Phenomena Based Methodology for Process Synthesis Incorporating

Article pubs.acs.org/IECR

Phenomena Based Methodology for Process Synthesis Incorporating Process Intensification Philip Lutze,*,†,‡ Deenesh K. Babi,§ John M. Woodley,‡ and Rafiqul Gani§ †

Laboratory of Fluid Separations, Department of Biochemical & Chemical Engineering, Technical University of Dortmund, Emil-Figge-Strasse 70, D-44227 Dortmund, Germany ‡ PROCESS, Department of Chemical & Biochemical Engineering, Technical University of Denmark, Søltofts Plads, Building 229, DK-2800 Kgs. Lyngby, Denmark § CAPEC, Department of Chemical & Biochemical Engineering, Technical University of Denmark, Søltofts Plads, Building 229, DK-2800 Kgs. Lyngby, Denmark S Supporting Information *

ABSTRACT: Process intensification (PI) has the potential to improve existing as well as conceptual processes, in order to achieve a more sustainable production. PI can be achieved at different levels, that is the unit operations, functional, and/or phenomena level. The highest impact is expected by looking at processes at the lowest level of aggregation which is the phenomena level. In this paper, a phenomena based synthesis/design methodology incorporating process intensification is presented. Using this methodology, a systematic identification of necessary and desirable (integrated) phenomena as well as generation and screening of phenomena based flowsheet options are presented using a decomposition based solution approach. The developed methodology as well as necessary tools and supporting methods are highlighted through a case study involving the production of isopropyl acetate.

1. INTRODUCTION Process intensification (PI) is one of many options to match current and future challenges of the chemical and biobased industry. These challenges are, for example, implementation of more sustainable production processes; adaption to quicker changing markets; and the potential to generate new innovative products in order to survive global competition. A global definition of PI is still not available, and the concept has been changed from technologies aiming at volume reduction as proposed in the early 1980s to definitions stressing that PI is an integrated approach for improvement of process efficiencies by the enhancement of transport phenomena. A broad overview about definitions and equipment of PI can be found elsewhere.1,2 Most implemented examples of PI are reactive distillation, dividing wall columns, and reverse flow reactors.3 In reactive distillation synergistic effects are used by integrating reaction and distillation, which, for example, enhances the reaction phenomenon by increasing the equilibrium conversion through in situ product removal via evaporation. Information on implementation of other PI technologies is scarce.3 This is because the development of intensified processes or intensified unit operations is not straightforward,1 and also, while a large number of process options for intensification are potentially available, the identification of the best potential candidate based on quantitative reasoning is time and resources consuming. In a previous work, we reported the development of a general systematic PI synthesis/design methodology with an efficient solution procedure of the (mathematical) synthesis problem, based on the decomposition approach.1,4 Even though improvements could be achieved in several cases studied, the methodology is limited to considering existing PI unit © 2013 American Chemical Society

operations that are also available in a knowledge base. In order to go beyond these unit operations and achieve even larger benefits, PI process synthesis/design must be investigated at a lower level of aggregation.5 Synthesis/design concepts/methods that may help to achieve PI beyond known unit operations are, for example, the development of novel reactor networks based on elementary process functions;6 the means-ends analysis;7 the generalized modular representation framework (GMR) for process synthesis;5 the phenomena based process synthesis based on manipulation and variation of process phenomena;8 and the phenomena based modularization approach.9 The approach based on elementary process functions6 tracks a fluid element through a reactor with possibilities to integrate separation and heating/cooling (thermal). In a stepwise procedure, the model depth is increased. The first level is the level of integration in which the optimal route in the state space is identified. In the second level operational constraints based on detailed mass and energy transport calculations are integrated within the design of level 1. In the last level, the unit operation is identified to screen for technical constraints of the design. Reactor parameters such as interfacial areas, residence time, and number of units are not defined a priori but are investigated through a stepwise procedure. The method has been illustrated for a SO2 oxidation reactor.6 Their Special Issue: PSE-2012 Received: Revised: Accepted: Published: 7127

September 27, 2012 January 23, 2013 January 23, 2013 January 23, 2013 dx.doi.org/10.1021/ie302513y | Ind. Eng. Chem. Res. 2013, 52, 7127−7144

Industrial & Engineering Chemistry Research

Article

promising attempts in synthesis/design to achieve PI beyond existing unit operation have been proposed, systematically finding the best intensified option for a whole process remains unsolved using the approaches mentioned previously. In this paper, a systematic synthesis/design methodology to achieve process intensification based on phenomena as building blocks is proposed. Phenomena are selected because all process intensification options can be described by their enhancement.1 Like the earlier developed unit operation based methodology,4 the decomposition approach handles large numbers of process options generated by combining the phenomena based building blocks. First, the problem is defined and analyzed to identify potential phenomena and their potential connection within the process for a targeted process improvement. Next, phenomena are connected to form process options, which are stepwise screened at the phenomena level for feasibility and performance constraints. The most promising phenomena based process options are then translated into unit operations. Here, additional constraints with respect to the unit operational level are integrated, meaning that the solutions are checked for feasibility and performance. The best option is identified by a final fine-tuning of the parameters and evaluating the feasible options with respect to an objective function. Application of the methodology requires a number of additional tools, subalgorithms, and methods, which are also discussed in this paper. The application of the methodology is highlighted through a case study which is the production of isopropyl acetate. The case study and the methodology have been presented briefly in a proceedings paper;11 in this paper, all details will be given.

