A New Retrofit Design Methodology for Identifying, Developing, and

1842

Ind. Eng. Chem. Res. 2005, 44, 1842-1853

A New Retrofit Design Methodology for Identifying, Developing, and Evaluating Retrofit Projects for Cost-Efficiency Improvements in Continuous Chemical Processes Eric Uerdingen,† Ulrich Fischer,*,† Rafiqul Gani,# and Konrad Hungerbu 1 hler† Safety & Environmental Technology Group, Institute for Chemical & Bioengineering, Swiss Federal Institute of Technology (ETH), CH-8093 Zurich, Switzerland, and Technical University of Denmark, Department of Chemical Engineering, CAPEC, DK-2800 Lyngby, Denmark

This paper introduces a new systematic retrofit design method for screening, identifying, and evaluating retrofit options targeted at improving the cost-efficiency of a continuous chemical process. The method is organized in five steps: (1) base case analysis, (2) generation of retrofit options, (3) rough economic evaluation of the retrofit options, (4) process optimization with regard to retrofit options that do not require investment, and (5) feasibility study as well as the economic profitability of the retrofit options that require investment. The first three steps were covered in a prior paper (Uerdingen, E. et al. AIChE J. 2003, 49, 2400-2418), while here Steps (4) and (5) are described in detail and the overall methodology is demonstrated using a case study from the fine chemical industry. For this industrial process, a number of interesting retrofit options were identified through the new method. The process was then optimized with regard to those retrofit options (optimization parameters) that do not require investment. For exemplary purposes, one of the remaining retrofit options that requires investment (structural retrofit alternatives) was evaluated in detail on its feasibility and economic profitability. The results obtained for this case study demonstrate the benefits of the new retrofit design method in systematically identifying and evaluating economically beneficial retrofit options for continuous chemical processes. Introduction The change toward globalization on the chemical market has caused an increasing competition in chemical industry in the past decades. In the face of emerging global markets, many chemicals that decades ago held the status of “speciality chemicals” have been subject to ever shrinking profit margins and gradually change to “commodity chemicals”. Consequently, improvements in cost-efficiency are more than ever needed to ensure competitiveness through price leadership on global markets that quite often carry overcapacities. To be still competitive, many existing production facilities require constant improvements for cost-efficiency. Building completely new plants usually demands large investment. This more and more favors less costly plant retrofit strategies as the only economically viable solution. Different methodologies have been used for evaluating the retrofit potential of a chemical process with regard to improved cost-efficiency. For example, Rapoport et al.2 presented a strategy using heuristic rules for the generation of retrofit options, while Jaksland et al.3 presented a thermodynamic insights based synthesis method for the generation of options applicable to new processes as well as retrofitting of existing processes, Tjoe and Linnhoff4 applied pinch analysis, previously introduced by Linnhoff and Flower,5 to retrofit design, * To whom correspondence should be addressed. Tel: +41 44 6325668. Fax: +41 1 6321189. E-mail: ufischer@ chem.ethz.ch. † Swiss Federal Institute of Technology (ETH). # Technical University of Denmark.

and among many others Ciric and Floudas,6 Jackson and Grossmann,7 and Sorsak and Kravanja8 used algorithmic approaches such as mixed integer nonlinear programming (MINLP). Also combinations of these methods have been applied; for example, Kovac and Glavic9 combined thermodynamic and algorithmic methods for retrofitting complex and energy intensive processes. A number of retrofit studies concentrated on single aspects of a chemical process; for example, Ku¨ru¨m et al.10 focused on entrainer selection, Dantus and High11 on waste minimization or process integration, Guntern et al.12 on optimizing the reactor, Gadalla et al.13 and Liu and Jobson14 on the optimization of distillation columns, Ciric and Floudas6 and Yeap et al.15 on heatexchange networks, Fraser and Hallale16 on massexchange networks, Jo¨dicke et al.17 and Kim and Smith18 on water-reuse and cooling-water systems, respectively, and van der Helm and High19 on waste minimization. Only a few studies have aimed at revealing retrofit opportunities of an entire chemical process. Fisher et al.20 and Rapoport et al.2 presented methodologies enabling a systematic analysis of process flowsheets with regard to their retrofit options. The method presented by Rapoport et al.2 uses rules based on heuristics and is organized in a hierarchical procedure, in which the improvement of the existing flowsheet, the selection and optimization of additional operation units, and the optimized design of heat integration are covered in this sequence. Fisher et al.20 presented a method for developing and also screening retrofit opportunities. It is again organized as a hierarchical procedure with the following sequence: (a) use an operating cost diagram for

10.1021/ie049065r CCC: $30.25 © 2005 American Chemical Society Published on Web 02/16/2005

Ind. Eng. Chem. Res., Vol. 44, No. 6, 2005 1843

Figure 2. Detailed description of Step 4 of the retrofit design methodology.

