Economic Evaluation for the Retrofit of Chemical Processes through

4566

Ind. Eng. Chem. Res. 1996, 35, 4566-4578

Economic Evaluation for the Retrofit of Chemical Processes through Waste Minimization and Process Integration Mauricio M. Dantus and Karen A. High* School of Chemical Engineering, 423 Engineering North, Oklahoma State University, Stillwater, Oklahoma 74078

This research involves the development of an economic-based methodology for waste minimization and reduction of energy consumption in the chemical industry by modifying existing processes. The methodology consists of identifying waste minimization options through a sensitivity analysis and flowsheet configurations through a hierarchical procedure. The alternatives identified together with the heat-exchanger network were used to construct a superstructure that was formulated as an MINLP problem. The superstructure was evaluated and optimized with ASPEN PLUS. An economic model based on the net present value method that incorporates both manufacturing and capital costs was used to select the most profitable configuration. The production of methyl chloride was used to evaluate the methodology. Results obtained identified the optimized base case as the best process. When considering byproducts as wastes, the hydrochlorination of methanol using an adiabatic plug flow reactor was identified as the best process. Introduction The chemical processing industry is faced with a need to manufacture quality products while minimizing production costs and complying with a variety of safety and environmental regulations. These regulations include the Clean Air Act (CAA), the Clean Water Act (CWA), the Resource Conservation and Recovery Act (RCRA), the National Emission Standards for Hazardous Air Pollutants (NESHAPs), and, recently, the Hazardous Organic NESHAPs (HON) (Zanetti, 1994). In 1990, the Environmental Protection Agency (EPA) promulgated the Pollution Prevention Act. This act declared that the national policy of the United States is to prevent or reduce pollution at the source, that pollution which cannot be prevented should be recycled in an environmentally safe manner, and that waste disposal should be employed only as a last resort. As part of its pollution prevention strategy the EPA initiated the 33/50 program, which is a voluntary program to reduce the emissions of 17 chemicals by 33% by 1992 and by 50% by 1995 (Freeman et al., 1992). As a result of the Pollution Prevention Act, the constant change in regulations, the risk of financial liability associated with regulatory compliance, and increasing pollution control costs which were estimated to be $152.6 billion by 1995 (Hydrocarbon Process., 1993), the end-of-pipe treatment approach is no longer feasible or recommended. A suggested approach to this problem is prevention at the source. By taking this approach, industry reduces the regulation requirements, follows the EPA waste management hierarchy (Mizsey, 1994), and generates a process that is less of a burden on the environment. This approach has been proven to be cost effective (see Table 1). Waste minimization, which is defined by the EPA as the reduction, to the greatest extent possible, of hazardous pollutants that are generated and subsequently treated, sorted, or disposed of (U.S. EPA, 1988) (see Figure 1), has been an area of intense research in process synthesis, design, and retrofit (Douglas, 1992; Fonyo et al., 1994; Farag et al., 1992; Manousiouthakis * Author to whom all correspondence should be addressed.

S0888-5885(95)00778-0 CCC: $12.00

and Allen, 1994; Hopper et al., 1993). To date, however, there is not much work in handling waste minimization in chemical engineering by using computer-aided process simulation and optimization strategies where economics are the major consideration. Waste minimization can be accomplished by retrofitting existing processes. This is done by performing minor modifications to the current process in order to increase their overall performance. When considering the retrofit problem, there can be several alternatives. These are in order of increasing costs (Grossmann et al., 1987) as follows: modify the operating conditions, redefine the use of the present equipment, modify the present equipment, and add new equipment. The three basic methodologies used for retrofit design are (Gundersen, 1989) as follows: (1) Hierarchical design approaches: These consist of a series of heuristic rules to screen process alternatives. (2) Pinch technology techniques: These techniques were developed during the energy crisis for the efficient use of energy. (3) Mathematical programming tools: These methods rely on optimization techniques to solve the mixedinteger nonlinear programming (MINLP) problem. These methods originally emerged as grassroot design methodologies for the synthesis of new processes but have been applied to retrofit problems. The present work developed a combined approach using process simulation, optimization, and economic analysis tools to formulate the problem as a mixedinteger nonlinear programming (MINLP) problem. This made it possible to economically evaluate the alternatives identified where the environmental impact and the energy consumption of an existing process were reduced. The alternatives were evaluated through the use of a superstructure. This superstructure included the alternatives identified using a sensitivity analysis and a hierarchical procedure. However, this superstructure does not include all the possible alternatives, since this represents an infinite number of alternatives. The objectives of this work were to (a) determine the advantages of applying waste minimization techniques © 1996 American Chemical Society

Ind. Eng. Chem. Res., Vol. 35, No. 12, 1996 4567 Table 1. Waste Reduction Projectsa company

accomplishments

Amoco Waste Minimization Program (1983)

Between 1983 and 1988, Amoco reduced its hazardous waste by 86%, saving the company about $50 million.

Chevron Save Money and Reduce Toxics Program (SMART, 1987)

From 1987 to 1990, Chevron reduced hazardous waste by 60% and saved more than $10 million in disposal costs.

Dow Waste Reduction Always Pays (WRAP, 1986)

SARA 313 overall releases are down from 12 252 tons in 1987 to 9659 tons in 1989, a 21% reduction. Offsite transfers are down from 2855 tons (1987) to 2422 tons (1989), a reduction of 15%. Air emissions for 1989 showed a 54% decrease from 1984.

General Dynamics Zero Discharge (1985)

Nearly 40 million lb. of hazardous waste discharge was eliminated from 1984 to 1988 (approximately 72%). Sales increased from $7.3 to 9.35 billion over the same period.

IBM

Hazardous waste generation was reduced 38% from 1984 to 1988; 84% of IBM’s hazardous waste was recycled in 1988; 28% of all solid waste from IBM U.S. operations was recycled in 1988; IBM U.S. emissions were reduced 20% from 1987 to 1988; and IBM U.S. had a decrease of 25% in its CFC emissions between 1987 and 1988.

Monsanto Priority One (TRI wastes)

From 1987 to 1990, Monsanto achieved a 39% reduction in hazardous air emissions.

