pinch anaylysis of cooling water

Upload: garywubo

Post on 04-Apr-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/29/2019 pinch anaylysis of cooling water

    1/13

    Published: October 06, 2011

    r 2011 American Chemical Society 12067 dx.doi.org/10.1021/ie200722z| Ind. Eng. Chem. Res. 2011, 50, 1206712079

    ARTICLE

    pubs.acs.org/IECR

    Simulation-Based Process Design and Integration for the SustainableRetrofit of Chemical Processes

    Aurora Hernandez Enrquez, Martin Tanco, and Jin-Kuk Kim*,

    Centre for Process Integration, School of Chemical Engineering and Analytical Science, The University of Manchester,Manchester M13 9PL, U.K.CITEM, Universidad de Montevideo, Luis P. Ponce 1307, Montevideo, UruguayDepartment of Chemical Engineering, Hanyang University, 17 Haengdang-dong, Seongdong-gu, Seoul, Republic of Korea, 133-791

    ABSTRACT: This research proposes a novel Retrofit Design Approach based on process simulation and the Response SurfaceMethodology (RSM). The approach comprises a diagnosis stage to select the promising variables through a sensitivity analysis, anevaluation stage to assess the impact of the promising variables, and to identify the most important factors through RSM. A reducedmodel from the process response behavior is built, and an optimization is carried out with the reduced model to identify optimalconditions and performance of the system, subject to objective function and model constraints. All these procedures are simulation-

    supported. The main advantage of the proposed approach is to handle a large industrial-scale design problem within a reasonablecomputational effort and to obtain a reduced model based on the most important factors. Limitations for the developed methodinclude that theglobal optimality of thesolutions found is not fully guaranteed, while large computational time for simulation may berequired when the large number of factors and levels need to be considered, although this is offset by the reduced optimization time.The proposed Retrofit Design Approach has been applied to the NGL (natural gas liquids) recovery process, in which steady-stateprocess simulation using Aspen Plus TM has been carried out, and complex design interactions existed for retrofit scenarioshave been systematically evaluated, leading to optimal strategies for retrofitting through the proposed design method. Both thecontinuous and discrete design options are considered in the retrofit design, and the results showed that the approach is effective toprovide reliable, cost-effective solutions which yield to economic and environmental improvements in the studied processes. Thepromising sets of retrofit design options were presented as a portfolio of investment opportunities for supporting further decision-making procedures.

    1. INTRODUCTIONMore and more companies are looking for solutions to reduce

    their greenhouse gas emissions or even to make no net contribu-tion to global warming. Globalization has also encouraged theindustry to look at increasing profits, reducing environmentalimpacts, being safer, and developing a commitment to sustain-ability in order to be competitive. Therefore, chemical and processindustry has turned their attention toward finding opportunitiesfor energy savings, cost reductions, increasing quality standards,and eliminatingbottlenecks through acquiring new equipment orrevamping old processes. Simulation and optimization techni-ques have been widely utilized as a decision-making tool in theprocess industries. However, it is not often straightforward to

    build a robust simulation and optimization framework and applyit for obtaining realistic and practical engineering solutions, dueto a large number of variables to be considered simultaneouslyand, consequently, complex system interactions among them tobe screened and analyzed. Although a very rigorous modelingframework can be formulated to reflect all the decision variablesand constraints, the development of a model requires significanthuman and computational resources if this should be utilized inpractice. Also, finding optimal or near-optimal results from thecomplex mathematical model and its optimization demandssignificant computational efforts.

    From the viewpoint of optimization methodology, conven-tional methods using deterministic or stochastic techniques have

    been widely employed to identify optimal solutions from theformulated optimization model in various industrial applications.However, it is very difficult to obtain optimal solutions withinacceptable computational time and resources unless the mathe-matical formulation is relatively simple and the problem size isnot unreasonably large.

    In addition to that, in most cases there also exists a verycomplex trade-off between various issues or competing perfor-mance criteria. A considerable works had been done with advancedalgorithms to solve those multiobjective optimization problems.14

    However, it is still difficult to incorporate a realistic operationalmodel of a plant into a decision-support tool that includes notonly the intrinsic process issues but also the complementary issuespreviously mentioned (i.e., financial, environmental, safety, andreliability issues). Furthermore, computational difficulties arisewhen attempts are made to solve more complex problems.

    Hence, further research should be undertaken to explore theseissues in the context of site-wide process retrofit design, in orderto determine the most cost-effective and practical solutions andprovide a reliable design tool that can be applied to variousprocess industries. This is the main motivation of the present

    Received: April 6, 2011Accepted: September 21, 2011Revised: August 18, 2011

  • 7/29/2019 pinch anaylysis of cooling water

    2/13

    12068 dx.doi.org/10.1021/ie200722z |Ind. Eng. Chem. Res. 2011, 50, 1206712079

    Industrial & Engineering Chemistry Research ARTICLE

    work which aims to develop a reliable and practical design approachto generate cost-effective and environmental-friendly retrofitdesign options.

    2. DESIGN METHODOLOGY

    This paper addresses a new design methodology for retrofitscenario which is based on the application of process simulationand Response Surface Methodology (RSM) simultaneously,which are rarely used together in the field of chemical engineer-ing for retrofit studies.58 The new design method will treat theprocess simulation as physical experiments and will be capable ofproducing reduced models that reproduce the techno-economicperformance of the system (named response in RSM nomen-clature) with an acceptable level of confidence. For this purpose,the most important factors will be first identified by the applica-tion of a screening design of experiments in combination with theprocess integration concepts.This will generate the knowledgeofthevariables that mostly affectthe process response. The reducedmodels can then be optimized with significantly reduced time, asthe form of the equations will be mostly quadratic and continuous

    in the parameters, yielding to the retrofi

    t design options. Besidesthis, environmental indexes can be inherently reflected in thesimulation results. The approach is based on three stages that aredescribed in the following section. Computational time andefforts for process optimization are considerably reduced byapplying the surface response methodology and optimizing thereduced model.

    The benefits of the proposed approach are as follows: To have the capacity to handle a large production plant, To generate a reliable retrofit design portfolio with econom-

    ic and environmental improvement, To require only a reasonable amount of time to reach

    pseudo-optimal solutions by optimizing reduced models, To quantify the effects of the most important factors toward

    the improvement which helps to understand the process, To be simple and generic enough to be applicable in thewide range of industrial applications.

    The limitations for the approach are as follows: Global optimality is not guaranteed. The computational time for simulation may be considerable

    in some cases, as a large number of simulations may be per-formed for sensitivity analysis and RSM, depending on thenumber of factors and details of study.

    The integer variables required for considering structuralchanges in a retrofit design may cause infeasibilities in theDoE applied.

    RSM is based on the work proposed by Box and Wilson 9 inwhich experimental design and procedures were introduced for

    determining a path of steepest ascent and for exploring maximaand ridges. Their purpose was to find optimal operating condi-tions in a chemical process, for example, the temperature, pre-ssure, and the concentrations of the various reactants which gavemaximum yield of the desired product. It was defined as a groupof mathematical and statistical techniques for analyzing problemsby investigating a response of interest influenced by variablesor factors to be optimized.10 Modeling can be performed byextracting quantitative data from a set of experiments with anappropriate experimental design and fitting them into mathema-tical equations.