knowledge base consisting of all constraints and unit operations, the set of rules necessary to identify novel reactor concepts from the functional analysis as well as a detailed algorithm, has not been fully presented yet. The basis of the means-ends analysis7 is the specification of the process inlet and outlet streams. Based on a set of rules, the tasks to satisfy those specifications are identified. By heuristic (expert knowledge) variations different process options are generated which are evaluated based on sets of performance criteria. The method has been illustrated by the generation of the reactive distillation unit for the production of methyl acetate. However, even though novel processes/units may be identified, the application is not simple. Rules and/or algorithms for identification and variations of tasks as well as for identification of unit operations have not been published. Furthermore, it does not aim to generate all potentially feasible options and based on this cannot guarantee to find the best flowsheet having taken into account PI. Process synthesis by the GMR-approach5 is based on heat and mass building blocks instead of defined (conventional) equipment. Heat and mass building blocks may or may not be connected using a set of connectivity rules. If a given connection of these blocks gives a feasible and promising solution, then in a subsequent step, unit operation(s) are identified for those. Until now, this approach has been successfully illustrated for column synthesis/design such as distillation, reactive distillation, and absorption.5,10 The selection of the initial search space of building blocks is based on heuristics and thermodynamic insights. A complete set of rules of how to identify unit operations beyond this has not yet been shown. The synthesis concept by Rong et al.8 is based on process phenomena. They classified process phenomena into “chemistry and chemical reaction phenomena, materials phases and transport phenomena, phases behaviour and separation phenomena etc.”. Process phenomena are characterized by surface materials, operation modes, flow pattern, facility medium, geometry, energy sources, and key variables as well as components and phases. Their methodology decomposes the synthesis problem into 10 hierarchical steps. The heart of their method is trial-and-error variations of the characteristics for the identified key process phenomena through seven different suggested PI principles. The method has been briefly illustrated for the production of peracetic acid.8 Details about algorithms and the stepwise procedure are not given, including the following: a definition/description and systematic identification of phenomena; as well as strategies for variations of these phenomena; techniques how to find all currently available options; and a solution approach how to identify the best option. For each of their conceptual example only the final design of the intensified process is presented. Another concept, using phenomena to synthesize potentially novel process solutions, is the modularization approach.9 In this approach the phenomena is classified at the structural level for the description of phases and interfaces and at the behavioral level for the description of mass transfer, phase change, energy, change conditions, and mechanical operations. Their concept is to aggregate phenomena to form phases. Phases can be aggregated to form tasks (such as stages or devices). Tasks can be aggregated to represent the whole process. Until now, only the library and classification of phenomena as well as the representation of one unit by phenomena have been presented. No details on algorithms, necessary tools, and solution techniques to synthesize processes based on their modularization approach have been published. Hence, even though

2. PHENOMENA BASED SYNTHESIS AND DESIGN METHOD TO ACHIEVE PROCESS INTENSIFICATION The general synthesis problem can be formulated mathematically by eqs 1-5. The quantitative comparison of different processes is realized by the objective function (eq 1) which is subject to a set of design (optimization) variables X, a set of binary decision variables Y, a set of parameters at the unit operation level d ,and a set of product/process specifications θ. Binary decision variables Y describe the existence of phenomena or units and streams F within a process option. This means that if Y = 1, then the corresponding stream, phenomenon, or unit is enabled in the process. A feasible and promising process option must match all constraints. Rules for process/operation synthesis are logical constraints (eq 2), for example to check if a stream has the correct state for entering a phenomenon. Structural constraints define the connectivity between phenomena (eq 3), for example if the integration of two phenomena is not advantageous then they cannot be combined. The process model hp is the set of equations (eq 5) which describe the behavior of the process. These can be built through models at unit operation level or at phenomena level. Those models are steady state or dynamic depending on the process scenario. Z = min FObj( Y , X , d , θ )

(1)

Subject to Y, X, d, θ with Y ∈ {0,1}, X, d, θ ∈ R and

7128

glogical ( Y ) ≤ 0

(2)

gstructural ( Y ) ≤ 0

(3)

goperational ( Y , X , d , θ ) ≤ 0

(4)

dx.doi.org/10.1021/ie302513y | Ind. Eng. Chem. Res. 2013, 52, 7127−7144


Article

Figure 1. Similarity of molecules and processes.

Figure 2. Search space reduction within process synthesis/design via decomposition approach.