Figure 1. Overview of the new retrofit design methodology (see text for further explanation).

the existing process to identify the incentive for raw material and energy savings, (b) determine the incentive for completely replacing the existing plant, (c) screen the process options, and find the best flow sheet if the existing plant is to be completely replaced, (d) modify the existing equipment sizes for the existing flow sheet or a structural alternative, and (e) refine the retrofit calculations. One element in the analysis of the existing process is that in the operating cost diagram all the significant operating costs are attached to stream arrows on the current process flowsheet. In this paper, a new systematic retrofit design method for screening, identifying, and evaluating retrofit options of a chemical process is presented. In a previous paper,1 we described in detail the aspects of screening and identifying retrofit options. This paper summarizes the complete retrofit design method and covers in detail the aspects of analyzing the generated retrofit options, optimizing the process first with regard to those retrofit options (optimization parameters) that do not require investment, and evaluating the feasibility as well as the profitability of retrofit options that require investment. The complete method is demonstrated using a case study from the fine chemical industry. Methodology of the Retrofit Design Problem. The new systematic retrofit design method described in this paper aims at improving the production costefficiency of a continuous process and supports decision making when selecting among the most profitable options. A flowchart that highlights the important steps of the method is given in Figure 1. In a previous paper,1 the screening part of the new retrofit design method was already described in detail (Steps 1-3, Figure 1). The method follows an evolutionary approach and combines process insights, process-specific knowledge, and general engineering practice with mathematical analysis. It comprises five steps organized in three phases: (1)

analysis of the base case, (2) generation of retrofit options, and (3) evaluation of the generated retrofit options. In the first phase (Steps 1 and 2) detailed mass and energy balances of a characteristic operational steady state of the investigated process are established either by direct measurements in the existing plant or by process flowsheet simulation. The process is visualized using graph theory and information from the mass and energy balances is attached to the vertices and edges of the resulting process graph. In Step 1, the process graph is decomposed into open and cycle component path flows with their respective flow-rates. The flowrate of each component path flow is then calculated. In Step 2, each component path flow is evaluated with an indicator framework that includes three indicators to measure economic performance (material-value added (MVA) and energy & waste costs (EWC) as well as the sum of these two expressed as total-value added (TVA)) and two indicators to measure physicochemical properties (reaction quality (RQ) and accumulation factor (AF)). The indicator RQ measures the effect a component path flow may have on the reactions that occur in its path. If the RQ value is positive, the path flow has a positive effect on the overall plant productivity. Negative values indicate an undesirably located component path flow in the process. The indicator AF is a way of measuring the accumulative behavior of individual components in recycles. The definition of these indicators is given in Uerdingen et al.1 In Step 3, which corresponds to the second phase, the most important component path flows from an economic perspective are systematically investigated using a list of generic retrofit actions that aim at an economic improvement of the process through means of retrofitting. The physicochemical indicators are thereby used to preselect appropriate generic retrofit actions from the complete list of actions. Applicable retrofit actions are selected for each component path flow. This procedure leads to the identification of important optimization parameters and to the generation of structural retrofit alternatives, both referred to as retrofit options, to realize the desired retrofit actions. Last, the magnitude of the impact on the variable process costs is evaluated in a first approximation for each resulting retrofit option by calculating its total cost impact potentials (see Uerdingen et al.1). The focus of this paper is on the third phase of the methodology that comprises Steps 4 and 5. In Step 4, the identified optimization parameters are investigated. Figure 2 shows a more detailed overview of the substeps followed in Step 4. In Step 4-1 a local sensitivity analysis is carried out by means of rigorous process