    Designof experiments (DoE)has been used to understand theeffect of the changes to the input parameters through systematic

    variations. The statistics analysis of these experimental data helps inthe understanding of design problem and generates the modelsdescribing its system behavior and characteristics. The modelsgenerated are commonly called mechanistic or empirical models,because these are obtained directly from experiments. However,it is also possible to develop simplified or reduced models fromtheoretical models with the aidof simulations if these are taken as

    experiments (i.e., computer experiments). In this manner, DoEefficiently explores the system of interest and gains useful statisticalinformation, with reasonable computational time and resources.

    RSM is a sequential procedure that follows model generationthrough DoE and searches for its optimum, subject to the objectiveand constraints. This method has been widely used in variousareas including chemistry, biology, electronics, and manufactur-ing, in which its main applications are related with determiningthe factors and levels that satisfy a set of requested specificationsin the searching space, and determine the optimum combinationof factors at a desired response. This allows gaining conceptualinsights and achieving quantitative understanding of the systembehavior. The computer experiments have been attractive forsolving industrial-type design problems. DoE for mechanical

    processes, such as the computational fluid dynamics of turbinesthat evaluated main and joint effects of input parameters on theturbine, was studied to understand the influences of the radialvelocity offluid at the inlet on the pressure recovery and energyloss factor.11 Various studies1215 demonstrated that RSMapplied to computer experiments is a promising tool for optimiz-ing real systems.

    The main advantage for RSM is its ability to gain under-standing of the main factors and interactions that affect thestudied response, which is used for the generation of the reducedprocess model. Thissimplifies the optimization procedure, whichreduces the computational efforts and time required for processdesign and optimization. Therefore, RSM can be used as analternative optimization methodology to reach pseudo-optimal

    solutions with a high statistical confidence for its results. Addi-tionally, it has been favored for the design of experiments in theindustry because of its promising cost, plant performance timereductions, and obtaining the effective and reliable results.57,16

    With these considerations and benefits, RSM, asa design methodfor retrofitting study, has been selected in this research.

    2.1. The Proposed Retrofit Design Approach. Figure 1 showsthe schematic diagram for the procedure to carry on the proposedRetrofit Design Approach. Three main stages explained below aresupported by simulation software.

    Step 1. The Diagnosis Stage. Variables to be changed orstructural modifications having a potential for improvement inthe process are identified. The improvement is measured by theincreases in a techno-economic performance index, for example,

    profit. Various indices, or objective functions, can be used accordingto evaluators needs. In this study, the economic profit has beenselected as a measure to evaluate the system performance. Thecosts, which are not heavily dependent on the system perfor-mance such as maintenance and salary, are not explicitly con-sidered in this study. Three types of profit, namely, net profit(NPr), marginal profit (MaPr), and normalized marginal profit(MaPr*), are defined as

    NPr SP SCO-VCRM-VCE 1

    where NPr = net profit [y1], SP = profit from the sales ofproduct [y1], SCO = profit from the sales of coproduct

  • 7/29/2019 pinch anaylysis of cooling water

    3/13

    12069 dx.doi.org/10.1021/ie200722z |Ind. Eng. Chem. Res. 2011, 50, 1206712079

    Industrial & Engineering Chemistry Research ARTICLE

    [y1],VCRM = cost of rawmaterial [y1],andVCE = cost of

    energy [y1]

    MaPr NPrSC-NPrHBC 2

    where MaPr = marginal profit [y1], NPrSC = net profit of

    studied case [

    y

    1

    ], and NPrHBC = net profi

    t of historical bestcase [y1]

    MaPr MaPr

    MaPrBC3

    where MaPr* = marginal profit normalized [-], MaPr = marginalprofit of studied case [y1], and MaPrBC = marginal profit ofbase case [y1].

    The main elements in the diagnosis stage are as follows:Selection of Key Design Variables. This is done by exploring

    all the controllable design parameters of the plant within theallowed operating range or a plant capacity and investigating itsimpacts on the performance of the plant. Structural changesavailable for debottlenecking in the plant are also considered for

    the assessment.Conceptual Understanding with the Aid of Process Inte-

    gration ( PI) Techniques. The promising options for operationalchanges or structural modifications are considered. These maybe continuous variables such as stream splits, changing existingequipment bypasses, modifications of heat exchange duty, etc. ordiscrete variables such as adding or eliminating equipment, relocat-ing or modifying existing equipment internally, etc. The feasi-bility of these structural modifications in practice, based on theusers experience and knowledge, as well as practical constraints(e.g., plant layout), must be considered. The evaluation of potentialoptions is carried out with process design methodologies, includingprocess integration methods.1719 For example, different options

    for the distillation sequencing, as retrofit options, are reviewed inthis stage. The component flow rates through each column, incombination with the column composition profile, are checkedto examine mixing effects, the feeding stage or the product extra-ction stage location, the feeding condition, and the number ofstages or their overall efficiency. Complex column arrangements,

    such as side-stream, prefractionator, and dividing wall column,are also included for testing and to find out whether they canproduce a considerable improvement in the existing process.These process integration options are useful for performanceevaluation because they can systematically identify realistic poten-tials fromoperational or structuralchangesthat can improveprocessperformance in product recoveries and energy recovery.2022

    The impact of each variable identified in the previous sectionsis assessed by a sensitivity analysis to select the most promisingvariable. It has been stated that this class of analysis, which iscommonly known as a one-factor-at-a-time (OFAT) analysis, isnot adequate for screening factors, as it does not considerinteractions between factors that may be important for somecases.10,23 It must be clarified that the sensitivity analysis carried

    out in this part is only performed as a tool to gain the necessaryknowledge of the process at this first diagnosis stage and, thus, toeliminate nonpromising parameters or structures. Althoughinteractions between factors may not be fully acknowledgedthrough sensitivity analysis, decision is made such that importantfactors showing significant effects on the performance index areselected at the discretion of users decision. This sensitivityanalysis is carried out by executing perturbations to the variablestudied and assessing the effect that these perturbations have inthe response of interest (e.g., profit).

    Following on from this sensitivity analysis, variables that donot show effect on the performance index of the process are notconsideredfurther in theRetrofit DesignApproach,at thediscretion

    Figure 1. The proposed Retrofit Design Approach.

  • 7/29/2019 pinch anaylysis of cooling water

    4/13

    12070 dx.doi.org/10.1021/ie200722z |Ind. Eng. Chem. Res. 2011, 50, 1206712079

    Industrial & Engineering Chemistry Research ARTICLE

    of users decision. The remaining set of promising variables isthen transferred to the evaluation stage. For the Retrofit DesignApproach, the initial size of the problem (i.e., number of factorsto study) and the ranges in which these can be varied (i.e., thelevels) are defined.