∂X = hP ( Y , X , d , θ ) ∂t

Several solution procedures for the CAMD problem exist. One solution approach is proposed by Harper and Gani14 in which the problem is decomposed into several steps. First, molecular groups are selected, and then all feasible molecular structures are generated. Connectivity rules between these ensure the chemical feasibility of the molecule. All feasible molecules are subsequently screened by their performance. This method is efficient because the performance calculation is decoupled from the combinatorial problem14 and has been adopted for unit operation based synthesis/design by Lutze et al.4 In this paper, the decomposition based solution method for solving the (mathematical) synthesis problem (eqs 1−6) is employed and illustrated in Figure 2. When all process options are generated, the initial search space is reduced in successive steps by solving logical and structural constraints (eqs 2 and 3) in a specified hierarchy. These generated (feasible) process options are represented by subsets of fixed binary variables. For each process option the corresponding process model (eq 5) and the operational constraints (eq 4) are solved simultaneously. This needs to be done at the phenomena level as well as at the unit operation level because a set of phenomena is only known when the physical surrounding is defined (for example, wall effects). The set of solutions that match all constraints are the feasible set of the solution. Those can be evaluated by performance criteria (eq 6) for the identification of the most promising process options. Afterward, the parameters of the most promising options can be fine-tuned by optimization of the whole set of equations (solving eqs 1−6) to identify the best solution. The obtained result of the

(5)

The objective function can also be substituted and/or supported by a set of performance criteria (eq 6). A set of performance metric proposed to evaluate PI is given by Lutze et al.1 Ψ( Y , X1..v , θ ) − Ψ Target( Y , X1..v , θ ) ≥ 0

(6)

The formulated general synthesis problem (eqs 1−5) is a mixed-integer nonlinear problem and dependent on the process scenario containing very complex process models. Generating processes by combining phenomena is a complex problem due to the potentially large number of possible combinations. Hence, dependent on size and complexity of the formulated problem the identification of the global optimal solution may sometimes be impossible. Therefore, an efficient and systematic solution approach is necessary. In the area of chemical engineering this type of combinatorial problems is not new. A similar problem is defined in computer-aided molecular design (CAMD) in which the objective is to identify a component or components with specific defined properties.12 The similarity of the structure of flowsheets and molecules has been reported before13 comparing molecules to processes consisting of groups or unit operations respectively. This analogy can be extended to the case of unit operations and phenomena since they are also building blocks in the same way as atoms in groups that represent a molecule (see Figure 1). 7129



Article

class of phenomena the number of inlet and outlet streams are defined which is important for connectivity between phenomena of different classes. Mixing phenomena have minimum one inlet stream and one outlet stream while dividing phenomena have one inlet and minimum two outlet streams. Reaction blocks are defined to have one inlet and one outlet stream. The inlet of phase transition phenomena is a mixture of two phases between which the mass transfer of components is realized. The outlet of a phase transition phenomenon is a mixture of two phases. Phase separation phenomena have one inlet, and the number of outlets is given by the number of phases separated from each other within this phenomenon. Energy transfer phenomena such as heating/cooling and pressurizing/expanding are defined to have either one inlet and one outlet or two inlet and two outlet streams, where the streams are not in contact. An overview of the phenomena available in the phenomena library is given in the supporting material of that paper. 2.2. Connection of Phenomena to Processes. In general, phenomena can be connected or interconnected depending on the overlap of their operating window. Phenomena can be interconnected into a simultaneous phenomena building block (SPB) in case the phenomena are occurring at the same time, at the same position, and having a combined operating window. All others can be connected sequentially when input of the first and output of the second phenomenon match. In order to differentiate this in a simple manner, the analogy to the SMILES code for molecules is used, in which a “-”-sign is used for a connection while the sign “=” is used for interconnection of phenomena. An interconnection can be necessary in case where a phenomenon cannot appear without a second (or more) phenomenon. For example, a phase transition phenomenon needs two phases between which components are transferred; therefore, a phase transition phenomenon has to be interconnected to a phase contact phenomenon. At the same time, the latter phenomenon needs an interconnection to a two-phase mixing phenomenon describing how the phases are mixed in each other (see Figure 3). It is defined that a dividing phenomenon can never be interconnected.

methodology will be the best one for the defined problem, the selected performance criteria, and constraints as well as dependent on the quality of the available data, parameters, and models. A global optimal solution is not guaranteed here − however, the solution obtained most likely would be the best that is possible to obtain. 2.1. Phenomena as Building Blocks. From a conceptual point of view (bio)chemical processes converting raw materials to products consist of streams and blocks that process the streams. Those blocks are conventionally unit operations, but here phenomena are used as building blocks (see Figure 1). Within a phenomena building block, the material or energy of the inlet stream is transferred or converted. Phenomena as building blocks consist of mass, component, energy, and momentum balances. Those are linked to constraint equations describing the phenomenon and additional constitutive equations (for example equations representing thermodynamic properties) as well as the inlet and outlet conditions of the streams. In chemical processing, single streams can be divided to two or more streams. Two or more streams can be mixed to one stream. Besides mixing, a single phase stream may change its concentration by reaction, may change its temperature by energy transfer, and/or may change its state by phase change (initiated by mixing or energy transfer). The state of a stream may be a single phase or multiple phases. Thermodynamically, different streams representing different phases may also interact. The behavior of those interacting streams is described by multiple phase mixing phenomena. Again, energy between those two phases may or may not be exchanged. In case the phases are in direct contact (initiated through mixing), the phase contact behavior need to be phenomenological described. In case a driving force between those contacted phases exists, phase transition takes place. The phases of a stream consisting of multiple phases may or may not be separated by a phase separation to form single phase streams. Therefore, phenomena are classified into 8 different classes which do not overlap and represent the full set of classes to build chemical processing. Those classes are as follows: • Mixing: Description of mass flow within one or between several phases. • Phase contact: Description of the contact and resistances at the phase boundary of phases. • Phase transition: Description of mass transfer of components between two phases (e.g., the vapor−liquid equilibrium relationship). • Phase change: Description of the state change of a complete stream at no phase transition (e.g., full condensation). • Phase separation: Description of the degree of separation of two phases (e.g., aerosols are results of nonideal phase separation). • Reaction: Description of a change in mass of a component or components generated or consumed between inlet and outlet. • Energy transfer: Description of the kind of energy transfer between sources and sinks of energy. • Stream dividing: Division of a stream into two or more streams (temperature, pressure, and concentrations remain unchanged). Each class of phenomena except for stream dividing can be further subclassified. A subclassification of mixing phenomena is, for example, the number of phases involved in the mixing, while for phase transition phenomena the subclasses are depending on the involved phases (e.g., V-L, L-L). For each

