You are on page 1of 6

16th European Symposium on Computer Aided Process Engineering and 9th International Symposium on Process Systems Engineering W.

Marquardt, C. Pantelides (Editors) 2006 Published by Elsevier B.V.

901

SIMULATION AND OPTIMIZATION IN 1,3-BUTADIENE PROCESS FROM C4-CUT USING GENETIC ALGORITHM
Farhang Jalali*, Raheleh Saffari, Chem. Eng. Dept., Faculty of Eng., Univ. of Tehran, Iran fjlali@ut.ac.ir

ABSTRACT Separation of 1,3-butadiene from the C4-cut is not possible by conventional distillation due to formation of several azeotropic mixtures and very close boiling points of the components. The BASF 1,3-butadiene extraction process takes advantage of highly improved relative volatilities of most of the components in the presence of a selective solvent. The solvent is n-methylpyrrolidone (NMP) which contains approximately 8.3% water. In the present work, first,1,3-butadiene extraction process is simulated in steady-state conditions. The results of the steady-state simulation are compared with plant data and show a good agreement between these values. It was found that the NRTL equation of state is able to predict the experimental data satisfactorily throughout the process. The binary interaction coefficients of the components were tuned in this study such that the equation of state best fits the real equilibrium data. An optimization framework is proposed in this work for a synthesis of extractive distillation sequence, based on a modified genetic algorithm coupled with a sequential process simulator. In the methodology developed here simulation models are automatically generated through a process and are evaluated for various candidate configuration of the system. These candidates are suggested by a genetic algorithm that automatically guides the system towards better solutions. Keywords: Simulation, Optimization, 1,3-Butadiene

mixture. This separation technique is widely used in chemical and petrochemical industries for separating azeotropic, close-boiling and low relative volatility components from the mixture. In extractive distillation, the solvent is chosen to interact differently with the components of the original mixture, thereby, altering their relative volatilities. Since these interactions occur predominantly in the liquid phase, the solvent is continuously added close to the top of the extractive distillation column such that an appreciable amount of solvent is present in the liquid phase on all the trays below. N-Methylpyrrolidone (NMP), a solvent with high solubility and selectivity for unsaturated compounds which has been used successfully in numerous industrial plants, proved its merits for 1,3-butadiene extraction. The BASF process for the recovery of high purity 1,3-butadiene from C4-cut employs NMP as the selective solvent. Table 1 gives the selectivity of NMP in different mixtures. The high selectivity of NMP for 1,3-butadiene versus the less soluble butanes and because and more readily soluble acetylene compounds versus butadiene in the solvent makes NMP an ideal aid in the recovery of butadiene with optimum purity. The selectivity of NMP is also sufficient to separate 1,2-butadiene from 1,3-butadiene. Selectivity of NMP for propyne is relatively low. However, the difference between the boiling points is large enough to separate propyne from 1,3-butadiene down to the permissible level. In addition to the selectivity, other properties of NMP such as low vapor pressure, stability and proper solubility for acetylenes has made this solvent the best for extraction of 1,3-butadiene. Table 1: Selectivity of NMP (40 C, 760 mm Hg)
Mixture 1,3-Butadiene / i- Butane 1,3-Butadiene / n- Butane 1,3-Butadiene / 1- Butene 1,3-Butadiene / 2-cis- Butene 1,3-Butadiene / Propyne 1,2-Butadiene / 1,3-Butadiene 1-Butyne / 1,3-Butadiene Vinyl acetylene / 1,3-Butadiene Selectivity 8.52 4.37 2.66 1.65 1.09 1.88 2.46 5.44

INTRODUCTION

1,3-Butadiene is a colorless, non-corrosive gas with mild aromatic or gasoline-like odor with the boiling point of 4.4 C at atmospheric pressure and liquid density 611 Kg/m3 at 20C. It is used primarily as a co-monomer for producing styrene-butadiene rubber. Separation of 1,3butadiene from C4-cut by conventional distillation is not possible due to the formation of several azeotropes and very close boiling points of most of the components. The alternative process for separation of 1,3-butadiene from the C4-cut is extractive distillation. Extractive distillation is defined as distillation in the presence of a miscible, high boiling, relatively nonvolatile compound (i.e., the solvent) which forms no azeotrope with the other components in the

902

F. Jalali and R. Saffari


as the overhead product. In the second distillation tower a mixture containing 1,2-butadiene and C5-hydrocarbons are separated as the bottom product. The 1,3-butadiene product is withdrawn from the overhead of the final distillation tower.