1844

Ind. Eng. Chem. Res., Vol. 44, No. 6, 2005

shows potential for economic profitability. If economic profitability is probable the technical implementation scenario has to be evaluated vis-a`-vis a number of important criteria such as compliance to environmental restrictions, process safety, plant space requirements, and others (Step 5-4). If the scenario does not comply or cannot be modified to comply with all of these criteria the scenario is rejected. Finally, in Step 5-5 the technical implementation scenario is studied in detail in the modified process flowsheeting simulation and optimized with regard to the variable process costs. This information is ultimately used to calculate the profitability of the scenario in detail with appropriate economic profitability measures. On the basis of these results, the decision maker can select the most profitable options satisfying the company-own minimum profitability criteria and including all other business risk factors. Fine Chemical Case Study. The presented fine chemical case study from a current industrial example is used to demonstrate the capabilities of the retrofit design method. Fine Chemical Process Description. The process flowsheet for the manufacture of the fine chemical product is shown in Figure 4. Reactant R1 is fed to a storage tank M1 to which mainly recycled reactant R2 is recovered. The mixture is then fed to reactor RK1 where an equilibrium reaction according to the equation

ν11R1 + ν12R2 a ν13I + ν14CP Figure 3. Detailed description of Step 5 of the retrofit design methodology.

flowsheeting simulation with the identified optimization parameters. In Step 4-2 the variable process costs are minimized in a parameter optimization by manipulating the most cost-sensitive optimization parameters. Process constraints encountered during local sensitivity analysis and parameter optimization are finally used to generate additional structural retrofit alternatives (Step 4-3). Once the process is optimized with respect to the identified optimization parameters, in Step 5 the method continues with the evaluation of the generated structural retrofit alternatives. Again the variable process costs are used as objective function. A more detailed flowchart on Step 5 of the method is given in Figure 3. In Step 5-1, attainable variable process cost savings are calculated by means of rigorous process simulation for all structural retrofit alternatives with regard to the previously optimized process as a benchmark. This step makes use of the capabilities of process flowsheeting simulators for setting up process conditions in a manner that operates outside the range of operations of the investigated process plant (e.g., outside equipment capacity constraints and actual selectivities obtained in the plant). In Step 5-2 the alternatives that incur the highest cost savings are selected and detailed technical implementation scenarios are formulated on the basis of general engineering knowledge and experience. For economic reasons, existing and decommissioned equipment will need to be reused as much as possible to reduce investment costs for new equipment. In Step 5-3, a preliminary investment cost study for each generated scenario (only order-of-magnitude estimation of investment costs) is carried out and the scenario that incurs the least investment costs is selected. Further evaluation steps should only be undertaken if this scenario

(1)

occurs to an intermediate product I and a coupled product CP. The reactor effluent is processed through a first distillation column (D1) via another storage tank M2 to remove coupled product CP through the bottom. Because R2 acts as an intermediate key in the distillation the bottom product essentially consists of a CP/ R2 mixture. The overhead product of column D1 contains all unconverted reactant R1 from the feed, the remaining reactant R2, and all of intermediate product I. In M2 intermediate product I from a different production plant is added to the distillate of column D1. The mixture is then partially converted to product P (reaction eq 2) in reactor RK2 as follows:

ν21I f ν22P + ν23R2

(2)

ν31I f ν32R1 + ν33B

(3)

Reaction eq 3 denotes the main side-reaction of intermediate product I to reactant R1 and byproduct B. The reactor output is further processed in a second distillation column (D2) to preseparate product P from reactants R1 and R2. Since low concentrations of R1 and R2 still remain in the distillate of column D2, product P is finally purified by water counter-current extraction in column EX. The water-phase is recycled to the process to distillation column D3. The bottom product of column D2, containing mainly reactants R1 and R2, is recycled to storage tank M1. The bottom product of column D1 is fed to column D3 where coupled product CP from reactor RK1 and water from extraction column EX are also separated from R2 and eliminated from the process. The main component in the distillate of column D3, reactant R2, is equally recycled to storage tank M1. Various heat integration measures lower the overall energy demand of the process. Since reactor RK1 operates at subzero temperatures heat exchanger H1 partially recovers cooling energy supplied to the reactor.

Ind. Eng. Chem. Res., Vol. 44, No. 6, 2005 1845

Figure 4. Process flowsheet of the fine chemical case study (see text for explanation).