Speciality Adhesives and Chemicals

An analysis of an amine production process increased the conversion, reducing waste by 95 tons/yr. By considering the recycling of excess reactant, an additional waste reduction of 70 tons/yr and a decrease of 20% of manufacturing costs were obtained.

a

Sources: Benforado and Ridlehoover (1991); Freeman (1992); Thayer (1992); Woodman (1989).

Figure 1. Waste minimization techniques (U.S. EPA, 1988).

to chemical processes during the retrofit phase, (b) develop a methodology for economically retrofitting existing processes to minimize waste, and (c) apply the methodology developed to an existing process for the production of methyl chloride, by incorporating environmental, economic, and energy efficiency constraints. Methodology The final decision criteria used to select a specific process alternative will rely on an economic evaluation

of this alternative. The economic model used to evaluate each alternative should take into account the capital and manufacturing costs (direct production costs, fixed charges, plant overhead costs, and general expenses) associated with a specific process. This takes into consideration the fact that the most environmentally friendly process might not be the most economical. The economic tool used to compare process alternatives was the net present value (NPV) method that represents the best profitability comparison tool (Peters and Timmerhaus, 1991; Brealey and Myers, 1988). When comparing

4568 Ind. Eng. Chem. Res., Vol. 35, No. 12, 1996

Figure 2. General methodology.

different process alternatives, the optimum alternative selected will be the one that satisfies production demand with a minimum cost. This is important, as the waste minimization approach is focused not only on the manufacturing process itself but on the product life cycle, from raw materials to final disposition (Freeman et al., 1992). A sensitivity analysis, together with a hierarchical method (Douglas and Stephanopoulos, 1994), was used as a starting point to identify possible alternatives and to generate the MINLP superstructure. By identifying the heat sources and heat sinks of a process, the pinch technology method (Linnhoff, 1982) estimates the minimum utility requirements, the number of heat-exchanger units, and the heat-exchanger area. Although this approach has been successful, it is a step procedure, where each step is affected by the previous one. A suggested approach (Ciric and Floudas, 1990) is to treat the heat integration as a simultaneous optimization problem, formulating it as an MINLP problem. The MINLP problem is solved using a modified algorithm based on the LP/NLP-based branch and bound algorithm (Raman and Grossmann, 1992). The algorithm can be modified to include logic constraints and the use of the disjunctive normal form (DNF) approach to determine the next unit to be branched as well as additional units to be fixed (Raman and Grossmann, 1993). Most of the work in the area of MINLP has been done with equation-based simulators (Diwekar and Rubin, 1993). However, for this study, the formulated problem was solved using ASPEN PLUS, a sequential modular simulator. Sequential modular simulators such as ASPEN PLUS are readily available and are user friendly. The built-in models as well as the in-line FORTRAN capabilities were used to model the process and solve the MINLP problem. The proposed methodology will consist of four sections (see Figure 2): (1) Development of a base case model.

(2) Generation of process retrofit alternatives. (3) Evaluation and optimization of process retrofit alternatives. (4) Evaluation of future scenarios. These four topics are discussed followed by the application of a case study to evaluate the methodology. The production of methyl chloride (CH3Cl) (AIChE, 1966) was used as a case study to evaluate the methodology. This process was selected because of its environmental impact, regulatory restrictions, and potential for improvement. Development of Base Case Model The modification of a process requires an incentive. This incentive can be economical, environmental, quality oriented, safety improvement, etc. A base case process model is used to evaluate the current performance of the process and serves as a guide to analyze the different retrofit alternatives. The process model will function as an inexpensive experimental tool that will allow us to evaluate the economic and environmental effect that the various process alternatives will have in the performance of our process. In the development of the base case process model, it is important to determine its scope by identifying the units and operations to be included. As more units are included, the model will become more accurate but will also make the solution more computationally intensive. For the case of waste minimization projects, it is necessary to include all the units that are point sources for each waste stream in the process. The process model should also include operations and units that have an important economic impact in the process. For this study, the ASPEN PLUS process simulator was used as the modeling tool. The economic performance of the base case model is used as the evaluation criteria to analyze each process alternative. The economic performance is evaluated in terms of profitability. This is to measure the amount

Ind. Eng. Chem. Res., Vol. 35, No. 12, 1996 4569

of profit that can be obtained from a specific process. The profitability criteria used is a function of both capital and manufacturing costs (direct production costs, fixed charges, plant overhead costs, and general expenses). The capital and manufacturing costs are included in order to consider that the most environmentally friendly process might not be the most economical. The Net Present Value (NPV) method represents the best profitability comparison tool (Peters and Timmerhaus, 1991; Brealey and Myers, 1988). This method will maximize the future worth of the company. Based on the NPV method and considering the salvage value as zero, the economic model applied is presented in eq 1. ny

NPV ) -CF +

[fD(Io(1 - Tx) + (CF)DfTx)] ∑ j)1

(1)

The operating income Io in eq 1 is a function of revenue obtained and manufacturing costs.

Io ) In - Mc

(2)

The income term In in eq 2 considers revenue obtained from the product and byproducts. The byproducts obtained in a process may be considered as a revenue. This might change in the future due to market and regulation changes. An evaluation of this effect should be performed. The alternatives to be considered will be evaluated in terms of their compliance with specified product demand. The manufacturing costs Mc consider the expenses associated with raw materials Rw, waste disposal W, operating costs O, and any additional expenditures Ac specific to the model considered.

Mc ) Rw + W + O + Ac

(3)

Pollution control and abatement costs are estimated to increase by 20-30% annually (LaGrega et al., 1994). The pollution control and abatement expenses of each process stream are included as part of the manufacturing costs. Therefore, the waste component in eq 3 corresponds to the costs associated with the final disposition of the waste leaving the process. All waste treatment expenditures are considered as part of the operating costs Oi and can be modeled as part of the flowsheet.

Oi ) Pri + Uri + Mri

(4)

where the operating costs of unit i are a function of the power, utilities, and maintenance requirements. By using this approach we can follow the EPA’s preferred hazardous management strategy by shifting the objective function toward waste reduction and away from waste disposal. The discount factor fD in eq 1 considers the cost of capital at an interest rate i. A suggested value for the interest rate is 15% (Peters and Timmerhaus, 1991).

fD )

1 (1 + i)j

(5)

The depreciation factor Df in eq 1 is estimated using the sum-of-the-years-digits method.