PROCESS DESCRIPTION

The block diagram of BASF butadiene extraction process is shown in Figure 1 and comprises of two major process steps: The extractive distillation section The distillation section Butanes and Butense, which have the smallest solubility in NMP, are separated as the overhead product (raffinate) of the extractive distillation. A vapor stream consisting mainly of 1,3-butadiene and C4-acetylenes is sent to the second extractive distillation step. In this tower the more soluble acetylenic components are removed by means of fresh solvent, NMP. Crude butadiene is withdrawn as the overhead product of the second extractive distillation step. The C4-acetylenes are fully absorbed by NMP and are withdrawn from the degassing section. In the degassing section dissolved hydrocarbons are completely removed from the solvent. Crude butadiene obtained from the

STEADY-STATE SIMULATION

The first step in process simulation is to determine the suitable thermodynamic system, i.e., a proper equation of state with attention to the components and process conditions. For polar or non-ideal chemical systems usually binary thermodynamic systems are employed. In this way, an equation of state (such as ideal gas, Peng-Robinson, SoaveRedlich-Kwang, etc.) is used to predict the vapor fugacity coefficient while another equation of state (usually based on the activity model) is employed for the liquid phase. In the present study, NRTL is used as the main equation of state, for liquid phase with its binary interaction coefficients being modified in order to best represent the vapourliquid equilibrium behaviour of the system. In two distillation towers because of absence of NMP, Peng Robinson is used as the thermodynamic model. The vapor phase is assumed to be ideal in this work. The flow diagram of process is shown in Figure (2). According to the PFD. The liquid C4-fraction from battery limit is fed to the feed surge drum where fluctuations of the flow to the process can be compensated. The reboilers which are heated up with hot NMP that is routed from the bottom of the degassing tower make the feed to be vapored.. The required flow rate of C4-hydrocarbons to the main washer is measured in the vapor line connecting to drum with mainwasher and adjusted by flow control. The vaporized C4-cut enters the main washer at the bottom together with the top gas from the rectifier and is washed counter-currently with the NMP-solvent. This packed column is simulated with a tray tower that the number of trays is calculated in attention with the type of the packing and HETP. In this way 1,3-butadiene is almost completely absorbed. The overhead product contains the bulk of propane, propene, propadiene, butanes and butanes present in the C4-feed stock of the plant. The solvent N-methylpyrrolidone (NMP) with approx. 8.3% (w/w) water, is fed on the top bed of the main washer. The solvent flow is controlled by FIC. The solvent withdrawn from the bottom of main washer is pumped on the top of the rectifier. The level controller of main washer acts on a flow controller to ensure that the flow of solvent can be kept sufficiently constant. In the upper part of rectifier the less soluble i.e. more volatile butene are stripped from the solvent by a counter-current vapor stream of the more readily soluble butadiene rising from the bottom. The gasous mixture of 1,3-butadiene and butene is leaving the top of the rectifier and is fed back to the bottom of the main washer. The concentration of 1,3-butadiene in

Raffinate C4-Cut Extractive Distillation

Propyne 1,3-butadiene Distillation

C4-acetylenes

C4/C5-HC

Figure 1- Block Diagram The extractive distillation section comprises the following subsections: 1. The extractive distillation section Extractive Distillation I Extractive Distillation Degassing 2. The distillation section The distillation section comprises the following subsections: Propyne Distillation Column Final Distillation Column extractive distillation is further purified in two subsequent distillation towers. In the first distillation tower propyne together with some 1,3-butadiene for dilution is withdrawn