Reactor RK2 instead operates at high temperatures. Subsequently, heat supplied to reactor RK2 is partially recovered in heat exchanger H7 and used to preheat its feed. The vapor overheads of distillation column D3 are used to heat the reboilers of distillation columns D1 and D2. Additional heating demand in both columns D1 and D2 are supplied by independent reboilers (H5 and H11). Finally, the total heating demand of column D3 is delivered by reboiler H16. Using graph theory, a chemical process flowsheet can be alternatively represented as a directed process graph. In this case, the process flowsheet is transformed into a state-equipment network (SEN21), a subkind of directed process graphs, where multiple unit-operations or equipment units of a process flowsheet can be represented by a single vertex of the process graph. Figure 5 shows the corresponding directed process graph for the fine chemical case study. Vertex MI1 represents storage tank M1 and pump P1, vertex RK1 represents heat exchangers H1, H2 and reactor RK1, vertex DI represents heat exchanger H3, distillation column D1, reboilers H4, H5, and condenser H6. Vertex MI2 represents storage tank M2, vertex RK2 represents heat exchangers H7, H8, H9, and reactor RK2, vertex DP represents heat exchanger H10, distillation column D2, reboilers H11, H12, and condenser H15, vertex EX represents heat exchanger H14 and extraction column EX, and finally vertex DH represents storage tank M3, pump P2, heat exchangers H15, H18, reboiler H16, distillation column D3, and condenser H17. A rigorous simulation model of the fine chemical process was established using the commercial flowsheeting software Aspen Plus. The distillation columns, including reboilers and condensers, as well as the extraction column of the process (Figure 3) are modeled with the RADFRAC model and make use of

Figure 5. Process graph of the fine chemical case study (vertices represent groups of unit-operations in the process flowsheet of Figure 4; see text for further explanation). Bold arrows denote demand and supply flows as defined in graph theory. The edges between the vertices correspond to material flows in the process flowsheet between the groups of unit-operations.

the UNIQUAC model for calculating the vapor-liquid equilibria. The heaters and coolers are calculated with the HEATER model, whereas the HEATX model represents the process heat exchangers. Reactor RK1 is modeled as a CSTR model with kinetics for the main reactions. Unfortunately, no reaction kinetics are available for the main and side reaction in reactor RK2 (R eq 2 and 3). Thus, the reactor needs to be represented by a nonrigorous RSTOIC model with constant conversion and selectivities to byproduct B. The heat integration measures between the distillation columns (see Figure 3) are included in the model as well. On the basis

1846

Ind. Eng. Chem. Res., Vol. 44, No. 6, 2005

Table 1. Component Path Flow Decomposition, Assessment Result, and Ranking Obtained for the Fine Chemical Case Study patha

mass flowb [kg/kg]

EWCc [kUS$/a]

MVAc [kUS$/a]

TVAc [kUS$/a]

Category 1: Open Path Flows, RQ e 0 1.51 × 101 0 3.10 × 102 0

1 90

-150 -1

-151 -91

Category 2: Open Path Flows, RQ > 0 2.01 × 101 0.50 4.98 × 102 0.98

8 128

-202 0

-210 -128

826 527 254 108

0 0 0 0

-826 -527 -254 -108

label

component

RQ [-]

AF [-]

O1 O2

B H2O

sRK2,ir-dRK2,or sEX,ip-dDH,op

O3 O4

R2 I

sRK2,ir-dDH,op sRK1,ir-dRK2,or

C1 C2 C3 C4

R2 R2 R2 R1

Category 4: Cycle Path Flows, RQ > 0 MI1-DI-DH-MI1 5.12 × 102 0.50 5.9 × 10-1 MI1-DI-DP-MI1 7.73 × 102 0.50 1.3 × 100 MI1-DI-DP-DH-EX-MI1 7.48 × 101 0.50 4.6 × 10-2 MI1-DI-DP-MI1 3.34 × 102 0.50 4.5 × 101

a The abbreviations refer to the vertices of the process graph in Figure 5. b All mass flow-rates denoted in this column are relative to the smallest component path flow-rate of the process; the values are therefore dimensionless. c The economic values in these columns are all scaled by a confidential factor.