    Step 2. The Evaluation Stage. In this stage, promising optionsselected from the diagnosis stage are further investigated. A

    screening DoE is applied to indentify the most important factorsfrom the promising setof variables. Further to theselection of themost importantfactors, a reduced model is obtained by fitting theprocess response behavior. Additional simulations can be exe-cuted further to account for a complementary DoE that accu-rately fits the surface model. The general procedure for theevaluation stage can be divided into twoparts, detailed as follows:

    Preliminary screeningconsists of a screening DoE, which is aFractional Factorial Design (FFD) at n levels with kfactors, where number of levels n can be any numberchosen by the user for all the factors and the k factors areones specified from the diagnosis stage. In practice, it isrecommended to set this FFD at two levels (n = 2) due to itssimplicity and relatively a low number of simulation runs

    needed when compared with a full factorial design (FD).10The level of resolution, which common values are III, IV, orV and, thus, the confounding pattern are selected, based onboth, the number of factors and the minimum number ofsimulation runs needed for a simple andclear factor analysis.This first FFD will result in a number of computer experi-ments (process simulations) to be carried out and thesubsequent estimation of the impact of factors in the response(e.g., profit). For this purpose, an Analysis of Variance(ANOVA) is applied to characterize simulations responses,andthe finalresult is theidentificationof the most importantfactors with which the surface model will be fitted.

    RSM is applied in the second part of the evaluation stage.The methodology generates the input parameters, based onthe most important factors identified, and produces theresults of itsperturbations in the objective response from thesimulation. When discrete parameters are present in the setof most important factors, infeasibility related with discretevariables in RSM can be avoided by setting the discreteparameter as a continuous one in the continuous range inthe response surface design. For this step, the surface responsedesign proposed in this study is a Central Composite Design,as this can be built with the previous FFD simulations whichare supported with additional computer experiments (axialand central runs). This results in a fewer number ofsimulationsthan a new experimental design and, consequently, reducesthe computational time. The response from simulations are

    then fitted into a function through a Linear Least Squares(LLS) method,24 which provides a reduced model with ahigh level of confidence for reproducibility within the rangesstudied. The accuracy of the reduced model is validated withthe Root Mean Square Error (RSME) and residual plots.

    Step 3. The Optimization Stage. The reduced model derivedfrom the RSM is optimized for obtaining the optimal values ofthe response withinthe ranges of parameters studied. In this step,similar to the evaluation stage, the optimization model does notinvolve binaryvariables, and newequipmentor facility tobe added isrepresented as a continuous variable, which means optimal valueof these variables indicates the capacity or size for the new unit.As the models are reduced and have no binary variables, Linear

    Programming (LP) or Nonlinear Programming (NLP) solverscan be used for optimization. Optimization can be carried outwith deterministic optimization solvers, for example, the Micro-soft Excel Solver and GAMS or stochastic optimization methods,including Genetic Algorithms and Simulated Annealing. Thesemethods inherently generate local optimal solutions and, there-fore, global optimization techniques, for example, BARON, may

    be considered for achieving global optimality of the solution. Theoptimization problem in this work is not highly nonlinear andnonconvex; the deterministic method is used in the case study.

    One issue with the NLP solvers used in optimization is that itoften results in a local optimal solution. One of the ways not totrap local optima is to apply conceptual knowledge of the problemor, alternatively, to carry out multiple runs of optimization withdifferent starting points, which different optimal solutions can bethen generated.

    The capital investment is estimated for new units consideredin the study as the annualized capital cost (ACCNewUnits)

    ACCNewUnits CCNewUnit 3AF 4

    where ACCNewUnit = annualized capital cost of the new unit[y1], CCNewUnit = capital cost of the new unit (acquisitioncost plus piping cost plus installation cost) [], and AF =annualization factor [y1].

    Finally, the payback periods achieved with the proposedretrofit design options can be estimated as

    PP ACC 3 PLP

    MaPr5

    where PP = payback period [y], ACC = annualized capital cost[y1], PLP = project life period [y], and MaPr = MaPrimprovements [y1].

    Afinal retrofit portfolio from the Retrofit Design Approachincludes all the changes to be made and its techno-economic

    benefits. If capital investment is necessary, payback periods willbe used for evaluating economic impact resulted from newequipment to be introduced.

    3. CASE STUDY

    3.1. Process Descriptions. The proposed approach is appliedto the retrofit of natural gas liquid (NGL) recovery plant asshown in Figure 2. The extraction of hydrocarbon liquids (C2+)from natural gas (NG) coming from gas wells is required. Theseparation process uses expensive cooling from refrigerationcycles as well as expansion devices to liquefy the feed sweetenedgas (SG) stream and then separate the components by distilla-tion in a demethanizer column. Maximum design capacity is

    173 kg 3 s1 of feed, with a recovery of 75% for the ethane (C2) and99% for the propane (C3). Typical specifications of the feed withkey recoveries for methane and ethane, product requirements,cost indices of all available utilities, raw materials, and productsare listedin Tables 1 to 4. The inlet SG enters at 65 kg 3 cm

    2 and43 C and is then fed to a first precooling heat exchanger usingcooling water, in which the gas temperature is reduced to 35 C.The outlet gas is dried to less than 0.1 ppm of water content.Following this stage, the dried gas is chilled to about 10 C.From this first section, a liquid stream is extracted and sent to afirst separator tank, from which the liquid is sent to the deme-thanizer column and its vapors to a second cooling train. Next, afirst turbo-expander expands the gas from 60 to 37 kg 3 cm

    2,

  • 7/29/2019 pinch anaylysis of cooling water

    5/13

    12071 dx.doi.org/10.1021/ie200722z |Ind. Eng. Chem. Res. 2011, 50, 1206712079

    Industrial & Engineering Chemistry Research ARTICLE

    followed by a separator tank from which other feed stream tothe distillation column is obtained. A Joule-Thompson valve isset in parallel to the first turbo-expander to bypass the gasgoing through in case of failure. The outlet gas stream of thistank is sent to the third cooling train to reduce the tempera-ture of the gas about 69 C. This is then sent to the secondturbo-expander to reduce its pressure from 35 to 20 kg 3 cm

    2

    approximately.

    Figure 2. NGL process.

    Table 1. Feed Stream and Key Recoveries

    components molar fraction

    N2 0.0507

    CO2 0.0001

    CH4 0.7742

    C2H6 0.1000

    C3H8 0.0438

    n-C4 0.0143

    i-C4 0.0065

    n-C5 0.0042

    i-C5 0.0039

    C6+ 0.0021

    H2O 0

    flow rate, kg/s 149.27

    flow rate, m3/s 170.26

    T, C 45.5

    P, kg/cm2 64.8

    propane recovery, % 99.3

    ethane recovery, % 65

    Table 2. Product Specification

    product parameter unit limits

    cryogenic liquids (C2+) methane content % vol e0.8

    residual gas high pressure propane content % mole e0.2

    CO2 + N2 content % vol 1.43

    humidity ppm e112

    total sulfur ppm e200

    H2S ppm e4.4

    calorific value kJ/m3 g35,443

    outlet pressure kg/cm2 66.792 ( 0.2

    res idual gas low pressure humidity ppm e112

    H2S ppm e4.4

    Table 3. Available Utilities

    hot utilities available temperature, C cost, kW1y1

    fuel gas 120

    high pressure steam 450 379

    medium pressure steam 360 358

    low pressure steam 180 242

    hot water 90 33

    Cold Utilitiescooling water 25 25

    propane 45 472

    Powerelectricity 300

  • 7/29/2019 pinch anaylysis of cooling water

    6/13

    12072 dx.doi.org/10.1021/ie200722z |Ind. Eng. Chem. Res. 2011, 50, 1206712079

    Industrial & Engineering Chemistry Research ARTICLE

    A Joule-Thompson valve is also set in parallel to this turbo-expander. The outlet stream goes to the final separator tank,which is also known as the cold tank and its pressure is about20kg 3 cm