Df )

2(ny - j + 1) ny(ny + 1)

(6)

Finally, the capital costs CF in eq 1 are estimated using the ASPEN PLUS cost blocks. The economic model has three variables that can be modified as required: the tax rate Tx, the interest rate i, and the number of years ny. Generation of Process Retrofit Alternatives There are two levels of possible retrofit alternatives: The zero investment level and the variable investment level. The zero investment level considers the modification of existing operating parameters such as temperature, pressure, and flowrate. This approach requires no major capital investments. The variables to be modified are identified through a sensitivity analysis where their effect is compared against the performance of the base case operation. The second level of alternatives, the variable investment level, considers different flowsheet configurations through the modification of the existing process structure by rearranging the existing units or the acquisition of new ones. The alternatives in this level are identified through the use of a hierarchical procedure (Chadha, 1994; Douglas and Stephanopoulos, 1994; Linnhoff, 1994). The flowsheet configurations for this study were identified using the hierarchical procedure by Douglas and Stephanopoulos (1994). When applying the hierarchical procedure, it is important to consider waste minimization options at each level. These options should be addressed as multimedia pollution prevention strategies, where the waste is eliminated, not changed from one media to another media. Apart from waste minimization options, other criteria should be included such as safety and operability. The process alternatives are compared against the economic performance of the base case. The optimum alternative will be the one that satisfies production demand with a minimum cost. Process Integration. The process integration tool was originally developed during the energy crisis in the 1980s. The main purpose is to maximize the heat transfer between the process streams and minimize the requirements of external utilities. Thus, by recovering the energy generated in a process, the utility requirements are satisfied and the utility cost is minimized. The retrofit of heat-exchanger networks (HEN) has been accomplished by formulating the problem as an MINLP problem, using the one-to-one approach (Ciric and Floudas, 1990). There is no reference in the literature regarding the identification of optimum HEN configurations using ASPEN PLUS. A detailed description of the implementation of the one-to-one approach using ASPEN PLUS, is presented in Appendix A. An evaluation of a heat-exchanger network modification needs to consider not only the utility costs and capital costs associated with heat-exchanger area but also the cost of reassigning existing units and repiping streams. The cost of repiping streams can vary between 4 and 20% of the capital costs (Peters and Timmerhaus, 1991). To accurately estimate this effect, layout considerations should be included in the economic model. The capital cost CF for each potential match was calculated using eq 7 (Douglas, 1988).

CF )

( (

) )

MES Q 101.3 280 U(LMTD)

0.65

Fc

(7)


Evaluation and Optimization of Process Retrofit Alternatives The retrofit alternatives identified and the HEN configurations are used to construct a flowsheet superstructure. The superstructure is used to select the best process flowsheet, by formulating it as a mixed-integer nonlinear programming (MINLP) problem (see eq 8).

max c(x,y) x,y

s.t.

h(x,y) ) 0 g(x,y) e 0 y ∈ {0.1}m,

x ∈ Rn

(8)

The problem is solved by varying the continuous variables x such as temperature, pressure, and flowrate, identified through the sensitivity analysis, and the discrete variables y that denote the existence of a specific unit, identified through the hierarchical procedure. The variables are varied in order to maximize the objective function c formulated using the economic model (see eq 1), subject to a set of equality constraints h and a set of inequality constraints g. The detailed description for the ASPEN PLUS MINLP algorithm is presented in Appendix B. The optimization results obtained using this method will produce a local optimum for the problem. This is especially true when ASPEN PLUS is used as the optimization tool. Very good initial guesses are required for the optimization routine to converge. The initial guesses should also be close to the optimum answer. A wider range in the optimization variables may cause the objective function to present discontinuities which are not easily handled with the ASPEN PLUS optimization subroutine. A mathematical procedure needs to be applied in order to verify that a global optimum was obtained. Due to the black box characteristics of ASPEN PLUS, the behavior of the objective function is unknown and the mathematical procedure cannot be applied. An alternative method that can be used consists of applying different initial guesses and verifying that the same optimum answer has been obtained.

larly, as the starting point in the Rochow synthesis (Morreto et al., 1985). Other uses include the manufacture of methyl cellulose, agricultural chemicals, quaternary amines, and butyl rubber. The methyl chloride U.S. demand was 363 and 374 million kg for 1994 and 1995, respectively, and is estimated to be 431 million kg by 1999 (Chem. Mark. Rep., 1995). Methyl chloride and the corresponding byproducts methylene chloride (CH2Cl2), chloroform (CHCl3), and carbon tetrachloride (CCl4) are considered hazardous wastes under RCRA and are regulated by CWA and CAA and included in HON. Methylene chloride, a suspected carcinogen, chloroform, and carbon tetrachloride are also included in the 33/50 program. Fugitive emissions of carbon tetrachloride need to be eliminated by the year 2000 according to the Montreal Protocol (Dep. State (U.S.) Bull., 1987), and a ban on its production goes in effect by 1996. The byproducts identified by Johnson et al. (1959) are also regulated under RCRA, CWA, CAA, and HON. Of these byproducts, methyl chloroform and trichloroethylene are also included in the 33/50 program. The process generates several waste streams. An attempt will be made to minimize or eliminate these waste streams: 1. Stream leaving the absorber containing HCl, water, and chloromethanes. 2. Waste streams leaving the drying towers. 3. The amount of byproducts generated. 4. Heavy ends after the distillation sequence. The base case reported (AIChE, 1966) was taken as the current process in operation (see Figure 3). The reported variables for the process used to model the base case situation are shown in Table 2. Synthesis Step. The synthesis step consists of a continuously stirred tank reactor (CSTR) where the primary and side reactions take place. The feed to the reactor is a mixture of chlorine and methane which react to produce methyl chloride. Subsequent reactions give methylene chloride (CH2Cl2), chloroform (CHCl3), and carbon tetrachloride (CCl4), as well as substantial amounts of hydrogen chloride (HCl) (see Scheme 1).