Simulation and Optimization in 1,3-butadiene Process from C4-cut


the vapor rises to its maximum between the rectifiers upper and lower part. From this location a butadiene-rich side-stream is withdrawn and fed into the bottom of the after-washer . The bottom of rectifier is divided into two compartments . the solvent loaded with hydrocarbons is drawn off from one compartment and routed to heat exchanger where it is heated up on the tube side of the exchanger train. Then it is flashed into the other compartment of the rectifier. By means of the pre-degassing a considerable amount of the dissolved hydrocarbons is vaporized. Butadiene still containing C4-acetylene is withdrawn as a gasous side-stream and fed into the bottom of after-washer. In this column butenyne (vinylacetylene) and 1-butyne (ethylacetylene) are removed from 1,3butadiene. The C4-acetylene are more soluble in NMP e.g. less volatile then 1,3-butadiene and are removed in counter-current with fresh solvent that is fed on the top of the column. The C4-acetylene absorbed in the solvent is

903

drawn from the bottom of after-washer and transferred with after washer pump to the rectifier. Water which had been present in the overhead vapor is separated from the organic phase in the bootleg of an accumulator. the pre degassing solvent from the rectifier is heated up again in solvent heater and flashed into the degassing tower. The hydrocarbons dissolved in the loaded solvent from rectifier have to be separated completely before the solvent is recycled to the extractive distillation. The degassing of the solvent is carried out in the degassing tower by stripping with solvent and water evaporated in reboiler. The crude butadiene is fed to the propyne distillation column and the bottom product of the propyne column is fed into the final distillation column. Here components with lower volatility than 1,3-butadiene are separated as the bottom product while 1,3-butadiene is withdrawn as the overhead products.

Figure 2: Process Flow Diagram

The results of the steady state simulation for some main parameters of the process are shown in Tables 2 to 6. The corresponding experimental values are also given in the table. As could be seen from this figures, there is a good agreement between the simulation and actual plant data. Table 2: Comparison between design and simulation values in main washer column
Parameter Mass fraction of 1,3butadiene in overhead stream Simulation value 0.18 Design value 0.19

Parameter Mass fraction of i-butene in overhead stream Mass fraction of 1,3butadiene in bottom stream Mass fraction of i-butene in bottom stream Temperature of bottom stream (C)

Simulation value 53.14 69.75 8.27 42.4

Design value 49.96 73.18 7.54 43.3

904
Table 3: Comparison between design and simulation values in Rectifier tower
Parameter Mass fraction of 1,3butadiene in overhead stream Mass fraction of 1-butene in overhead stream Mass fraction of i-butene in overhead stream Mass fraction of butane in overhead stream Mass fraction of 1,3butadiene in overhead stream Simulation value 69.64 3.4 16.7 1.8 Design value 68.92 3.6 17.28 1.93 78.9 Mass fraction of 1,3butadiene in outletstream Mass fraction of i-butene in outlet stream Mass fraction of transe-2butene in outlet stream Mass fraction of cis-2butene in outlet stream

F. Jalali and R. Saffari


Table 6: Comparison between design and simulation values in final distillation tower
Parameter Simulation value 99.46 0.0 0.022 0.30 Design value 99.7 0.045 0.21 0.22

4
Table 4: Comparison between design and simulation values in After washer tower
Parameter Mass fraction of 1,3butadiene in overhead stream Mass fraction of 1butyne in overhead stream Mass fraction of butenyne in overhead stream Mass fraction of 1,2butadiene in overhead stream Simulation value Design value

OPTIMIZATION

98.7 0.04 0.03 0.41

99.08 0.00 0.00 0.43

Table 5: Comparison between design and simulation values in degassing tower


Parameter Mass fraction butadiene in stream Mass fraction butadiene in stream Temperature of stream (C) of 1,3overhead of 1,2overhead overhead Simulation value Design value