    2 and temperature of89 C. The top vapors of this coldtank are sent backto coola part ofthe inlet stream and then tothefuel gas network as residue gas low pressure (RGLP). Part of thisstream is sent through the coupled compressors with the expanderstoward the high-pressure compression section. The cold tankliquid stream is pumped to the first stage of the demethanizationcolumn. The column normally operates at 25 kg 3 cm

    2 and atemperature of79 C in the top; it has a reboiler which uses alow-pressure steam to provide the duty but does not have acondenser. The reflux ratio to the column is controlled by the topinlet liquid stream coming from the cold tank; the overhead

    vapor product, mainly methane (i.e., 93.5% mol CH4),isfedbackto the first and second turbo-expanders and then to a series ofgas-to-gas heat exchangers, in which it exchanges heat with theinlet gas stream. Following this stage, the gas is compressed byhigh pressure steam turbines to the desired pipeline pressure setby the end users (i.e., 70 kg 3 cm

    2) and then is heat-integratedwith the demethanizer column bottoms. The existing design ofthe heat recovery system in the entire unit is highly integrated,and the plant layout is too compact to accommodate new pipingor modifying existing equipment, no structural changes for heatrecovery systems, for example, adding new heat exchanger areas,repiping and resequencing, are considered in this particular case.

    To estimate the annualized capital cost of the equipmentrequired in the retrofit design, a ten-year project life with a 12%interest rate is assumed. The items considered in the calculationof net profit given in eq 1 (NPr) are profits from three products(residue gas low pressure, residue gas high pressure, and C2+product) and three byproducts (low pressure steam, mediumpressure steam and condensates), operating expenses of rawmaterial (sweetened gas) and energy-used (high pressure steam,cooling water, electricity), and penalty associated with nitrogencontent

    NPr SP SCO-VCRM-VCE N2Pe 6

    N2Pe SRGHP 3GCVRGHP

    GCVRGHP@5%N2

    !7

    where N2Pe = nitrogen penalty [y1], SRGHP = sales of RGHP

    [y1], GCVRGHP = gross calorific value of RGHP [kJkg1],

    and GCVRGHP@5%N2 = gross calorific value of RGHP at inletcontent of 5% mole of nitrogen [kJkg1].

    The level of nitrogen content in the feed reduces the grosscalorific value (GCV) for the final RGHP, and its selling price isbased on its caloric value. The penalty to the NPr related to thelevel of nitrogen is included and estimated as shown in eq 7.

    The objective function in this case study is set to maximizeMaPr. The number of annual working days for the plant isassumed to be 350 days per year, as the maintenance period is 30days for every 2 years. The capital cost for the new equipment in

    this case study was considered the Free-On-Board (FOB)investment cost, plus piping costs and associated arrangements,which is assumed to be 40% of the equipment cost.

    3.2. Process Simulation. The simulation of the plant wasperformed with Aspen Plus simulator 2006.5 SM. The PengRobinsonSoave (PRS) equation ofstate wasset forthecalculationof thermodynamic properties. The standard modules available in

    the Aspen Plus library were used, except the dehydration unit.However, to avoid hydrate formationproblems in the system, theassumption of having the inlet streamwater content at 0.1 ppm waskept. A number of assumptions were made based on the normaloperating strategy of the plant. The simulation results showeda good level of agreement with real operating data as shownin Table 5.

    The major operating cost of the plant is energy cost. Mechan-ical power demand for compressors and electricity requirementsfor pumps are provided from central utility systems throughsteam turbines and steam distribution networks, while steamfrom utility systems is required to provide heat in the reboiler ofdemethaniser column. Therefore, reducing steam and power de-mand is important to improve MaPr. Atmospheric emissions

    associated with energy usage of the plant are estimated by usingthe correlation factors given in EPA AP-4225 for uncontrolledfurnaces with fuel gas and duty less than 100 MMBtu 3 hr

    1, asfollows

    AirEi Fi 3 FrFG 3 GCVFG 8

    where AirEi = air emission of i [t i (kt hydrocarbon)1]; i

    includes SO2, NOx, CO2, and VOC, Fi = factor in EPA 42 forcompoundi [(t is) (kt hydrocarbonkj)1],FrFG = flowrate offuel gas [kgs1], and GCVFG = gross calorific value of fuel gas[kJkg1]

    SEC TEC

    FrSGP

    9

    where SEC = specific energy consumption [GJkg HC pro-cessed1], TEC = total energy consumption [GJY1], andFrSGP = flow rate of sweetened gas processed [kg HC pro-cessedY1].

    Table 5 presents these indexes, while water discharges andwaste indexes cannot be estimated because of the lack of specificdata for the plant. The EuropeanIntegratedPollution Preventionand Control Bureau (EIPPCB)26 proposed for these indexes andthe MexicanLawregulated limits27 are also presented as a reference.

    3.3. Application of the Proposed Retrofit Design Ap-proach. 3.3.1. Diagnosis Stage. a). Selection of Key DesignVariables.All 25 independent variables of the plant were studied,plus the nitrogen componentin the inlet gas. The impact of these

    variables for the improvement in the response MaPr (objectivefunction) was assessed by the sensitivity analysis. The variableswere ranged along the safe operating limits. The analysis identifiedthree parameters and the inlet nitrogen composition, which arelistedin Table 6. The rest of theother23 variables showed almostno changes on MaPr. The proposed approach continues thenwith the second part of the diagnosis stage, in which the promisingoptions for operational changes or structural modifications basedon process integration options are explored.

    b). Conceptual Understanding of Retrofitting. Options forfurther considerations were feed location, the number of stagesand their efficiency, adding or adjusting pumparounds in thedemethanizer, and adding or adjustingpower generation capacity

    Table 4. Raw Material and Products Unit Costs

    component type unit cost [/m3]

    sweetened gas (SG) raw material gas phase 0.134

    cryogenic liquids (C2+) product liquid phase 83.6

    residual gas (RG) product gas phase 0.109

  • 7/29/2019 pinch anaylysis of cooling water

    7/13

    12073 dx.doi.org/10.1021/ie200722z |Ind. Eng. Chem. Res. 2011, 50, 1206712079

    Industrial & Engineering Chemistry Research ARTICLE

    in the turbo-expanders. A good level of knowledge about theprocess was gained after performing the first part of the diagnosisstage, from which a first group of factors were selected. Toidentify the effects of these factors and their interactions, a total

    of twelve factors, as listed in Table 7, are studied in the evaluationstage for MaPr.

    3.3.2. Evaluation Stage. a). Preliminary Screening. Promis-ing variables listed in Table 7 are assessed to determine theimpact of the factors considered. The problem has twelve factors,with at least two possible levels available for the evaluation ofeach factor. No geometrical restrictions in the outputs for thesearching space (geometrical form) were found in the diagnosisstage. A first screening experimental design, a fractional factorialdesign (FFD) at two levels with twelve factors, was applied toidentify the most important factors, which is followed by fitting areduced model based on those. Matlabis applied to automaticallyfind and generate an FFD on two levels based on the twelve

    factors (k), the maximum number of runs (2kp) with p numberof generators, and the resolution level by using the functionfracfactgen. A minimal resolution level of IV and six generators(p) was proposed in the Matlab to provide a good balance betweenthe number of runs and the confounding level. The minimumnumber of simulations suggested by the generated design is64. The generators were as follows: X3X4X5X6, X2X3X5X6,X2X4X5X6, X1X4X5X6, X1X3X5X6, and X1X2X3X4X5X6. Thecorresponding natural variables (real operational value for eachfactor) are presented in Table 8. Interactions between three or

    more factors were assumed to have a lower effect on the MaPrresponse, so these were not taken into consideration.