Figure 3. Relationship between 2-phase mixing, phase contact, and phase transition phenomena and their necessary interconnection.

Formally, two SPBs can be connected in case the phase, temperature, and pressure of outlet and inlet stream match. However, not all those connections make sense. For example, if a stream enters a reaction phenomenon building block and not all reactants are given yet, the reaction will not take place and this arrangement would be formally feasible. But a reaction may also come before mixing in case the inlet stream contains all reactants. Therefore, a priori definition of the order is not 7130



Article

implied. The concept used here is to first interconnect all phenomena which have a common operating window into SPBs, afterward connect all SPBs with matching inlet and outlet stream specifications, and afterward screen the generated structures by applying logical and structural constraints. For example, the case of a stream entering a reaction phenomenon building block without all reactants given yet would be screened out by logical constraints. SPBs having at least two phases can be connected in three different flow arrangements which are cocurrent-flow and crossflow as well as counter-current-flow. The SPBs can be connected within arrangements of one or more stages to form unit operations operation. Here, another set of rules is used which ensures matching of inlet and outlet streams of connected SPBs. A sequence of unit operations forms a flowsheet. 2.3. General Workflow. The general workflow of the phenomena based synthesis/design methodology is shown in Figure 4. It consists of six steps and has been developed in analogue to the unit operation based workflow published before.4 In each step the user has to perform a number of

substeps and have to take decisions/actions supported by a set of tools and methods. Each step is presented in the following sections. The tools and methods are described in sections 2.4 and 2.5. The list of rules is given in the Supporting Information. The starting point of the methodology is either a base case design of an existing process or the conceptual design of a process. The method is explained below. 2.3.1. Step 1: Define Problem.4 The purpose of step 1 is to define the synthesis/design problem incorporating PI in terms of objective, process/operation scenario, process boundaries, underlying assumptions, and the performance metric for screening PI options. This input is collected in the first four substeps, while in the last substep 1.5 this input is transformed into a mathematical form. Step 1.1: Define the objective function Fobj (eq 1). Step 1.2: Define the design and process scenario. The design scenario is either the improvement of the design of a whole process or specific parts of the process. The process scenario is either a batch or a continuous process. Step 1.3: Define the process and product specifications θ which all generated process options must match. This includes a list of raw materials, quality and quantity of the product, reactions in the process, and also safety specifications. Step 1.4: Define performance metric PM. Note: The performance metric PM can be based on sustainability requirements such as operating and capital cost, safety, energy consumption, waste generation, and efficiency as well as intensification metrics,1 such as simplification of the flowsheet and volume reduction. Simplification is defined by the number of unit operations in the flowsheet. Step 1.5: Translate θ and PM into logical (eq 2), structural (eq 3), operational constraints (eq 5), and performance criteria Ψ (eq 6). Step 1.5.1: Translate items of θ and PM to logical constraints (eq 2) by applying the rules 1.1−1.5 (see the Supporting Information). Step 1.5.2: Retrieve structural constraints (eq 3) for the items of θ and PM from the knowledge base. Step 1.5.3: Translate items of θ and PM into operational constraints (eq 4) or into a performance metric (eq 6) by applying the rules 1.6−1.8. 2.3.2. Step 2: Analyze the Process.4 The purpose of step 2 is to analyze the base case design in order to identify the limitation of the process and the corresponding phenomena. Therefore, all mass and energy data are collected in step 2.1. In step 2.2, the flowsheet of the base case is transformed into task and phenomena in order to decouple the analysis from physical unit operations. In step 2.3, the flowsheet is analyzed to identify limitations/bottlenecks of the process by employing a knowledge-based and a model-based approach. The identified limitations are subsequently linked to corresponding phenomena and tasks inside (step 2.4) and outside (step 2.5) of the unit operation in which they occur. For example, a difficult separation may occur because of impurities formed in a sidereaction in the reactor. Hence, the problem has its source outside the unit operation in which it occurs. In that case also the selectivity of the reaction should be targeted for improvement. Step 2.1: Collect mass and energy data for the base case design: Step 2.1.1: Decide if the energy data of the base case design is required by using rule 2.1.