81.15 6.48 99.14

80.79 6.39 99.4

Evolutionary algorithms (EAs) are based on the paradigm of evolution. Of course, the mechanisms of the nature for improving the extremely complex biological organisms can not easily be transferred to an algorithm for the optimization of technical systems. But several approaches with different degrees of complexity are developed as discussed for instance by Baeck and Schwefel (1993). In the terms of EAs a representation of the technical system to be optimized is called an individual. A population is a set of individuals competing with each other with respect to their target function values (fitness) Starting from an initial random population of individuals, new populations are iteratively created and their target function values are calculated by respective invocations of an evaluation function. Only the best individuals are selected as the basis (parents) for the next population. The process of generating new individuals from two randomly chosen parents is modeled by the genetic operators recombination, shown in figure 1 and mutation, a further stochastic variation for each single parameter setting of the newly created individual. The definition of these variations is strongly dependent on the representation of the individual, corresponding to different types of EAs. Genetic algorithms are stochastic methods based on the idea of evolution and survival of the fittest. In a GA, a set of values of the optimization variables forms an individual, usually codified in a chromosome through a series of bits (01). The algorithm starts generating a random population (a group of individuals), and then repetitively evolving it with three basic genetic operators: selection, crossover, and mutation. For a detailed explanation on genetic algorithms and operators. In the present work, Genetic Algorithm which is one of the most important of EAs is implemented to optimize some operation conditions and parameters in Butadiene extraction plant. It is calculated that Uniform crossover was preferred over single point, with 0.75 as the best

Simulation and Optimization in 1,3-butadiene Process from C4-cut


value for the crossover probability .Generations of 25 individuals gave the best results in comparison with larger or smaller populations and the best values for the probability of the mutation is 0.005 which is a governing factor in the performance of the GA.
Optimal design Model construction

905

user

Codification and paremeters

5 IMPLEMENTION OF GENETIC ALGORITHM WITH A COMMERCIAL SIMULATOR


Distillation is still one of the most important operations in process engineering, The modeling and formulation of such models, however, is difficult and time consuming. In addition to the time and expertise needed to formulate these models, the synthesis and design of distillation sequences pose other difficulties for their solution. The use of rigorous design and thermodynamic models leads to very large non-convex models and very difficult to converge. Moreover, taking into account structural and design decisions, such as the existence of stages, columns, condensers and reboilers, have lead to the inclusion of integer variables further increasing the difficulty of solving the model. To compensate for these difficulties, it is often necessary to supply initial values for the optimization variables very close to the actual solution, something that is not always an easy task, and even recent works have used simplifications for the design model, thermodynamics, hydraulics, or the cost functions to obtain feasible solutions or to examine complex superstructures in synthesis problems . An alternative to reduce some of these problems is the use of commercial simulators. These tools include a variety of highly efficient rigorous thermodynamic and design models that allow the process engineer to evaluate different flowsheets and modeling options in an easy way. However, the computational structure of modular simulators has not allowed the complete incorporation of the latest improvements in mathematical optimization algorithms, usually based on the evaluation of the gradients of the model, information that is not directly available in the simulator. A commercial simulation software is used as the simulator to evaluate the individuals generated by the GA. The values of the optimization variables are passed to the simulator through a program implemented in Visual BASIC to control the simulator. Through the client object the simulated file is open, closed, and manipulated to modify the structure of the flowchart process. The input and output data in the simulation are organized where the values of the variables within the simulation are read and modified. In the architecture used in this work, the user interacts with the GA defining the parameters of the algorithm, explicit constraints and convergence options, with the simulator, to select the mathematical models. This process is depicted in Fig. 3.

simulator

Codified data, simulation input

GA

Simulation output, fitness

Fig3. Implemented architecture between simulator and the GA

CONCLUSION

In this section, the optimization on operation conditions in an extractive distillation systems is presented to illustrate the advantages of the proposed framework. We have considered the After washer tower which is one of the main parts of this plant that the feed of distillation section is withdrawn from this tower. The objective function in this case is to maximize the mass percent of 1,3-Butadiene that is the overhead stream.

