of separation of mixtures like volatile organic compounds. Otherwise, in the study of on line production scheduling and planning of fine chemical plants, there is a lack of quick but reliable and more accurate than black box models in batch distillation. In this work, a robust and rapid mathematical model has been developed, based on mass balance and relative volatility. The model has been extensively tested in a variety of scenarios using a batch distillation pilot plant that has been constructed adhoc. The satisfactory results obtained allows to use the model in real time for a better control and management of batch distillation, resulting in an improved production scheduling/rescheduling of batch processes.

Hybrid Model for Optimization of Crude Distillation Units

2015

Planning, scheduling and real time optimization (RTO) are currently implemented by using different types of models, which causes discrepancies between their results. This work presents a single model of a crude distillation unit (preflash, atmospheric, and vacuum towers) suitable for all of these applications, thereby eliminating discrepancies between models used in these decision processes. Product TBP curves are predicted via partial least squares model from the feed TBP curve and operating conditions (flows, pumparound heat duties, furnace coil outlet temperatures). Combined with volumetric and energy balances, this enables prediction of crude distillation on par with a rigorous distillation model, with 0.5% RMSE over a wide range of conditions. Associated properties (e.g. gravity, sulfur) are computed for each product based on its distillation curve and corresponding property distribution in the feed. Model structure makes it particularly amenable for development from plant data.