Scheme 1 Evaluation of Future Scenarios A final analysis is performed to evaluate the flexibility of the optimum process. This analysis will determine the possible effect that any future changes in our current constraints such as product demand, environmental regulations, economic data, etc., will have on the process performance. This analysis can be carried out using a sensitivity analysis to determine the effect of specific variables in the overall process performance. Case Study: Production of Methyl Chloride This section presents the application of the proposed methodology to a specific process: the production of methyl chloride by the thermal chlorination of methane. This process was selected because of the environmental impact and potential for improvement. There is not a reported study in the literature where a simulation of the complete process, an economic analysis, or a waste reduction study has been done. Development of the Base Case Model Methyl chloride (CH3Cl), also known as chloromethane, is mostly used for the manufacture of silicones, particu-

CH4 + Cl2 f CH3Cl + HCl CH3Cl + Cl2 f CH2Cl2 + HCl CH2Cl2 + Cl2 f CHCl3 + HCl CHCl3 + Cl2 f CCl4 + HCl The ASPEN PLUS block RCSTR was used to model the reactor. This block uses the power law expression to define the reaction kinetics (see eq 9 and Table 3).

r ) ATae-Ea/RT

∏(Ci)b

i

(9)

The feed to the reactor must be heated to 300 °C for the reaction to be initiated. It has been found in commercial operations that an operating temperature of 400-450 °C is necessary to have a stable, selfsustaining reaction. The reactor should not be operated higher than 500 °C since pyrolysis can occur. This is a very exothermic reaction and can lead to possible reactor explosion. The methane feed should contain a minimal amount of hydrocarbon impurities (100 ppm, excluding nitrogen)


Figure 3. Base case process. Table 2. Reported Input Variablesa variable

0.3 isothermal CSTR 450 25 7.8 -50 72.6

Source: AIChE (1966).

Table 3. Kinetic Parameters for the Base Case Model (Scheme 1)a reaction no. 1 2 3 4 a

component

H1 (m3 atm/(kg mol))

H2 (m3 atm K/( kg mol))

temp range (°C)

CH2Cl2 CHCl3 CCl4

0.058 0.072 0.078

16.2 19.6 18.7

10-35 10-35 10-35

value

feed ratio (Cl2/CH4) reactor type reaction temp (°C) reactor effluent cooling system (°C) compressor outlet pressure (atm) condenser temp (°C) product demand ((kg mol)/h) a

Table 4. Henry’s Law Constantsa

rate eq

preexponential factor (s (kg mol)/m3)-1

activation energy (kJ/(kg mol))

k1[Cl2][CH4] k2[Cl2][CH3Cl] k3[Cl2][CH2Cl2] k4[Cl2][CHCl3]

2.56 × 108 6.28 × 107 2.56 × 108 2.93 × 108

82 000 71 100 82 000 87 200

Source: Scipioni and Rapisardi (1961).

(DeForest, 1979). Under certain conditions (Johnson et al., 1959), such as high impurities and both low and high mole feed ratios, the remaining hydrocarbons may be chlorinated, generating several byproducts. Apart from the regulation status of these byproducts, they can generate subsequent problems during the separation sequence. The methane used in the process is usually obtained from natural gas, which can be an important source for impurities. The methane is purified to remove other hydrocarbons through cryogenic distillation. Separation Step. The reactor effluent is cooled to the specified temperature of 25 °C and is then passed through an absorber where HCl is removed by using water as the absorbing agent. To adequately represent

a

Source: Gossett (1987).

this operating unit, the dissociation of HCl in water has to be taken into account. Among the different data banks and chemistries available in ASPEN PLUS, eq 10 (ASPEN PLUS data bank H2OHCL) gives the best

HCl T H+ + Cl-

(10)

results when compared to experimental data (Oldershaw et al., 1947). The data package H2OHCL uses the electrolyte NRTL and Henry’s law models. The latter requires Henry’s law constants for interactions between water and the different components. Although ASPEN PLUS has several data sources for interaction parameters, ASPEN PLUS does not include information for the interactions between H2O-CH2Cl2, H2O-CHCl3, and H2O-CCl4. The reported values (Gossett, 1987) for these interactions were correlated as a function of temperature T in the form of eq 11. The final values

H ) exp(H1 - H2/T)

(11)

are shown in Table 4. The next step in the separation sequence is the removal of the water introduced in the previous step. The concentration of water throughout the process is recommended to be less than 5 × 10-5 ft3/ft3 of solution (DeForest, 1979). This minimizes corrosion when the streams contain HCl or chlorine and prevents the

4572 Ind. Eng. Chem. Res., Vol. 35, No. 12, 1996 Table 5. Summary of Base Case Blocks

Table 7. Base Case Results

operating unit

block used

variable

value

reactor cooling system absorber dryer condenser compressor distillation columns

RCSTR HEATER RADFRAC SEP HEATER COMPR RADFRAC

methyl chloride flowrate waste cooling water requirements refrigerant system requirements hot utility requirements manufacturing costs

2.99 × 107 kg/yr 1.69 × 108 kg/yr 1.95 × 1014 J/yr 7.82 × 1013 J/yr 1.30 × 1014 J/yr 1.141 × 108 $/yr

Table 6. Economic Data for the Base Case Process raw materials chlorine methane (includes purification) water waste cost utilities refrigeration condenser system compressor system heating utilities electricity

0.249 $/kg 0.159 $/kg 4.36 × 10-4 $/kg 0.70 $/L 1.10 × 10-8 $/J 1.42 × 10-8 $/J 5.65 × 10-9 $/J 0.07 $/kW h

hydrolysis and decomposition of the product and byproducts in subsequent steps. The removal of water is usually accomplished with a series of drying columns by using sulfuric acid and sodium hydroxide as drying agents. There is not much information available regarding this process. Since one of the proposed alternatives is the removal of water, the elimination of this step will be considered. This step will be represented with an ASPEN PLUS SEP block. The economic criteria used to evaluate the performance of this unit are based on the amount of waste generated, which is a function of the amount of water to be removed. The final step in the separation sequence is a series of distillation columns where the product and each of the byproducts are separated. The ASPEN PLUS DSTWU shortcut distillation method was used to obtain an initial estimate of the basic parameters. The DSTWU block uses Winn’s method to estimate the minimum number of stages, Underwood’s method to estimate the minimum reflux ratio, and Gilliland’s correlation to estimate the required reflux ratio or the required number of stages. The values obtained from applying the DSTWU block were incorporated into a more rigorous separation model RADFRAC. A summary of the blocks used to represent the different operating units is shown in Table 5. The base case reported (AIChE, 1966) was taken as the current process in operation and used to construct a model using ASPEN PLUS. The economic performance of the base case is evaluated using eq 1 and the economic data in Table 6. The methane cost presented in Table 6 includes the purification expenses (Jordan, 1972). The waste costs presented in Table 6 correspond to the treatment and disposal of chlorinated hydrocarbons by fuel blending and reclamation of the chlorine (Fox, 1994). Since the base case represents the existing process in operation, it does not consider capital investment. Based on this, the economic performance for the base case is presented in Table 7. Alternatives that perform better economically than the base case will be selected as viable. The byproducts obtained in a process may in some cases be considered as a revenue, but this might change in the future. To overcome this uncertainty, the byproducts methylene chloride, chloroform, and carbon tetrachloride were not used as a source of revenue. These byproducts were considered as wastes due to the trend