CASE 1
The aim is to optimize the mass percent of 1,3-Butadiene in the over head stream of after washer tower. The number of stages, the feeding stage and the reflux ratio are constant. The pressure at the top plate(the stage of solvent entrance) is 490 kPa and at the bottom( the stage of feeding) is 510 kPa. The solvent flow enters at 42C and the mass flow of solvent is 33000Kg/hr. The column was simulated in the simulation software program. The thermodynamic properties were calculated with the equation of state of NRTL which were modified. Several numerical experiments were studied to analyze the behavior of the GA and the proposed strategies under different conditions, the variables in this case are the temperature of the solvent and the pressure of feed that should be optimize which were codified in the genetic algorithm resulting in a chromosome. The population size of the GA was set to 25 individuals and 80, 100 and 200 generations were defined. The table (7) shows the results of this optimization. The results shows that the mass percent of Butadiene in overhead stream will be increased by decreasing the temperature of entrance solvent and increasing the pressure of the feed. Its clear that because the solution of hydrocarbons in NMP is exothermic, the reduction of temperature and pressure increasing will maximize the mass percent of Butadiene in overhead stream.

906
Item 1 2 3 Generation 80 100 200 Solvent temperature (C) 38.5 38 38 Feed pressure (Kpa) 513 513 513 1,3-Butadiene (%) 0.9759 0.9760 0.9760

F. Jalali and R. Saffari


Time (s) 25 45 130

Table 7: The result of operation conditions in after washer column(case 1)

CASE 2
In this case, again, the aim is to optimize the mass percent of 1,3-Butadiene in the overhead stream of after washer tower. But this time the variables are temperature and mass flow of solvent. Like case 1, The column was simulated in the simulation software program. The thermodynamic properties were calculated with the equation of state of NRTL which were modified. The population size Solvent temperature (C) 38.5 38 38

of the GA was set to 25 individuals and 80, 100 and 200 generations were defined. The table (8) declares the results of this optimization. The results shows that the mass percent of Butadiene in overhead stream will be increased by increasing the solvent temperature and mass flow.

Item 1 2 3

Generation 80 100 200

Solvent mass flow (Kg/hr) 38935 38950 38950

1,3-Butadiene (%) 0.974 0.975 0.975

Time (s) 40 70 150

Table 8: The result of operation conditions in after washer column(case 2)

ACKNOWLEDGMENTS
The research described in this paper is supported by Amir Kabir petrochemical company. The authors would like to be gratefully for technical and financial supports.

REFRENCES
Amir Kabir petrochemical company Documents. William L.Luyben. 1990. Process modeling, simulation and control for chemical engineers. McGraw-Hill International Editions. William D.Mccain,Jr. 1990. The properties of petroleum fluids. PennWell publishing company. J,Vladimir de Oliveria,A.M.Cohen Uller, 1995, Solubility of pure 1,3 butadiene an methyl propene and their mixture in pure n-methyl-2-pyrrolidone and its aqueos solutions; Fluid Phase Equilibria 118(1996) 113-14. Mario Llano-Restrepo, Jaime Aguilar-Arias, 2002 Modeling and simulation of saline extractive distillation columns for the production of absolute ethanol, Computer and Chemical Engineering 27(2003) 527-549.

Goldberg, David E. Genetic Algorithms in Search ,Optimization and Machine Learning. AddisonWesley Pub. Co. 1989. David A Coley, An Introduction to Genetic Algorithms for Scientists and Engineers, World Scientific, 1998. Goldberg, David E. The Design of Innovation : Lessons from and for competent Genetic Algorithms, Boston, MA: Kluwer Academic Publishers, 2002. Perry R.H & Chilton C.H. Chemical Engineering Handbook, 5 th ed, MC Graw Hill, 1973. SRI(Stanford Research Institue), document NO.247, Butadiene with NMP, Page 60-64 Kefeng Wang, Yu Qian!, Yi Yuan, Pingjing Yao, Synthesis and optimization of heat integrated distillation systems using an improved genetic algorithm, Computers and Chemical Engineering 23 (1998) 125-136. Jose Leboreiro, Joaquin Acevedo, Processes synthesis and design of distillation sequences using modular simulators, a genetic algorithm framework, Computers and Chemical Engineering , 28 (2004) 1223 1236.

You might also like