Figure 4. Workflow of the methodology (abbreviations of the subalgorithms: MBS: model based search; LBSA: limitation/bottleneck analysis; APCP: analysis of pure component properties; AMP: analysis of mixture properties; AR: analysis of reactions; OPW: operating process window; DS: development of superstructure; SoP: selection of phenomena; AKM: apply the extended Kremser method; KBS: knowledge base search). 7131



Article

Step 2.5.1: Identify potential tasks for additional analysis of a LB by the rules 2.12−2.15. Step 2.5.2: Apply rule 2.16 to check if the limitation is in the list LB. 2.3.3. Step 3: Identification of Desirable Phenomena. The objective is to identify additional phenomena for synthesis of flowsheets which match the targets defined in step 1 and improve the process performance of the necessary phenomena (identified in step 2). In step 3.1, the identified limitations/ bottlenecks and the corresponding phenomena and tasks are used to identify tasks and phenomena which need to be improved to achieve the desired performance (here, those are called desirable phenomena). For the selection of suitable desirable phenomena, property analysis of single component (step 3.2) and mixtures of components (step 3.3) as well as reactions (step 3.4) in the system are needed. From the set of suitable desirable phenomena the most promising phenomena are selected in step 3.5. As explained in section 2 some phenomena cannot stand for its own, hence, additional accompanying phenomena are selected in step 3.6. Finally, the operating window of all identified phenomena in the search space is determined which is needed input for step 4. Step 3.1: Collect PI for identified targets (KPI): Identify desirable tasks as well as phenomena ΩPI to enhance the existing necessary process phenomena using substeps 3.1.1− 3.1.3. Step 3.1.1: Prepare a list of keywords KPI for identification of PI possibilities including all identified limitations/bottlenecks, tasks, and the corresponding objective for PI (Fobj, PM). Step 3.1.2: Apply the algorithm KBS to identify the phenomena to be targeted enhanced for process improvement within a task. Step 3.1.3: Apply the algorithm KBS to identify additional desirable tasks of the process. Step 3.2: Use algorithm APCP to identify potential phenomena for each item of ΩPI identified in step 3.1. Step 3.3: Use algorithm AMP to identify potential phenomena for each item of ΩPI concerning mixture properties. Step 3.4: Identify reaction phenomena if identified as additional desirable task Step 3.4.1: Check if a reaction task is identified (rule 3.1) Step 3.4.2: Retrieve reaction data from databases/scientific literature by using the purpose of the task as the keyword (rule 3.2). Step 3.4.3: Enable phenomena applying rule 3.3. Step 3.4.4: Use algorithm AR to analyze/compare/select potential reaction phenomena blocks for each task. Step 3.5: Enter subalgorithm SoP to select the best phenomena for each for each item of ΩPI. Step 3.6: Retrieve accompanying phenomena information from the knowledge base and select accompanying phenomena from the phenomena library for each identified phenomena. Step 3.7: Apply the subalgorithm OPW to determine the operating window for each phenomenon. 2.3.4. Step 4: Generate Feasible Operation/Flowsheet Options. The general objective is to generate all phenomena based process options and to screen them according to logical and structural constraints. The input to this step is the initial search space of identified phenomena (nP,tot) and its corresponding operating windows. In step 4.1, all phenomena are connected to form feasible simultaneous phenomena building blocks (SPBs). Because SPBs are connected to form unit operations and unit operations are combined to form

Step 2.1.2: Collect the mass and/or energy data of the base case design and the phase description of each stream. These data are provided as simulation data or directly as operational data from a plant. The simulation data can be provided from simulation-results of the base case design. Step 2.2: Transform the flowsheet into task based and phenomena based flowsheet. This involves three substeps. Step 2.2.1: Identify for each unit operation of the base-case design the task of the process. For this a set of rules 2.2−2.8 is used to identify the tasks: reaction, separation, heat supply/ removal, pressure increase/decrease, phase change, mixing, and dividing. Step 2.2.2: Identify the split of each separation task by analyzing the split factors (σsep,i) for each component involved in a separation (eq 7) by applying rule 2.9 and afterward rule 2.10. σisepu = ni , out1/ni , in

(7)