in environmental regulations. The evaluation on the effect of this assumption is presented later. Generation of Process Retrofit Alternatives The zero investment level through a sensitivity analysis was used to evaluate each variable of the base case process. The variables that showed an important effect in the overall performance of the process were identified as possible retrofit alternatives for waste minimization. The final variables identified were the reaction temperature, reactor pressure, molar feed ratio and compressor outlet pressure. The variable investment level served to identify the feasible flowsheet structure alternatives: (1) Alternative chemistry: The waste minimization approach requires the evaluation of possible alternative reaction schemes that will be more environmentally friendly. Through this analysis, the use of the hydrochlorination of methanol was selected as an alternative chemistry. The use of the new chemistry has several advantages over the current chemistry. The environmental impact is greatly reduced due to the fact that the only byproduct, methyl ether (C2H6O), is not regulated by the EPA and it is not produced in measurable quantities (Becerra et al., 1992). The use of a cryogenic unit is eliminated, reducing the operating costs; to maintain a low byproduct composition, the reactor has to be operated at a very low conversion of methane per pass, thus requiring large volumes of gas to be recycled. Finally, in the thermal chlorination process, big amounts of HCl are produced, unless a good use can be obtained from this byproduct, the HCl may become a burden on the process (DeForest, 1979). The principal drawback in the new chemistry is the increased cost of raw materials. Compared to the thermal chlorination (see Scheme 1), the hydrochlorination of methanol (alternative) uses hydrogen chloride as a raw material. This hydrogen chloride is reacted with methanol in the presence of a catalyst (see Scheme 2).

Scheme 2 CH3OH + HCl f CH3Cl + H2O 2CH3OH + HCl f (CH3)2O + H2O (CH3)2O + 2HCl f 2CH3Cl + H2O The power law model available in ASPEN PLUS was used to represent the solid catalyst system using γ-alumina as the corresponding catalyst (see eq 12).

r ) Ae-Ea/RTCHClCCH3OH A ) 2.93 × 108 (s (kg mol)/m3)-1 Ea ) 90 000 kJ/(kg mol)

(12)


Figure 4. ASPEN PLUS superstructure representation: section 1.

In Scheme 2, hydrogen chloride is usually fed in excess to reduce the formation of methyl ether. The unreacted hydrogen chloride can be recycled back to the reactor after the removal of water. However, the recovery of hydrogen chloride is difficult since it is at a concentration where an azeotropic mixture is formed with water. The reaction can be carried out in an isothermal or adiabatic PFR reactor. Both alternative reactor systems were considered as possible retrofit alternatives. The reactor effluent is cooled and separated. The liquid phase containing mostly hydrogen chloride and water is recycled back to the reactor. The vapor phase containing mostly methyl chloride is dried with sulfuric acid to remove the remaining amount of water and to eliminate the methyl ether in the final product. The vapor phase purification is similar to the drying step for the thermal chlorination process and will be evaluated on the amount of water to be removed. A simulation of the hydrochlorination of the methanol process was done using ASPEN PLUS to compare the accuracy of the model to reported results available in the literature. A sensitivity analysis was performed to identify the variables that affect the performance of the hydrochlorination of methanol (alternative). The variables identified were the reactor temperature, condenser temperature, and flash vessel pressure. (2) Alternative reactor system: The actual reactor system consists of one isothermal continuous stirred tank reactor (CSTR). Another alternative includes the use of an isothermal plug flow reactor (PFR). (3) Alternative separation sequence: The main difficulty in the separation sequence comes from the need to remove the hydrogen chloride. In the base case, this is done through the absorption with water. The presence of water can decompose the products and can create corrosion problems in streams containing chlorine and hydrogen chloride. A first alternative consists of performing an initial separation of the reactor outlet stream. The reactor outlet stream is flashed, the vapor stream returns to the reactor, and the liquid stream continues through the separation sequence. The advantage of this alternative is the reduction of the amount of product to be exposed to water.

The second alternative is to substitute the water with another solvent. Forlano (1974) suggests the use of part of the chlorinated hydrocarbons as the solvent to remove the hydrogen chloride. This eliminates the use of water or any external solvent which could further be a waste source. The absorber pressure and methyl chloride flowrate were identified through a sensitivity analysis as the critical variables of the new absorber. Evaluation and Optimization of Process Retrofit Alternatives Formulation of Superstructure. The base case and all of the retrofit alternatives including the HEN configurations are used to formulate an overall flowsheet superstructure. The superstructure was composed of three different sections: base case alternatives keeping the water as the absorbing agent, base case alternatives using methyl chloride as the absorbing agent, and hydrochlorination of the methanol alternative process. The superstructure developed is reproduced by dividing it into the previous three sections in order to increase its readability (see Figures 4-6). For the same reason the HEN configuration is only presented in Figure 6. The results obtained from the base case simulation, the simulation of the hydrochlorination of the methanol (alternative) process, and a simulation of each alternative were used as the starting point for the optimization subroutine. Logic constraints were incorporated into the MINLP problem to reduce the number of alternatives to evaluate. The final MINLP problem consists of an objective function (see eq 1), 9 continuous variables, 10 discrete variables, and 15 constraints (see Table 8). Solution of the MINLP Problem. A simulation for each alternative was performed to use as the initial guess for the optimization subroutine and to formulate the MINLP problem. The ranges for most of the variables were obtained from the literature. The discrete variables consisted of a variation of the fraction flow. The fraction flow will have a value of 1 to represent the existence of a specific unit or process and a value of 0 to represent the absence of the specific unit or process.