of plant data, a representative base case steady-state was calculated. The resulting mass and energy balances build the basis for the application of the new retrofit design method. Steps 1 & 2: Component Path Flow Decomposition, Assessment Results, and Ranking. Table 1 shows the results of the path flow decomposition, assessment procedure, and ranking applied to the fine chemical case study. As already mentioned, a detailed description of the flow decomposition and assessment methodology is given in Uerdingen et al.1 The path flows are sorted according to ascending TVA values in each path flow category. For demonstration purposes only component path flows with total cost savings potentials (TVA values) lower than -50 kUS$/a are considered in the following analysis. In Category 1 of Table 1, only two open path flows with neutral RQ values remain after the ranking and cutoff procedure, while RQ values of 0.5 and 0.98 are calculated for the open path flows in Category 2 of Table 1. Although open path flows O1 and O3 have a low flowrate and low EWC values, they nevertheless cause raw material losses while generating little or no output value (negative MVA values) and therefore account for negative TVA values. In turn, open path flows O2 and O4 with considerably higher flow-rates, show negative TVA values primarily due to their high EWC values. The four cycle path flows in Category 4 generate high EWC values where especially the three reactant R2 cycle path flows (C1, C2, C3) have an important impact on the variable process costs. Albeit a relatively low flow-rate of R2, cycle path flow C3 however shows the highest specific EWC value in this category (ratio of EWC value per mass flow-rate of a given component path flow), closely followed by cycle path flow C1. All four path flows show the same positive RQ values, while reactant R1 in cycle path flow C4 exhibits by far the highest AF value (AF ) 45). Step 3: Total Cost Impact Potentials of Identified Optimization Parameters and Structural Retrofit Alternatives. Steps 3-1 & 3-2: Identification of Optimization Parameters and Structural Retrofit Alternatives. Using the generic retrofit actions given in Uerdingen et al.,1 various optimization parameters and structural retrofit alternatives can be identified. As all of the identified optimization parameters and structural retrofit alternatives would require a lengthy discussion, only selected examples are presented in Tables 2 and 3 and are subsequently discussed.

The reflux ratio of distillation column DH could be manipulated to vary the R2 content in the bottom product such that reactant R2 losses are minimized. At the same time, the distillate rate of column DH needs to be adjusted accordingly to keep the content of coupled product CP in the distillate constant (parameter P1, Table 2). Alternatively, better column internals could be introduced to improve the separation efficiency (alternative A1, Table 3). Since the occurrence of byproduct B is undesirable in reactor RK2, open path flow O1 exhibits a negative TVA value. The pressure and feed-temperature of reactor RK2 could be selected as possible optimization parameters (parameters P2 and P3, Table 2). Open path flow O2 (water) also displays a negative TVA value in Table 1. To reduce its flow-rate the following process parameters could be manipulated: (i) the freshwater feed-rate to extraction column EX could be manipulated, and (ii) the reflux ratio of distillation column DP could be manipulated to vary the reactant R1 recovery in the distillate while the product P recovery is kept constant in the distillate (parameters P4 and P5, Table 2). Cycle path flows C1, C2, C3, and C4 exhibit high EWC values due to their high flow-rates. Multiple solutions can be formulated that aim at increasing the low equilibrium conversion of the main reaction (eq 1) in reactor RK1. A manipulation of the coolant flow-rate to reactor RK1 (indirectly varies the outlet-temperature in reactor RK1) and of the feed ratio of reactants R1 and R2 (indirectly manipulated by varying the fresh R2 feed-rate to the process) is proposed (parameters P6 and P7, Table 2). Alternatively, two structural retrofit alternatives are found to increase the conversion: (i) the cooling system could be modified to cool at even lower temperatures or (ii) a different reactor type that reaches higher conversions (e.g., a different catalyst) could be introduced (alternatives A2 and A3, Table 3). The high specific EWC value of reactant R2 cycle path flow C1 indicates that rerouting a part of the flow-rate to cycle path flow C2, which shows a notably smaller specific EWC value, is a possible option. The reflux ratio of column DI can be manipulated such that the R2 recovery in the distillate varies in response. At the same time, the distillate rate of the column needs to be adjusted to keep the content of coupled product CP constant in the distillate as CP represents a poison for the solid-bed catalyst in reactor RK2 (parameter P8, Table 2).