    The ANOVA for the 64 simulation responses was made withthe statistic toolbox of Matlab 7.0.1 with a significance levelchosen in 99%. The ANOVA results for all the second-orderinteractions of the 12 factors visualized that all of them yielded top-values larger than 0.01. Therefore, these interactions are notsignificantly important. Table 9 presents the most importantfactors with its p-values and effect. The magnitude of the factorseffect can be ranked based on its absolute values of MaPr*.

    The most significant effect on the MaPr response is givenwhen the inlet nitrogen content has its highest value of 20%molein the range of interest. However, this cannot be controlled due

    Table 6. Continuous Factors in the Plant

    description unit

    range

    disturbance best setting

    improvement

    on MaPr, %

    improvement on

    C3+ recovery, %

    reduction on high pressure

    compression power, %

    pressure in stage 1 of demethanizer kg/cm2 10 to +10% 10% 2.6 0.06 0.04

    power generation capacity of second turbo-expander MW 5 to +15% +15% 0.99 0.05 3.45

    power generation capacity offirst turbo-expander MW 3 to +6% +6% 0.06 0.03 0.48

    % mole nitrogen % mole 5 to 30% 5% (base case) 0 0 0

    Table 7. Feasible Changes for the Cryogenic I Plant

    factor description

    X1 Pressure in stage 1 of demethanizer.

    X2 Power generation capacity of second turbo-expander.

    X3 Power generation capacity offirst turbo-expander.

    X4 Nitrogen content at inlet stream.

    X5 Increasing the current amount of pumparound flow rate.

    X6 Adding one pumparound and one heat exchanger from stage

    number 23 in demethanizer to second separator tank liquid outlet.

    X7 Adding one pumparound from stage number 28 in demethanizer to

    the heat exchanger outlet of the high pressure compressors.

    X8 Varying the feeding stages position of the current ones in the

    demethanizer column.X9 Increase the number of stages in the demethanizer column.

    X10 Replace existing column trays with new one in the demethanizer

    column (type of trays) to increase its efficiency.

    X11 Increase power generation capacity offirst turbo-expander by

    installing other in parallel to the current one.

    X12 Increase power generation capacity of second turbo-expander by

    installing other in parallel to the current one.

    Table 5. Environmental Indexes Estimated for the Base Case

    emission EIPPCB Mexican law regulated limits27 case study (simulation)

    SO2 (tonnes SO2/Mt HC) 306000 50 0

    NOx (tonnes NOx/Mt HC processed) 60500 190 0.979

    CO2 (tonnes CO2/Tonne HC processed) 0.020.82 N/A 0.001

    VOC (tonnes VOCs/Mt HC processed) 506000 N/A 44.244

    SEC (GJ/tonne HC processed) 14 N/A 1.066

    Table 8. Natural Variables for the Twelve Factors at TwoLevels Used in FFD

    factor units level (1) level (+1)

    X1 kg/cm2 22.5 26.5

    X2 kW 2803 3101

    X3 kW 2391 2541

    X4 % mole 5 30

    X5 Kgmol/h 2268 6804

    X6 Kgmol/h 0 4536

    X7 Kgmol/h 0 4536

    X8 number 18, 8, and 6 16, 6, and 4

    X9 number 30 48

    X10 % 58 63X11 kW 0 2466

    X12 kW 0 2953

  • 7/29/2019 pinch anaylysis of cooling water

    8/13

    12074 dx.doi.org/10.1021/ie200722z |Ind. Eng. Chem. Res. 2011, 50, 1206712079

    Industrial & Engineering Chemistry Research ARTICLE

    to feed specifications and is not shown in Table 9 accordingly.The Retrofit Design Approach continues with the second part ofthe evaluation stage, in which the application of the RSM wasconducted based on the previously identified most importantfactors, leading to a reduced model.

    b). Application of RSM. Before proceeding to the RSM, twomain considerations were made in the methodology. First, inorder to cover thewide range of nitrogen composition of the inletto the system, and not to generate infinite cases for the number ofpossible compositions, it was decided to fix them with four levelsof 5, 10, 15, and 20 mol %. The second consideration was forfactor X1, for which Table 9 showed a decrease in MaPr, and itwas set at its lowest operating value of 22.5 kg 3 cm

    2 and toprovide improvement in MaPr. This also simplified the system tobe investigated, as factor X1 can be treated as constant in thereduced model. It should be noted that considerations for X1 andX4 have been possible because there were no interactions presentbetween MaPr and them, which means that factors can be setindependently of each other.28 Consequently, the RSM wasapplied to find a reduced model with fixed X1 at lowest leveland X4 at different nitrogen contents, which predicted the MaPrresponse based on X7, X11, and X12. The employed approach tofit the surface response is a Central Composite Design (CCD)because there were no geometrical restrictions in the outputs forthe searching space and it can be built based on the previousFFD. The additional points for the proposed surface responsedesign were placed at values which is given as

    ( number of factors1=4 31=4 1:316

    An additional central point was set to verify the calibration ofthe simulations along the CCD (i.e., no settings were movedaway from the initial point). Integer parameters infeasibilitieswere avoided. This could be addressed easily in this stage, as all ofthe structural changes (binary factors) were found to improveMaPr when existing; therefore, all three factors were set in thecontinuous range. The simulations, based on the CCD applied,are shown in Table 10 with both coded variables (i.e., levels foreach factor) and its corresponding natural variables (i.e., realvalues for the factor). These define thevalues of variables for eachsimulation run.

    Fifteen simulations were run by following this CCD design.MaPr* responses were obtained, and the linear least-squares(LLS) method was carried out in Matlab for fitting the corre-sponding model. The same procedure was applied to each of thefour fixed inlet nitrogen compositions, and best fit models foreach case were obtained. These results for MaPr* response arepresented in Table 11, together with its root-mean-square error(RMSE). Factors X7, X11, and X12 are coded (i.e., ranged from1.316 to 1.316). The RMSE indicated a good reproducibilityfor the MaPr* response, which are supported by the plot ofresiduals as shown in Figure 3.

    3.4. Optimization Stage. Optimization of RSM Model. Thereducedmodels obtained andpresented in Table 11 are nonlinear in

    nature and are a function of three variables. The models inTable 11 were optimized by the NLP solver in Excel to maximizeMaPr*, andvarious starting points, for example, in theorder of 10

    points, were tested for the three factors within ranges, with whichthe best solution is selected and presented in Table 12. Figure 4visualizes MaPr* values for both the base case and the optimumsolution at each nitrogen scenario. Additionally, as suggested inthe approach, a set of confirmatory simulation runs near thesolution identified are carried out to generate a feasible solution.The results are given as the percentage of difference between thereduced model and the simulation in the last column of Table 12.