Figure 5. ASPEN PLUS superstructure representation: section 2.

Figure 6. ASPEN PLUS superstructure representation: section 3. Table 8. Optimization Variables variable range continuous variables (Scheme 1) methane fresh feed ((kg mol)/h) reaction temp (°C) reaction pressure (atm) compressor outlet pressure (atm) methyl chloride absorber (atm) discrete variables (Scheme 1) Scheme 1 (Y5) isothermal CSTR (Y6) isothermal PFR (Y7) base case separation sequence (Y9) alternative separation sequence (Y10) alternative separation sequence w/water (Y11) alternative separation sequence no water (Y12) continuous variables (Scheme 2) hydrogen chloride fresh feed (kg mol)/h) reaction temp (°C) condenser outlet temperature (°C) flash vessel pressure (atm) discrete variables (Scheme 2) Scheme 2 (Y1) isothermal PFR (Y2) adiabatic PFR (Y3)

120-150 350-500 1-5 2-15 1-5 0, 1 0, 1 0, 1 0, 1 0, 1 0, 1 0, 1 85-95 260-370 0-65 1-5 0, 1 0, 1 0, 1

The first step in the MINLP procedure requires the establishment of a fixed set of variables, this set will determine the upper bound to the solution. The selected

Table 9. Hydrochlorination of Methanol Optimization Results variable

final value

hydrogen chloride fresh feed (kg/h) reaction temp (°C) condenser outlet temp (°C) flash vessel pressure (atm) cooling water requirements (J/yr) hot utility requirements (J/yr) waste (kg/yr)

86.2 284 52 1.5 2.27 × 107 0 1.66 × 107

set should give a good estimate of the optimum point in order to reduce the number of MINLP iterations. For this case, the base case was selected as the initial set. By varying the specified continuous variables, the base case was optimized. The optimum flowsheet configuration corresponds to the hydrochlorination of methanol (alternative) by using an adiabatic PFR (see Table 9). This alternative considers the case of completely replacing the current process. Even though this alternative requires high investment costs, an increase in savings for both operating costs and waste disposal costs makes this a feasible alternative. As was mentioned previously, the economic performance of the reaction Scheme 1 considered the byproducts as wastes. If the byproducts are considered a


source of revenue, the optimized base case (Scheme 1) represents the best alternative. A comparison between short-term and long-term effects including the trend in environmental constraints should be made to determine the flowsheet configuration to be selected. Future Scenarios An analysis on the effect of waste costs shows that this was a critical factor in the selection of the optimal flowsheet. The economic model used assumed the byproducts as wastes, even though revenue can be obtained from them, and a demand still exists for these products. There are reported cases in the literature (DeForest, 1979) where, in order to reduce raw material requirements, a combined process uses both the thermal chlorination of methane and the hydrochlorination of methanol (alternative). The hydrogen chloride produced in Scheme 1 is used as a raw material in Scheme 2. This alternative reduces the excess of hydrogen chloride and still has the flexibility of obtaining the different chlorinated hydrocarbons. The combined process might also be a solution in order to obtain a gradual change from the old process into the new process. The economic model did not consider the type of waste or the risk associated with each chemical and did not consider the possibility of recycling, reusing, regenerating, or treating the waste generated, but in any case the increase in the amount of waste will be directly related to the operating costs and thus affect the overall profit of the operation. This increase will greatly enhance the decision to switch to the hydrochlorination of the methanol alternative, this being a more environmentally friendly process. Several other factors were not included in the economic model such as safety and operability factors. It was established that the temperature inside the reactor for the thermal chlorination of methane requires a strict control. A PFR gives the possibility of better temperature control during the reaction than a CSTR, but since the CSTR gives a better product distribution, this type of reactor was chosen over the PFR. The hydrochlorination of the methanol (alternative) process is a safer process. This occurs due to the absence of the reaction temperature restrictions available in the thermal chlorination process. Conclusions An economic-based methodology was developed to retrofit existing processes in the chemical industry. This methodology was aimed at reducing the waste generated at the source and improving the energy consumption, while satisfying all the environmental and product specifications and still remaining profitable. The methodology accomplished the combined use of the three retrofit techniques: hierarchical design procedure, heat integration technology, and mathematical programming methods. Based on these techniques, a combined methodology was proposed which consisted of four general steps: development of a base case model, generation of process retrofit alternatives, evaluation and optimization of process alternatives, and evaluation of future scenarios. The final decision criteria used to select a specific process alternative will rely on an economic evaluation. An emphasis was made in the development of an economic model based on the net present value method to accurately compare each alternative. The proposed

methodology was successfully applied to the manufacture of methyl chloride. Acknowledgment This present work was made possible through the support of Consejo Nacional de Ciencia y Tecnologı´a (CONACYT), the Fulbright Program, and University Center of Water Research at Oklahoma State University. Nomenclature Roman Letters a ) temperature exponent A ) preexponential factor Ac ) additional costs b ) concentration exponent c ) objective function C ) concentration CF ) fixed capital cost Df ) depreciation factor Ea ) activation energy f ) flowrate Fc ) design factor fD ) discount factor g ) inequality constraints h ) equality constraints H ) Henry’s law constant H1, H2 ) Henry’s law parameters i ) interest rate I ) cold streams Io ) operating income In ) income J ) hot streams LMTD ) log mean temperature difference Mc ) manufacturing costs Mri ) maintenance requirements of process unit i MES ) Marshall and Swift index ny ) number of years Oi ) operating costs of process unit i P ) list of open nodes Pri ) power requirements of process unit i Q ) heat load R ) gas constant r ) rate of reaction Rw ) raw materials costs T ) temperature Tx ) tax rate U ) overall heat-transfer coefficient Uri ) utility requirements of process unit i W ) waste cost x ) continuous variables y ) integer variables z ) lower bound Greek Letters ∆Tmin ) minimum temperature difference