Ind. Eng. Chem. Res., Vol. 44, No. 6, 2005 1847 Table 2. Results of Applying Step 3 of the New Retrofit Design Method to the Fine Chemical Case Study: Total Cost Impact Potentials of the Identified Optimization Parameters

label

optimization parametersa

P6

vary the outlet temperature in reactor RK1 (vary the coolant flow-rate) vary the reactant feed ratio to reactor RK1 (vary the R2 feed-rate sMI1,ip) vary the reflux ratio of distillation column DI to vary the R2 recovery in the distillate while keeping the CP content in the distillate constant vary the pressure of distillation column DI

P7 P8

P9 P1

P5

P3 P2 P4

vary the reflux ratio of distillation column DH to vary the R2 content in the bottom product while keeping the CP content in the distillate constant vary the reflux ratio of distillation column DP to vary the R1 and R2 contents in the distillate while keeping the P recovery in the distillate constant vary the reactor RK2 output temperature vary the reactor RK2 pressure vary the freshwater flow-rate sEX,ip to extraction column EX

impact onb

energy & waste cost impact potential [kUS$/a]

material cost impact potential [kUS$/a]

total cost impact potential [kUS$/a]

C1, C2, C3, C4, othersd

1728

0

1728

C1, C2, C3, C4, othersd

1728

0

1728

O4, C1, C2

1481

0

1481

condens, heating & cooling duties of DIc O3, C1, othersd

1290

0

1290

834

212

1046

C2, C3, C4, othersd

890

29

919

O1, othersd, heating & cooling duties of RK2c O1, othersd O2, othersd

400

216

616

17 91

216 30

233 121

a The abbreviations refer to the vertices of the process graph in Figure 5. b Refers to the important energy and waste cost-sensitive component path flows and the important energy cost-sensitive unit operations. The component path flow abbreviations refer to Table 1. c Only the EWC value is impacted. d Component path flows not listed in Table 1 are impacted as well.

Table 3. Results of Applying Step 3 of the New Retrofit Design Method to the Fine Chemical Case Study: Total Cost Impact Potentials of the Generated Structural Retrofit Alternatives

label

structural retrofit alternativesa

impact onb

energy & waste cost impact potential [kUS$/a]

A2

improvement of reactor RK1 cooling system to increase the equilibrium conversion reduce pressure in distillation column DI below capacity constraint of reactor RK2 other column internals for higher separation efficiency in distillation column DH more selective catalyst for reactor RK2

C1, C2, C3, C4, othersd

1728

0

1728

condens, heating & cooling duties of DIc

1290

0

1290

O3, othersd, conden & cooling duties of DHc

54

212

266

O1, othersd

13

216

229

A4 A1 A3

material cost impact potential [kUS$/a]

total cost impact potential [kUS$/a]

a

The abbreviations refer to the vertices of the process graph in Figure 5. b The component path flow abbreviations refer to Table 1. c Only the EWC value is impacted. d Component path flows not listed in Table 1 are impacted as well.

Finally, since cycle path flows C1-C4 show high EWC values, a variation of the pressure in distillation column DI (parameter P9, Table 2) might positively affect the energy consumption of that unit-operation. Structural alternative A4 will be discussed below as a result of Step 4-3 of the new retrofit methodology. Steps 3-3 & 3-4: Total Cost Impact Potentials of Optimization Parameters and Structural Retrofit Alternatives. The energy cost impact potentials, material cost impact potentials, and total cost impact potentials were calculated for the identified optimization parameters and structural retrofit alternatives as described in Uerdingen et al.1 The optimization parameters and structural retrofit alternatives are sorted by descending total cost impact potentials in Tables 2 and 3. The optimization parameters that score the highest total cost impact potentials in Table 2 are the outlet temperature in reactor RK1 (parameter P6) and the

fresh reactant R2 feed-rate (parameter P7). The outlet temperature of RK1, manipulated through the coolant flow-rate, determines the maximal possible conversion according to the equilibrium reaction (R eq 1). Therefore, it impacts on reactant R2 cycle path flows C1, C2, C3, reactant R1 cycle path flow C4, and a path flow of inert components not listed in Table 1. Parameter P7 has a similar effect on the aforementioned component path flows as it influences the feed-ratio of reactants R1 and R2 to reactor RK1 and consequently also acts on its reaction conversion. The same total cost impact potential is calculated for alternative A2 (Table 3) as parameter P6 and A2 aim on the same effect on reaction conversion. The variation of the R2 reactant recovery in the distillate of distillation column DI (parameter P8) is expected to impact mainly on reactant R2 cycle path flows C1 and C2. Probably the recovery of intermediate I in the distillate might be altered as well, if the R2 split