    Table 13 shows the actual MaPr* and its improvements for theoptimal results at 10% mole nitrogen content scenario. Mainproduct compositions, percentages of variation in demethanizerreboiler duty, C3+ recovery, and high-pressure compressionpower respectively are also shown. Figure 5 shows the optimalresults which were compared with the base case at the 10% mole

    inlet nitrogen scenario.There are a few comments and observations regarding im-

    provement made in the case study.Demethanizer Reboiler Duty.At a fixed nitrogen level of inlet

    feed, a reduction in the reboiler heat duty can be achieved byprocess optimization. This resulted from the introduction of anew pumparound, which reduces the duty needed by the reboilerandproduces a better heat distributionalong thecolumn, makingfractionation much easier. The difference in reboiler dutybetween base case and optimum case increases considerably asthe inlet nitrogen increases. This is because the C2+ inlet to theplant is relatively reduced, due to theincreasing nitrogen content,which results in a decrease in the total duty required by the reboiler.

    Table 9. p-Values and Estimated Effects for the Most Important Factors

    factor description p-value factors effect, absolute units

    X1 pressure of demethanizer 0 11,567

    X7 additional pumparound from demethanizer to high pressure heat exchanger 0 11,493

    X11 additional turbo-expander in parallel with first turbo-expander 0.0001 7034

    X12 additional turbo-expander in parallel with second turbo-expander 0.00117 4215

    Table 10. CCD for the Case Study

    coded variables natural variables

    number of

    simulation

    X7 absolute

    units

    X11 absolute

    units

    X12 absolute

    units

    X7

    kgmol/h

    X11

    kW

    X12

    kW

    1 1 1 1 2041 2219 2658

    2 1 1 1 2041 2219 3248

    3 1 1 1 2041 2713 2658

    4 1 1 1 2041 2713 3248

    5 1 1 1 2495 2219 2658

    6 1 1 1 2495 2219 3248

    7 1 1 1 2495 2713 26588 1 1 1 2495 2713 3248

    9 1.316 0 0 2566 2466 2953

    10 1.316 0 0 1970 2466 2953

    11 0 1.316 0 2268 2791 2953

    12 0 1.316 0 2268 2141 2953

    13 0 0 1.316 2268 2466 3342

    14 0 0 1.316 2268 2466 2564

    15 0 0 0 2268 2466 2953

  • 7/29/2019 pinch anaylysis of cooling water

    9/13

    12075 dx.doi.org/10.1021/ie200722z |Ind. Eng. Chem. Res. 2011, 50, 1206712079

    Industrial & Engineering Chemistry Research ARTICLE

    C3+ Recovery. At a fixed nitrogen level, C3+ recovery isincreased at the optimal conditions with having different operat-ing conditions and introducing two additional turbo-expanders,which increases in the liquid entering the demethanizer. Whennitrogen is increased in the feed, marginal increase for C3+recovery is observed for the higher nitrogen level in the feed. Itshould be noted that increase of nitrogen content in the feedproduces less amount of liquid product in the system.

    Power Requirement for High Pressure Compressor. Themechanical power for driving high-pressure compressors de-creases notably when new optimal conditions are applied. Thisoccurs because two new additional turbo-expanders in the

    system, through its coupled compressors, provide more thanenough power to compensate additional power requirement ofcompression. With the increasing nitrogen content, the volumeof gas in the vapor phase is greater than pure Residue Gas(Residue Gas plus inert nitrogen), which results in a higherpower required to run the high-pressure compressors at the highinlet nitrogen level than at the low level.

    MaPr*. Overall profit is increased when optimal operatingconditions are applied with the introduction of additional turbo-expander and pumparound, as a net result of the reduction indemethanizer reboiler duty, an increase in C3+ recovery, and areduction of compression power for high-pressure compressor.To completethe evaluation stage,it was necessaryto considerthe

    capital costs associated with the structural changes proposed bythe retrofit. In order to do this, the case with 10% mole of inletnitrogen was selected for the following detailed study. In general,capital investment was estimated for the new additional unitsconsidered in the current study as in eq 4 for the new units,namely, additional pumparound (X7), additional turbo-expander(X11), and additional second turbo-expander (X12). The index

    used to compare different schemes was defined as

    CIPA CI

    PA

    3 100 10

    where CIPA = index of capital investment over the referencevalue [%], CI = capital investment for each retrofit scenario [],and PA = reference value for capital investment ( 176 m).

    Table 14 shows the calculation basis for the new units, basedon eq 4 and data of capital costs available in Timmerhaus et al.23

    The optimum solution identified from the proposed approach ispurely based on the objective function used in the methodology,but, when there is budget restriction, it is necessary to look at

    Table 11. Best Fit Models at 4 Nitrogen Levels for MaPr* (Normalized)

    % mole inlet N2 best fit model % RMSE

    5 MaPr* = 108.8 102 + 0.05 102X11 0.24 102X7

    2 0.24 102X112 0.11 102X12

    2 0.13

    10 MaPr* = 73.1 102 + 0.1 102X11 + 0.1 102X12 + 0.1 10

    2X122 0.1 102X11X12 0.03

    15 MaPr* = 36.9 102 + 0.1 102X11 + 0.2 102X12 + 0.02 10

    2X122 0.01

    20 MaPr* = 0.2 102 + 0.1 102X11 + 0.2 102X12 0.02 10

    2X122 0.09

    Figure 3. Plot of residuals for the best fit model @ 5%mole nitrogen.

    Table 12. Coded Variables for Optimal Results at Each N2Case

    coded variables

    inlet N2 X7 X11 X12

    difference reduced

    model (simulation)

    5% mole 0 0.1 0.02 0.4%

    10% mole 1.316 0.334 1.316 0.2%

    15% mole 1.316 1.316 1.316 0.1%

    20% mole 0.002 1.316 1.316 0.01%

    Figure 4. MaPr* for optimal results and base cases at each N2 case.

  • 7/29/2019 pinch anaylysis of cooling water

    10/13

    12076 dx.doi.org/10.1021/ie200722z |Ind. Eng. Chem. Res. 2011, 50, 1206712079

    Industrial & Engineering Chemistry Research ARTICLE

    different retrofitting options with different levels of investmentrequired. Therefore, in this study, subproblems are generatedand reoptimized by allowing one or two design variables to beadjusted from three potential factors considered in the optimalretrofit cases. For example, optimization can be carried out to

    find an optimal solution when only the introduction of newturbo-expanders is allowed in the existing plant (i.e., X11 and X12are considered, not X7), which is suboptimal, but requires lesscapital investment. The increase of MaPr from each sub-problem and corresponding capital requirement is obtained,

    Table 13. Optimal Results vs Base Case Comparison at 10% Inlet N2a

    streams residue gas high pressure residue gas low pressure C2+

    molar fraction bc OC % diff BC OC % diff BC OC % diff

    N2 0.12 0.12 0 0.15 0.15 0 0 0 0

    CO2 0 0 0 0 0 0 0 0 0

    CH4 0.86 0.87 1 0.83 0.84 1 0.01 0.01 0C2H6 0.02 0.01 34 0.02 0.01 64 0.53 0.54 2

    C3H8 0 0 0 0 0 0 0.27 0.27 0

    n-C4 0 0 0 0 0 0 0.09 0.09 0

    i-C4 0 0 0 0 0 0 0.04 0.04 0

    n-C5 0 0 0 0 0 0 0.03 0.02 3

    i-C5 0 0 0 0 0 0 0.02 0.02 0

    C6+ 0 0 0 0 0 0 0.01 0.01 0

    H2O 0 0 0 0 0 0 0 0 0

    rate, kg/s 95.914 94.408 1.5 10.985 10.858 1.1 46.281 47.831 +3.3

    T, C 43 43 0 44 44 0 21 21 0

    P, kg/cm2 71.4 71.4 0 10 10 0 25.3 25.3 0

    demethanizer reboiler duty, kW 5332 3838 28

    C3+ recovery, % 99.64 99.89 +0.2

    high pressure compression power, kW 12652 10959 13.4

    MaPr*, absolute units 0.644 0.735 9.0a BC = base case (Operating data used are based on August 2008 when feed contains 5% mole nitrogen.); OC = optimum case.