Step 2.2.3: Retrieve the list of phenomena involved in each unit operation from the PI knowledge base. Note: For example, an isothermal ideal liquid phase CSTR consists of an ideal liquid phase mixing phenomenon, a reaction phenomenon, and energy transfer phenomena. Step 2.3: Identify limitations/bottlenecks LB of the base case. This objective is achieved by analyzing the collected data. Two methods are applied: A knowledge based method (step 2.3.1) and a model based method (step 2.3.2). Step 2.3.1: Collect the limitations of the process by applying the algorithm KBS. Step 2.3.1.1: Prepare a list of keywords K containing the following items: the process system, the reaction system, the tasks, and the components. Step 2.3.1.2: Apply the algorithm KBS to identify the limitations of the process. Save all retrieved limitations in a list LB. KBS is described in section 2.4. Step 2.3.2: Apply the algorithm MBS for a model based search of limitations in the base-case design. Step 2.3.3: Identify most important limitations/bottlenecks by rule 2.11. Step 2.4: Analyze the obtained limitations/bottlenecks LB and their corresponding tasks in the base-case design. The objective is to identify the phenomena causing the limitation. This requires an analysis of pure component, mixture, and reaction properties together with an analysis of the operational boundaries of the corresponding parts of the process. Step 2.4.1: Analyze pure component properties by applying the algorithm APCP (see section 2.4). Step 2.4.2: Analyze mixture properties by applying the algorithm AMP (see section 2.4). Note: The following mixture properties MP are analyzed: formation of azeotropes; miscibility gaps; and the formation of an explosive atmosphere. Step 2.4.3: Analyze reactions which have been identified to be a limitation/bottleneck by applying the algorithm AR (see section 2.4). Step 2.4.4: Determine the operating window of each task under investigation by applying algorithm OPW (see section 2.4). Step 2.5: Link a limitation/bottleneck to a task and a corresponding phenomenon outside the unit operation in which it occurs. This is achieved through substeps 2.5.1−2.5.2. 7132



Article

Step 4.2.1: Select for each SPB the identification and application of the necessary tools for identification of the configuration by applying rules 4.1- 4.2. Step 4.2.2: Calculate the phase diagram for each SPB containing a phase-transition phenomenon. Step 4.2.3: Identify the key component for the SPB. Step 4.2.4: Apply the DF method15 to determine minimum reflux Rmin (see section 2.5). Step 4.2.5: For each SPB (including a phase transition phenomenon) apply the Kremser method using the subalgorithm AKM to identify the number of stages for counter-flow, cocurrent-flow, and cross-flow and the potential outlets. Step 4.2.6: Compare all SPB in their performance (outcome and number of stages) to select the flow connection (counterflow, cocurrent-flow, and cross-flow) by using rules 4.5−4.8. Step 4.2.7: Compare all SPB in their performance (outcome and number of stages) to identify the minimum number of stages necessary fulfilling the necessary outlet specification of the task. Step 4.3: Generate the number of feasible operation/process options: Step 4.3.1: Retrieve the corresponding stage connection superstructure from the model library. Step 4.3.2: Generate all operation/process options by inserting the SPBs in the search space into the stages of the superstructure. Note: The theoretical maximum number of operation options NOOmax depends on the number of SPBs, the number of (separate) tasks, the connection, and the integration through recycles. For a crossflow arrangement, it can be calculated using eq 9. Input is the total number of stage building blocks NSPB fulfilling each separate task, the maximum number of stages within the operation nS,max, and the number of recycles nR (depending on the number stages, stage arrangements, and number of tasks nT):

processes, the number of stages and the way to connects those stages (see section 2.2) to achieve a process target by using a set of SPBs need to be identified. This is done in step 4.2 by applying an extended Kremser method (explained in detail in section 2.5.5) which is a simple method to determine the necessary number of stages to achieve a defined performance (e.g., conversion, recovery). With the knowledge of the number of necessary stages, the SPBs are connected to a set of process options in step 4.3 using generic superstructures (from a model library). Not all of them are feasible and structural promising. Hence, screening based on logical (step 4.4) and structural constraints (step 4.5) defined in step 1 are performed. Step 4.1: Generate all feasible simultaneous phenomena building blocks SPB from the phenomena in the search space: Step 4.1.1: Identify the maximum number of phenomena within an SPB nP,max. This is calculated by np,max = TPB − (TE − 1) − (TM − 1) − TD, where TPB is the total number of phenomena building blocks, TE is the total energy phenomena, TM is the total number of mixing phenomena, and TD is the total number of dividing phenomena. Note: Each phenomenon can only be present once in an SPB. Opposing energy transfer phenomena such as heating and cooling or pressurizing and expanding and different mixing phenomena such as perfectly mixed and flow mixing can only be present once within an SPB. A dividing phenomenon is defined to be an unconnected single stage. Step 4.1.2: Interconnect all phenomena in the search space to SPBs by using the connectivity rules for each phenomenon stored in the knowledge base. The theoretical maximum number of SPBs can be calculated by giving the total number of phenomena in the search space nP,tot and the maximum number of phenomena within an SPB nP,max (from step 4.1.1) using eq 8: nP , max

NSPBmax =

∑ k=1

⎛ (nP , tot − 1)! ⎞ ⎜⎜ ⎟⎟ + 1 ⎝ (nP , tot − k − 1) ! k! ⎠

nS , max

NOOmax,t = nT

(8)

∑ k=1

Note: Eq 8 has been developed on the basis that a dividing phenomenon is always a single SPB, and therefore the number on the right-hand side of eq 8 is increased by one. The use of eq 8 is illustrated through an example. Considering the following phenomena in the search space (nP,tot=4): Mixing (A), Heating (B), Pressurizing (C), and dividing (D); the number of maximum phenomena within an SPB is nP,max=3 (note that dividing is an SPB on its own) and the total number of phenomena is nP,tot=4. Hence, it gives 8 options (A=B=C, A=B, B=C, A=C, A, B, C, D) or using eq 8:

(NSPBT k + nR , k NSPBT k − 1)

(9)

Step 4.3.3: Screen the operations/processes by connectivity rules for stages based on matching inlet/outlet specifications. The number of remaining feasible stage building blocks is NOOt. Step 4.4: Screen options by logical constraints. Remaining options are NOOL or NPOL respectively. Step 4.5: Screen options by structural constraints. Remaining options are NOOS or NPOS respectively. 2.3.5. Step 5: Fast Screening for Process Constraints. The objective of step 5 is to identify the most promising phenomena based process options and identify for each of those the corresponding feasible necessary unit operations. Therefore, the remaining phenomena based process options from step 4 are screened by operational constraints (in step 5.1) and, afterward, evaluated by its performance (step 5.2). The remaining most promising options are transformed to unit operations using a set of rules (step 5.3). These options are additionally screened by operational constraints and performance criteria at the unit operation level in step 5.4. This step is necessary because certain constraints are linked to the physical unit operation, e.g. maximum sizes of vessels. Step 5.1: Identify process options satisfying the operational constraints (eq 4) by solving operational constraints (eq 4) and the process model (eq 5) for each of the remaining process

Step 4.1.3: Screen the SPB by connectivity rules for matching operating windows of the involved phenomena (retrieved from the knowledge-base). The number of remaining feasible stage building blocks is NSPB. Step 4.2: Analyze the performance of each SPB to determine the potential connections of those within stages to fulfill a task. This is achieved by applying the substeps 4.2.1−4.2.7. 7133



Article

Table 1. Short Description of the Subalgorithms subalgorithm

short description

MBS (model based search) LBSA (limitation/ bottleneck sensitivity analysis) APCP (analysis of pure component properties) AMP (analysis of mixture properties): AR (analysis of reactions) OPW (operating process window): DS (development of a superstructure) SoP (selection of phenomena) KS (knowledge base search) AKM (apply the extended Kremser method)

Determination of limitations/bottlenecks through a model based search. Quick determination of the limitations/bottlenecks which have the potential to improve the process performance most when this task is enhanced. Collection, generation and analysis of pure component properties based on thermodynamic insights (see section 2.5.3). Collection, generation, and analysis of mixture properties. Analysis of the involved reactions. Determination of the operating window of each phenomenon and of each unit operation. Generation of the superstructure based on the items in the search space Ω. Selection of the initial search space of phenomena and screening for potentially most promising phenomena to enhance a specific task. Retrieval of existing knowledge about PI from a knowledge base tool (see section 2.5.1). Data in the knowledge base can be searched through different keywords simultaneously in a forward or a reverse manner or integrating both. Identify of the configuration of SPBs (co-, counter, crossflow-current) and the number of stages necessary to achieve a defined task in the process (see section 2.5.5).The full algorithm is given in the supplementary material of that paper.

2.4. Subalgorithms. At different steps, the workflow needs a number of subalgorithms (see Figure 4). A short description of each of them is presented in Table 1. Detailed information of the subalgorithms MBS, LBSA, APCP, AMP, AR OPW, DS, SoP, and KS can be found elsewhere.4 The algorithm AKM has not been published before and is given in the supplementary material of that paper. 2.5. Supporting Methods and Tools. A number of supporting methods and tools are needed to apply the described workflow for phenomena based synthesis/design. They are briefly introduced in this section. 2.5.1. Knowledge Base Tool. At different steps of the developed workflow for phenomena based synthesis/design knowledge about existing PI processes, unit operations as well as PI principles is needed. This knowledge is provided by a knowledge base tool. The following main items stored are used in the workflow: list of phenomena and their corresponding operating window; known PI equipment and their corresponding operating window; known PI processes; knowledge for integration of task/phenomena; rules for translation of performance criteria into logical, structural, and operational constraints at different levels of abstraction: unit operation, phenomena; rules for the transformation of phenomena into unit operations. The ontological structure for fast and efficient knowledge storage and retrieval has been explained before.4 Currently, the knowledge base contains around 12000 information items for 42 different phenomena and 135 PI equipment and internals as well as around 200 different process configurations.1,4 2.5.2. Model Library. For this approach, different models in different modeling depths as well as superstructures are necessary. Once developed and validated, the following models and superstructures are stored in a model library: models for phenomena; models for (PI) equipment; superstructure library for connection of units; and superstructure library for connection of phenomena. 2.5.3. Method Based on Thermodynamic Insights.16 In step 3 of the workflow, suitable phase transition phenomena may need to be quickly identified. Here, the method based on thermodynamic insights16 is applied. Based on differences in pure component properties between components targeted to be separated, suitable phase transition phenomena are identified. For example, a large difference in the boiling point