1848

Ind. Eng. Chem. Res., Vol. 44, No. 6, 2005

between cycle path flows C1 and C2 changes. Therefore, the intermediate product I open path flow O4 is taken into account in the total cost impact potential. Impacts on other component path flows are expected to be of a much lesser degree and are not included in the calculated total cost impact potential. A variation of the pressure in distillation column DI (parameter P9) mainly affects its energy consumption with probably little effect on involved component path flows if the separation efficiency is held constant. The total cost impact potential thus includes the utility costs associated to the reboiler, condenser, and bottomproduct after-cooler. A variation of parameter P1 (Table 2) mainly influences reactant R2 open path flows O3, reactant R2 cycle path flow C1, and other component path flows not listed in Table 1. A tradeoff between energy costs for recycling R2 to reactor RK1 and raw material costs for fresh makeup R2 to storage tank MI1 is expected. A similar variation in distillation column DP (parameter P5) mainly impacts reactant R2 cycle path flows C2, C3, reactant R1 cycle path flow C4, and other component path flows. Another cost tradeoff between costs for recycling and costs for separation efficiency (reflux ratio) is also expected here. Varying the outlet-temperature and pressure of reactor RK2 (parameters P2 and P3) should mostly influence the formation of byproduct B in open path flow O1 and other component path flows not listed in Table 1. Almost the same reasoning can be followed in calculating the total cost impact potential of alternative A3 (Table 3). The energy & waste cost impact potentials differ only slightly. Additionally, the outlet temperature variation will impact the external heating and cooling requirements for reactor RK2. This further increases the total cost impact potential of parameter P3. Finally, varying the freshwater flow-rate for extractor column EX (parameter P4) is expected to affect water open path flow O2 and another component path flow not mentioned in Table 1. As reactant R2 is very soluble in water the performance of the extraction will probably remain largely unaffected. Using more efficient column internals in distillation column DH (alternative A1) mainly taps the TVA values of reactant R2 open path flow O3 and other component path flows not included in Table 1. It also taps the energy costs due to condensing the unused vapor distillate (heat integration) and the costs for cooling the bottom product of distillation column DH. Step 4: Sensitivity Analysis, Process Optimization, and Generation of New Structural Retrofit Alternatives. Step 4-1: Sensitivity Analysis. In Step 4-1 of the retrofit design method the optimization parameters identified in Step 3 are studied in a local sensitivity analysis (ceteris paribus). As mentioned before the RK2 reactor pressure and temperature (parameters P2 and P3, Table 2) cannot be varied because no reaction kinetics are available for the main and side reactions (R eq 2 & 3). The optimization parameters are varied by a maximum of (40% as compared to their base case values, while the production rate of product P is kept constant by a design specification. Most parameters are however varied in a smaller range because process constraints are either encountered earlier or costs increase dramatically. For each optimization parameter approximately 6-8 values above or below its base case value

Figure 6. Effect of the variation of the R1/R2 reactant feed mass ratio of reactor RK1 on the variable process costs relative to the base case.

are simulated (perturbation steps). For exemplary purposes only two optimization parameters (parameters P7 and P9) are shown and discussed individually in the following. The abbreviations used for the unit-operations always refer to the process graph in Figure 5. The R1/R2 reactant feed mass-ratio at the inlet of reactor RK1 (parameter P7) influences the equilibrium conversion of the main reaction (eq 1). This ratio is directly controlled by fresh reactant R2 feed to the process. A higher fresh reactant feed decreases the ratio whereas a lower feed increases it. The cost sensitivity of this parameter variation is shown in Figure 6. The fresh R2 reactant feed-rate is varied from -10 to +4% of its initial base case value causing the R1/R2 feed mass-ratio to vary from -16 to +38% relative to its base case value. In the direction of an increasing ratio, the reactor conversion first increases to a maximum and then decreases again. In the opposite direction the conversion only decreases over the investigated interval. Due to the specification of constant productivity, the flow-rates of both R1 and R2 vary in dependence of conversion. The resulting total costs vary considerably between -7 to +15% of the base case costs. The total cost curve itself indicates a cost minimum at the reactant ratio of +17% and exhibits a strongly nonlinear proportionality. The total costs are mostly influenced by the steam costs that result from changing utility requirements in distillation columns DH and DI. Unexpectedly, the change in raw material costs is negligible due to a strong sensitivity of the process behavior to the R2 reactant feed-rate. In particular, ratios of R1/ R2 higher than +17% relative to its base case value and the resulting decrease in conversion lead to unfavorable process conditions, i.e., a more difficult separation in column D1 which results not only in higher utility costs but also an increased recycling of the product P. Varying the pressure in distillation column DI (parameter P9) affects its separation efficiency. The optimization parameter is varied from -20 to +15% of the base case value, which causes the total costs to vary from -3.5 to +2.8%. The resulting cost curves are shown in Figure 7. The total and the steam cost curves exhibit a linear proportionality to the pressure in the investigated interval. The total costs are mostly influenced by the steam costs. The sensitivity of cooling water, raw material, waste, and electricity costs is comparatively weak. As expected, a pressure reduction decreases not only the fresh steam consumption (heat supply through heat