Appendix A: ASPEN PLUS Heat Integration Using the One-to-One Approach The one-to-one approach (Ciric and Floudas, 1990) consists of the representation of all of the possible HEN configurations through the use of a superstructure. The input to the superstructure corresponds to a set of cold streams I and a set of hot streams J. A cold stream is defined as a process stream that is to be heated, and a hot stream is defined as a process stream that is to be cooled. Each potential match in the superstructure corresponds to the existence of a heat-exchanger unit


constraints are already specified by ASPEN PLUS, the only additional constraint will be to specify the value for ∆Tmin. Additional constraints may be added to specify temperatures and to not allow stream splitting. An initial guess for the heat flow of each stream is required, as was needed for the flowsheet superstructure. The starting point can be estimated by calculating the maximum heat load required by a specific stream. Appendix B: MINLP Algorithm for ASPEN PLUS

Figure 7. HEN superstructure.

which may be present or purchased. The bypass of a specific heat exchanger unit is also considered through the use of splitters and mixer blocks. The heat duty supplied by the utilities will correspond to the heat load that could not be supplied internally by the process streams. This is accomplished by adding the utility matches at the end of the superstructure (see Figure 7). The following assumptions are made: the heat exchangers are of the countercurrent type, only one match between two streams is allowed, and no matches between hot streams or between cold streams are allowed. The use of ASPEN PLUS for the formulation of the HEN superstructure presents several advantages due to the existing internal constraints such as mass and energy balances that are not required to be specified as part of the optimization block. Each potential heatexchanger unit is represented in ASPEN PLUS with a HEATER block, and the mixer and splitter units are represented with FSPLIT and MIXER blocks, respectively. The minimum temperature difference ∆Tmin between streams that are to exchange heat represents a critical parameter in a process integration study. The minimum temperature difference value used will have an effect on utilities and capital costs and can vary between 10 and 40 °C (Linnhoff, 1982). A value of 10 °C was assumed as the minimum temperature difference. An approximation for the logarithmic mean temperature difference (LMTD) is used to avoid numerical difficulties when the approach temperatures of both sides of the exchanger are equal (Yee et al., 1990). The MINLP problem for retrofitting heat-exchanger networks using ASPEN PLUS consists of the representation of the HEN using a superstructure (see Figure 7) and an objective function (see eq 1). The continuous variables x will be represented by varying the heat flow to each possible match, through the use of heat streams. The discrete variables y are represented using ASPEN PLUS FSPLIT blocks by varying the flow fraction between 0 and 1. Each possible match is represented by two HEATER blocks; therefore, additional constraints are added to consider both blocks as one unit. Since all the physical

The retrofit alternatives identified are used to construct a flowsheet superstructure. The superstructure is used to select the best process flowsheet by formulating it as a mixed-integer nonlinear programming (MINLP) problem (see eq 8). The problem is solved by varying the continuous variables x such as temperature, pressure, and flowrate and the discrete variables y that denote the existence of a specific unit. The variables are varied in order to maximize the objective function c formulated using the economic model (see eq 1), subject to a set of equality constraints h and a set of inequality constraints g. The sequential quadratic programming (SQP) method built in ASPEN PLUS was used as the optimization algorithm. Other than the specific optimization subroutine that is used in the evaluation of the superstructure, two factors are critical in order to obtain an optimum solution: a feasible search region, and an initial starting point. The feasible search region will determine the boundaries of the solution. It is important to be aware that the boundaries selected will enclose the local optimum but not necessarily the global one. These boundaries are generally fixed by physical constraints or can be determined through several case studies. A good initial starting point is required to guarantee that the result obtained represents a true optimum and to reduce the number of iterations. The closer the initial guess is to the optimum, the faster the optimization subroutine will converge. The results obtained from the base case simulation and from the simulation of each alternative are used as the initial guess for the optimization subroutine. The identified variables are then varied between the selected ranges until an optimum answer is found. The MINLP subroutine considers the possibility of having zero flows to specific operating units within the flowsheet. This is to simulate a particular part of the flowsheet being ignored. To adequately handle the zero flows with ASPEN PLUS, the tear streams used to converge the flowsheet should only use the material flow as a convergence variable. The default state variables of pressure and enthalpy should not be used. The use of such variables will cause convergence problems for the optimization block. The objective function sampling variables such as heat duties, power requirements, and cost variables need to be initialized to prevent FORTRAN errors that occur due to zero flow calculations. The initialization is done using a FORTRAN block as the first step in the computational sequence. For the case of heat-exchanger duties, these need to be initialized as part of the objective function. Logic constraints are incorporated into the MINLP problem to reduce the number of alternatives to evaluate. The superstructure is simplified by representing the splitter and mixer units as Z blocks and the other units as Y blocks. The logic constraints are then


Figure 8. ASPEN PLUS MINLP algorithm.

formulated by using Boolean decision variables and the logic representation of the superstructure (Raman and Grossmann, 1992). The algorithm used for solving the MINLP problems was based on the LP/NLP based branch and bound algorithm (Quesada and Grossmann, 1992). This algorithm was modified to include a disjunctive normal form approach (DNF) (Raman and Grossmann, 1993). The DNF approach is useful in cases when the number of feasible alternatives is significantly smaller than the number of 0-1 combinations. The DNF approach consists of applying the branching rule and the logical chaining. The branching rule is used to determine the next variable to be branched in the solution tree in order to search among branches that will lead to integer solutions as soon as possible. The logical chaining attempts to fix as many of the other binary variables as possible by analyzing the logic representation of the

superstructure. A summary of the MINLP algorithm for ASPEN PLUS is presented in Figure 8. Literature Cited AIChE. 1964 Problem: Meeting Market DemandssChloromethanes Plant. In Student Contest Problems and First-Prize-Winning Solutions 1959-1965; American Institute of Chemical Engineers: New York, 1966. Becerra, A. M.; Castro, A. E. L.; Ardissone, D. E.; Ponzi, M. I. Kinetics of the Catalytic Hydrochlorination of Methanol to Methyl Chloride. Ind. Eng. Chem. Res. 1992, 31, 1040. Benforado, D. M.; Ridlehoover, G. Pollution Prevention: One Firm’s Experience. Chem. Eng. 1991, 98 (9), 130. Brealey, R. A.; Myers, S. C. Principles of Corporate Finance; McGraw-Hill: New York, 1988. Chadha, N. Develop Multimedia Pollution Prevention Strategies. Chem. Eng. Prog. 1994, 90 (11), 32. Chemical Profiles: Methyl Chloride. Chem. Mark. Rep. 1995, March 6.