    Figure 5. Optimal results vs base case at 10% inlet N2. (BC = base case; OC = optimum case).

  • 7/29/2019 pinch anaylysis of cooling water

    11/13

    12077 dx.doi.org/10.1021/ie200722z |Ind. Eng. Chem. Res. 2011, 50, 1206712079

    Industrial & Engineering Chemistry Research ARTICLE

    which enables to have a series of feasible retrofitting options atdifferent budget levels.

    All the combinatorial opportunities are presented in Figure 6,in terms of MaPr and CIPA. The change of operating pressure ofdemethanizer (X1) does not require capital investment and itsimpact is presented for comparison purpose. Figure 7 presentsthe payback periods estimated with eq 5 for each case shown in

    Figure 6. This portfolio can be very useful for supporting decision-making procedures in practice.

    The first option (i.e., change of pressure in demethanizer, X1)provides 2.6% increase in MaPr. Besides this first no-investment option, there are three levels of investment ac-cording to the capital investment for all the available options;these levels are identified with dashed lines in both graphs.

    Table 14. Calculation Basis for Capital Costing23

    item features sizing and capacities

    capital cost

    estimated, MM/Y

    1. Additional Pumparound from Demethanizer to High Pressure Heat Exchanger (X7)

    pump centrifugal pump, electric motor

    included 1220, API-610 cast steel casing

    flow rate: 0.187 m3/s,

    pressure: 33.742 kg/cm20.015

    pipe purchased cost of cast-iron pipe, bell

    and spigot pipe, 10351725 kPa

    diameter: 0.1524 m

    length: 30 m 0.0001

    installation arrangement demethanizer and heat exchanger

    arrangements plus installation costs

    40% of purchase cost 0.0058

    total 0.02

    2. Additional Turbo-Expander in Parallel with First Turbo-Expander (X11)

    com pressor purchased cost of compressors , inc luding dr ive,

    gear mounting, base plate, normal 1228,

    centrifugal turbine, carbon steel

    power capacity: 2500 kW 0.172

    turbine purchased cost of turbine and internal combustion

    engine drivers 1235, steam turbine

    power capacity: 2500 kW 0.0178

    installation arrangement compressor and turbine copling

    and placement in the system

    40% of purchase cost 0.0760

    total 0.27

    3. Additional Turbo-Expander in Parallel with Second Turbo-Expander (X12)

    compressor purchased cost of compressors, including

    drive, gear mounting, base plate, normal

    1228, centrifugal turbine, carbon steel

    power capacity: 2953 kW 0.204

    turbine purchased cost of turbine and

    internal combustion engine

    drivers 1235, steam turbine

    power capacity: 2953 kW 0.0193

    installation arrangement compressor and turbine

    copling and placement in the system

    40% of purchase cost 0.0892

    total 0.31

    Figure 6. MaPr and capital costs comparison (10% inlet N2). Figure 7. Payback period comparison (10% inlet N2).

  • 7/29/2019 pinch anaylysis of cooling water

    12/13

    12078 dx.doi.org/10.1021/ie200722z |Ind. Eng. Chem. Res. 2011, 50, 1206712079

    Industrial & Engineering Chemistry Research ARTICLE

    The choice of the investment level mainly depends on the budgetavailability.

    INV1. Among the first level of investment, the introduction ofpumparound on a standalone basis (X7) is the least expensiveoption, with only a 0.7% of CIPA, while it provides the lowestincrease in MaPr (2.9%). This is mainly due to its contributiontoward a lower heat duty in the demethanizer reboiler.

    INV2. The second level of investment lies in the range of 10%for CIPA, whichcovers four options. As thepayback periodfor allfour options in this set is very close at 0.29 years on average, it issuggested that the physical feasibility and operational difficulty(control system) must be taken into account to choose the finaloption.

    The firstoption in this level is the additional turbo-expander inparallel with the first one existing (X11), which achieves a 7.4%increase in MaPr with 8.5% of CIPA. The increase is mainly dueto the high-pressure compression power reduction and thehigher C2+ recovery achieved. This is considerably higher thanthe first option, as is the increase in its capital cost. The paybackperiod is 0.26 years which is the lowest of all the options in thislevel. As this is only one turbo-expander added in parallel, it

    would not introduce too much complexity in the operation andcontrol and therefore this can be a most promising option.The second option in this level is the additional turbo-

    expander to the second existing one (X12). The effect of this inMaPr is similar to the first additional turbo-expander but higher(9.4%) due to its additional power generation capacity. Thelarger size also contributes to higher capital cost, resulting inhigher payback period of 0.30 years, compared to other optionsin this level.

    The third and fourth option in this level is the simultaneousintroduction of two new units: i) a pumparound and a turbo-expander to the first existing one (X7+X11) and ii) the additionalpumparound and turbo-expander to the second existing one(X7+X12). These two structural changes can have potential

    difficulties on process control and practical limitation in therevamping, due to existing plant layout.

    INV3. This final level of investment includes two options:i) simultaneous implementation of two turbo-expanders and ii)all the available factors are considered.

    The optimum solution found with the three parameters op-timized has the highest profit improvement among all possibleoptions but requires the highest capital investment. It is im-portant to mention that the process changes proposed in thiswork are generated by the new design methodology in this paperwith the aid of process simulation and computational optimiza-tion. This implies a further feasibility study required to confirmthe applicability and implementability of the proposed changes inpractice, for example, the hydraulic of the systems, plant layout,

    maintenance, andoperability, etc., although it is beyond thescope ofthis study to carry out further detailed engineering studies.

    4. CONCLUSIONS

    In this research, a simulation-based Retrofit Design Approachhas been proposed and applied to an industrial case study. Theclassical methods of experimental design have been evolved withthesystematic andstrategic useof process simulation andprocessintegration concepts, along with RSM methodology, which leadsto provide generic design methodology for retrofit analysis ofprocess industries. The new design methodology has beentested,and its usefulness for developing a portfolio of retrofit design

    opportunities has been demonstrated with existing gas proces-sing NGL recovery plant. Retrofit Design Approach has beenshown to be a practical and reliable approach to achieve pseudo-optimal solutions within a reasonable time scale without relyingon building and solving very complex optimization frameworks.Incorporation of process integration tools and concepts in themethodology plays an important role, in order to fully understand

    system interactions existed in a retrofi

    t study, as well as to screenand evaluate systematically structural and operational options.

    AUTHOR INFORMATION

    Corresponding Author*Phone: +82 (0)2 2220 2331. E-mail: [email protected].