Ind. Eng. Chem. Res., Vol. 44, No. 6, 2005 1849

Figure 7. Effect of the variation of pressure in distillation column DI on the variable process costs relative to the base case.

integration remains constant) in the reboiler of distillation column DI but also the necessary cooling duties. The pressure variation shows almost no effect on the energy requirements of other unit-operations. The “constraint” label in Figure 7 indicates a known process constraint due to equipment capacity limitations in reactor RK2 (conversion along with residence time). The sensitivity analysis shows that the most important cost sensitivities essentially result from changing steam requirements in the reboilers of the three distillation columns DI, DH, and DP. In almost all parameter variations the raw material, waste, electricity, and cooling cost sensitivities are far below the steam cost sensitivities. Table 4 shows the type of proportionality of the total costs to the varied optimization parameter, the base case values of each optimization parameter, and the nearest process constraint (equipment capacity or quality requirement) for each studied optimization parameter. Especially the variation of parameters P6, P7, and P9 exhibit important total cost sensitivities and result in cost savings if their respective values are lowered. Parameter P6 cannot be lowered below approximately -0.2%, because the cooling system itself operates near its minimum cooling temperature.

Parameter P9 can only be lowered to a pressure minimum of -15% relative to the base case value according to the capacity constraint discussed above. Step 4-2: Parameter Optimization. The optimization is carried out with the SQP algorithm (successive quadratic programming) that is integrated in the flowsheeting software. As for the sensitivity analysis, the variable process costs are used as objective function. The variation range of the manipulated parameters in the parameter optimization is fixed according to the results of the local sensitivity analysis. The base case values of the manipulated parameters are used to initialize the optimization. A number of optimization runs is then performed with modified initialization values to increase the chances of finding the minimum variable process costs, although it cannot be said if the calculated optimal solution coincides with the global optimum. The optimized values of the parameters are compared to their base case values in Table 4. If an optimization parameter is indirectly varied by manipulation through another process parameter the table also shows its base case and optimized values. As expected the optimized value for the RK1 reactor outlet temperature (parameter P6) corresponds to the minimum possible outlet temperature (-0.2%, see Table 4). Surprisingly, the cooling duty in the cooling system drops by 6%, although the outlet temperature actually decreases. This occurs because the total flow-rate at the entrance of reactor RK1 decreases as a result from other parameter changes and therefore less cooling duty is needed to reach the minimum cooling temperature difference. The reactant feed mass-ratio to reactor RK1 (parameter P7) increases by roughly 25%, which massively reduces the excess of reactant R2 at the reactor entrance by reducing the fresh R2 reactant feed-rate by 16% accordingly. The R2 reactant recovery in the overheads of the distillation column DI (parameter P8) is decreased by 5%, which is achieved by increasing the column mass reflux ratio by 16%. As expected from the sensitivity analysis, the pressure in distillation column DI is optimally set to the pressure constraint of -15% of the initial base case value. The column mass reflux ratio of distillation column DH (parameter P1) is only

Table 4. Sensitivity and Optimization Results of the Identified Optimization Parameters

label

optimization parametersa

P6

vary the outlet temperature in reactor RK1 (vary the coolant flow-rate) vary the reactant feed ratio to reactor RK1 (vary the R2 feed-rate sMI1,ip) vary the reflux ratio of distillation column DI to vary the R2 recovery in the distillate while keeping the CP content in the distillate constant vary the pressure of distillation column DI vary the reflux ratio of distillation column DH to vary the R2 content in the bottom product while keeping the CP content in the distillate constant vary the reflux ratio of distillation column DP to vary the R1 and R2 contents in the distillate while keeping the P recovery in the distillate constant vary the freshwater flow-rate sEX,ip to extraction column EX

P7 P8

P9 P1

P5

P4

total cost proportionalityb almost linear

constraintc

optimized value (relative to base case)d