4578 Ind. Eng. Chem. Res., Vol. 35, No. 12, 1996 Ciric, A. R.; Floudas, C. A. A Mixed Integer Nonlinear Programming Model for Retrofitting Heat-Exchanger Networks. Ind. Eng. Chem. Res. 1990, 29, 239. DeForest, E. M. Chloromethanes. Encyclopedia of Chemical Processing Design; Dekker: New York, 1979; Vol. 8, p 214. Diwekar, U. M.; Rubin, E. S. Efficient Handling of the Implicit Constraints Problem for the ASPEN MINLP Synthesizer. Ind. Eng. Chem. Res. 1993, 32, 2006. Douglas, J. M. Conceptual Design of Chemical Processes; McGrawHill: New York, 1988. Douglas, J. M. Process Synthesis for Waste Minimization. Ind. Eng. Chem. Res. 1992, 31, 238. Douglas, J. M.; Stephanopoulos, G. Hierarchical Approaches in Conceptual Process Design: Framework and Computer-Aided Implementation. Proceedings of the Foundations of Computer Aided Process Design; American Institute of Chemical Engineers: New York, 1995. Environmental Processes ’93: Challenge in the 90’s. Hydrocarbon Process. 1993, 72 (8), 71. Farag, I. H.; Chen, C. C.; Wu, P.; Rosen, J. B. Modeling Pollution Prevention. CHEMTECH 1992, 22 (1), 54. Fonyo, Z., Kurum, S.; Rippin, D. W. T. Process Development for Waste Minimization: The Retrofitting Problem. Comput. Chem. Eng. 1994, 18, S591. Forlano, L. Production of Methylene Chloride. U.S. Patent No. 3,848,007, May 12, 1974. Fox, J. Personal communication, Oklahoma State University, 1994. Freeman, H.; Harten, T.; Springer, J.; Randall, P.; Curran, M. A.; Stone, K. Industrial Pollution Prevention: A Critical Review. J. Air Waste Manage. Assoc. 1992, 42, 619. Gossett, J. M. Measurement of Henry’s Law Constants for C1 and C2 Chlorinated Hydrocarbons. Environ. Sci. Technol. 1987, 21, 202. Grossmann, I. E.; Westerberg, W. A.; Biegler, L. T. Retrofit Design of Processes. Proceedings of the First International Conference on Foundations of Computer Aided Process Operations; Elsevier: New York, 1987. Gundersen, T. Retrofit Process Design Research and Applications of Systematic Methods. Proceedings of the Third International Conference on Foundations of Computer-Aided Process Design; Elsevier: New York, 1989. Hopper, J. R.; Yaws, C. L.; Ho, T. C.; Vichailak, M. Waste Minimization by Process Modification. Waste Manage. 1993, 13, 3. Johnson, P. R.; Parsons, J. L.; Roberts, J. B. Making Chloromethanes. Ind. Eng. Chem. 1959, 51, 499. Jordan, C. H. Liquified Natural Gas: Natural Gas Processing at Low Temperatures. Chem. Eng. Prog. 1972, 68 (9), 52. LaGrega, M. D.; Buckingham, P. L.; Evans, J. C. Hazardous Waste Management; McGraw-Hill: New York, 1994. Linnhoff, B. Use Pinch Analysis to Knock Down Capital Cost and Emissions. Chem. Eng. Prog. 1994, 90 (8), 32. Linnhoff, B.; Townsend, D. W.; Boland, D.; Hewitt, G. F.; Thomas, B. E. A.; Guy, A. R.; Marsland, R. H. User Guide on Process

Integration for the Efficient Use of Energy; Pergamon Press: Great Britain, 1982. Manousiouthakis, V.; Allen, D. Process Design for Waste Minimization. Proceedings of the Foundations of Computer Aided Process Design; American Institute of Chemical Engineers: New York, 1995. Mizsey, P. Waste Reduction in the Chemical Industry: A Two Level Problem. J. Hazard. Mater. 1994, 37, 1. Montreal Protocol: Controlling Substances that Deplete the Ozone Layer. Dep. State (U.S.) Bull. 1987, 87, 60. Morreto, H.; Manfred, S.; Wagner, G. Silicones. Ullmann’s Encycl. Ind. Chem. 1985, A24, 57. Oldershaw, C. F.; Simenson, L.; Brown, T.; Radcliffe, F. Absorption and Purification of Hydrogen Chloride from Chlorination of Hydrocarbons. Chem. Eng. Prog. 1947, 43 (7), 371. Peters, M. S; Timmerhaus, K. D. Plant Design and Economics for Chemical Engineers, 4th ed.; McGraw-Hill: New York, 1991. Quesada, I.; Grossmann, I. E. An LP/NLP Based Branch and Bound Algorithm for Convex MINLP Optimization Problems. Comput. Chem. Eng. 1992, 16, 937. Raman, R.; Grossmann, I. E. Integration of Logic and Heuristic Knowledge in MINLP Optimization for Process Synthesis. Comput. Chem. Eng. 1992, 16, 155. Raman, R.; Grossmann, I. E. Symbolic Integration of Logic in Mixed-Integer Programming Techniques for Process Synthesis. Comput. Chem. Eng. 1993, 17, 909. Scipioni, A.; Rapisardi, E. Sul rapporto di formazione dei clorometani in funzione della cinetica di reazione. Chim. Ind. (Milan) 1961, 43, 1286. Thayer, A. M. Pollution Reduction. Chem. Eng. News 1992, 70 (46), 22. U.S. Environmental Protection Agency. Waste Minimization Opportunity Assessment Manual; U.S. EPA: Washington, DC, 1988; EPA/625/7-88/003. Yee, T. F.; Grossman, I. E.; Kravanja, Z. Simultaneous Optimization Models for Heat IntegrationsI. Area and Energy Targeting and Modeling of Multi-stream Exchangers. Comput. Chem. Eng. 1990, 14, 1151. Woodman, H. J. Why not Zero Waste? EPA J. 1989, Nov/Dec, 39. Zanetti, R. The Growing Cost of Cleanup. Chem. Eng. 1994, 101, 5.

Received for review December 27, 1995 Revised manuscript received September 18, 1996 Accepted September 18, 1996X IE9507787

X Abstract published in Advance ACS Abstracts, November 1, 1996.

Economic Evaluation for the Retrofit of Chemical Processes through

Recommend Documents