    NOMENCLATUREACCNewUnits annualized capital cost for new unitsAF annualization factor (y1)AirEi air emission of i (t i/MMt HC)ANOVA analysis of varianceBC base case

    CCD central composite designCCNewUnit capital cost of the new unit ()CIPA index of capital investment over reference value

    of plant assets (%)CI capital investmentDoE design of experimentsEPA42 factor EPA-42 emissions factor (MMt CO2/MMBtu)FD factorial designsFFD fractional factorial designsFi factor in EPA 42 for compound i [(t i*s)/

    (MMt HC*kJ)]FrFG flow rate of fuel gas (kg/s)FrSGP flow rate of sweetened gas processed (t HC pro-

    cessed/y)

    GCV gross calorific value (kJ/kg)GHG greenhouse gasHC hydrocarbonLP linear programmingMaPr marginal profitMaPr* marginal profit normalizedN/A not availableNGL natural gas liquid recovery processNLP nonlinear programmingN2Pe nitrogen penalty NPr net profitLPG gas liquefied from petroleumOD operating dataPA plant assets

    PI process integrationPLP project life periodPP payback period (y)RG residue gasRGHP residue gas high pressureRGLP residue gas low pressureRMSE root mean square errorRSM response surface methodologySCO profit from the sales of coproductSEC specific energy consumption (GJ/t HC processed)SG sweetened gasSM simulated resultsSP profit from the sales of product

  • 7/29/2019 pinch anaylysis of cooling water

    13/13

    12079 dx.doi.org/10.1021/ie200722z |Ind. Eng. Chem. Res. 2011, 50, 1206712079

    Industrial & Engineering Chemistry Research ARTICLE

    SRGHP sales of RGHPVCE variable cost of energyVCRM variable cost of raw material

    REFERENCES

    (1) Pintaric, Z. N.; Kravanja, Z. The Two-level Strategy for MINLP

    Synthesis of Process Flowsheets under Uncertainty. Comput. Chem. Eng.2000, 24, 195.

    (2) Pintaric, Z. N.; Kravanja, Z. A Strategy for MINLP Synthesis ofFlexible and Operable Processes. Comput. Chem. Eng. 2004, 28, 1105.

    (3) Lazzaretto, A.; Toffolo, A. Evolutionary Algorithms for Multi-objective Energetic and Economic Optimization in Thermal SystemDesign. Energy 2002, 27, 549.

    (4) Jia, X.; Zhang, T.; Wang, F.; Han, F. Multi-objective Modelingand Optimization for Cleaner Production Processes. J. Clean. Prod.2006, 14, 146.

    (5) Ilzarbe, L.; Tanco, M.; Alvarez, M. J. Practical Applications ofDesign of Experiments in the Field of Engineering: A BibliographicalReview. Qual. Reliab. Eng. Int. 2008, 24, 417.

    (6) Tanco, M.; Viles, E.; Ilzarbe, L.; Alvarez, M. J. How is Experi-mentation Carried Out by Companies? A Survey of Three European

    Regions. Qual. Reliab. Eng. Int. 2008, 24, 973.(7) Tanco, M.; Viles, E.; Alvarez, M. J. Barriers Faced by Engineerswhen Applying Design of Experiments. TQM J. 2009, 21, 565.

    (8) Francheschini, G.; Macchietto, S. Model-based Design of Ex-periments for Parameter Precision: State of the art. Chem. Eng. Sci. 2008,63, 48464872.

    (9) Box, G. E. P.; Wilson, K. B. On the Experimental Attainment ofOptimum Conditions. J. R. Stat. Soc. 1951, Series B, 1.

    (10) Montgomery, D. C. Design and Analysis of Experiments;John Wiley & Sons: New York, 2005.

    (11) Cervantes, M. J.; Engstorm, T. F. Factorial Design Applied toCFD. J. Fluids Eng. 2004, 126, 791.

    (12) Davim,J. P.; Cardoso, R. Thermo-mechanical Model to Predictthe Tribological Behaviour of the Composite PEEK-CF30/steel Pair inDry Sliding using Multiple Regression Analysis. Ind. Lubr. Tribol. 2005,

    57, 181.(13) Liaua, L. C.; Yang, T. C. K.; Tsai, M. T. Expert System of aCrude Oil Distillation Unit for Process Optimization using NeuralNetworks. Expert Syst. Appl. 2004, 26, 247.

    (14) Tong, K. W.; Kwong, C. K.; Yu, K. M. Intelligent ProcessDesign System for the Transfer Moulding of Electronic Packages. Int. J.

    Prod. Res. 2004, 42, 1911.(15) Kabir, C. S.; Chawathe, A.; Jenkins, S. D.; Olayomi, A. J.; Aigbe,

    C.; Faparusi, D. B. Developing New Fields Using Probabilistic ReservoirForecasting. SPE Annual Technical Conference and Exhibition proceedings,Paper Number 77564-MS DOI, 2002.

    (16) Tanco, M.; Ilzarbe, L.; Alvarez, M. J. Implementation of Designof Experiments projects in industry. Appl. Stochastic Models Bus. Ind.2009, 25, 478.

    (17) Triantafyllou, C; Smith, R. Design and optimization of fullythermally coupled distillation columns. Chem. Eng. Res. Des. 1992, 70 (A2),118132.

    (18) Shah, P; Kokossis, A. NewSynthesisFramework forthe Optimisa-tion of Complex Distillation Systems. AIChE J. 2002, 48, 527550.

    (19) Smith, R. Chemical Process Design and Integration; John Wiley &Sons: West Sussex, 2005.

    (20) Linnhoff, B.; Townsend, D.; Boland, D.; Hewitt, G.; Thomas.,B.; Guy, A.; Marsland., R. User Guide on Process Integration forthe EfficientUse of Energy; IChemE: Rugby, 1982.

    (21) Klemes, J.; Friedler,F.; Bulatov, I.; Varbanov,P. Sustainability inthe Process Industry: Integration and Optimization; McGraw-Hill:New York, 2010.

    (22) Nemet, A.; Varbanov, P. S.; Klemes, J. Waste-to-energy tech-nologies performance evaluation techniques. Chem. Eng. Trans. 2011,25, 513520.

    (23) Myers, R. H.; Montgomery, D. C.; Vining, G. G.; Borror, C. M.Response Surface Methodology: A Retrospective and Literature survey.

    J. Qual. Technol. 2004, 36, 5378.(24) Box, G. E. P.; Draper, N. R. Empirical Model-Building and

    Response Surfaces; John Wiley & Sons: New York, 1987.(25) Office of Air Quality Planning and Standards. Emission Factor

    Documentation for AP-42 Section 1.4 Natural Gas Combustion (EPA-R09-OAR-2006-0635-0013); US Environmental Protection Agency:

    Research Triangle, 2006.(26) European Integrated Pollution Prevention and Control Bureau.

    Best Available Techniques for Mineral Oil and Gas Refineries; EuropeanCommission, 2003.

    (27) Mexican Government. Official Journal of the Federation NOM-001-SEMARNAT-1996, 2007.

    (28) Myers, R. H.; Montgomery, D. C. Response Surface Methodol-ogy: Process and Product Optimization Using Designed Experiments; John

    Wiley & Sons: New York, 2002.(29) Timmerhaus, K.; Peters, M. S. Plant Design and Economics for

    Chemical Engineers (5th ed.); McGraw-Hill: New York, 2003.