options to determine a feasible set of variables X. This step is divided into four substeps. Step 5.1.1: Determine the degrees of freedom DoF of the model (eqs 4 and 5). Step 5.1.2: Specify process variables X not included in the operational constraints to match DoF for each of the remaining process options. Step 5.1.3: Solve process model (eq 5) to identify a set of process variables satisfying the operational constraints (eq 4). Step 5.1.4: Remove redundant process options by rule 5.1. Step 5.2: Identify the set of most promising options through performance screening NOOP or NPOP respectively: Step 5.2.1: For each of the remaining options, calculate the performance metric Ψp from the set of process variables X determined in step 5.1.3. Step 5.2.2: Rank all process options according to Ψp and select the most promising options. The number of promising process options is NOOP. Step 5.2.3: Check for necessity of screening by Fobj by rule 5.2. Step 5.2.4: Calculate the objective function Fobj. Step 5.2.5: Rank all process options according to Fobj and select the most promising options. The number of promising options is NOOP or NPOP respectively. Step 5.3: Identify unit operations by using the knowledge base tool by using five rules 5.3−5.7. Additionally, a list of earlier translated unit operations and its corresponding additional corresponding operational constraints exists. Step 5.4: Identify process options satisfying the operational constraints and based on this the set of most promising options at the unit operational level. For this, steps 5.1−5.2 are repeated at the unit operation level. 2.3.6. Step 6: Solve the Reduced Optimization Problem. The objective of this step is to identify the best PI process option. This is achieved by optimization of the objective function Fobj with respect to the full mathematical model (eqs 1−5) simultaneously for each remaining process option (step 6.1) and ranking all options by their corresponding performance (step 6.2). Step 6.1: For each process option, solve separately the reduced optimization problem (eqs 1−5). Step 6.2: Rank all options by their objective function and select the best option. 7134



Article

(binary ratio ≫1.05) of two components identifies phase transition by relative volatility as a potential candidate. 2.5.4. Driving-Force (DF) Method.15 From the method based on thermodynamic insights, potentially a large number of phase transition phenomena are identified. In order to screen them the driving force is used which is a measure of the ability to separate components from each other by a two-phase system. The driving force Dij between a binary pair of component i and component j to be separated is defined by the difference in the concentration of one component (here i) between the two phases (eq 10). Dij = yi − xi

with y*0 as the vapor concentration in equilibrium to the entering liquid. Through material balances in each stage from the top to the bottom, the stages are linked through inlet and outlets. The recovery can be derived for cocurrent (eq 15), crossflowcurrent (eq 16), and counter-current (eq 17) flow arrangements by a simple equation. The complete derivation of eqs 15−17 is given elsewhere.17

(10)

If the driving force is small a separation by this phase transition phenomenon is resource demanding (if DAB = 0 the separation of A and B is not possible) and not recommended. 2.5.5. Extended Kremser Method. The Kremser method is used to identify the most promising arrangement of phenomena building blocks and the necessary number of stages to achieve the best or a defined performance (a certain conversion or separation). As stated earlier, SPBs can be connected in cocurrent, counter-current, or crossflow-current arrangements. In order to determine for each SPB, the minimum number of stages for each of the flow arrangements and also to quickly identify the most promising options, an extended Kremser method has been developed. The Kremser method has been selected because • it allows the comparison of three different configurations using the same method; • it covers phase-transition with and without simultaneous reaction by extension; • it is simple to use and needs only phase or reactive phase diagram calculations; • it is a nongraphical method (and therefore not restricted to binary or ternary diagrams); • it can be extended to multicomponent systems. Originally, the counter-current blocks were defined for absorption and stripping processes.17 In the absorption block the feed enters as a liquid, and the desired component transfers from the gas phase into the liquid phase. Similarly, in the stripping block, the feed enters as vapor (stripping agent), and the desired component transfers from the liquid to the gas phase. The phase transition relationship is defined by eq 11. K = yi /xi

SE = (K ·V )/L = 1/AE

(13)

n+1

0

α=1−

1 (1 + AE /N )n

(16)

(17)

y = L /V ·x + D/V ·xD

(18)

Transformation of the distillate flow rate D: y = L /V ·x + (V − L)/V ·xD

(19)

Assuming that V and L are constant, V and F equal and defining a ratio of flows R with

R = L /V

(20)

gives for eq 19 y = R(x − xD) + xD

(21)

Inserting this equation into the definition for the driving force (eq 10) gives y = R(y − xD)/(x − xD)

(22)

In which the x is the equilibrium concentration to y. Stripping section: Feed is liquid (xin = xFeed) A mass balance around the stripping section gives

Absorption and stripping factors (eqs 12 and 13) have been adapted for different phase transition phenomena (see the Supporting Information). In cases of simultaneous reaction and phase transition phenomena, the element based approach is used to generate reactive phase diagrams to calculate the necessary K-values.18 The recovery of n stages (with countercurrent flows) is defined as y − y1 α = n+1 y − y*

(15)

For the application of the extended Kremser method a reliable estimate for the ratio of the two phase flows (L/V) is necessary. Here, the driving force method is used to obtain good initial values for a low energy consumption Rmin. For simplification, it is shown here for the phase transition phenomena by relative volatility. However, the results can be generalized for all types of phase transition phenomena or integrated reactive phase transition phenomena. Two sections may exist in the arrangement: an absorption section in which the feed is vaporized and moves upward and a stripping section on which the feed is liquefied and moves downward. Absorption section: Feed is vaporized (yin = xFeed) A mass balance around the absorption section gives

Additionally, absorption and stripping factors are defined (eqs 12 and 13). (12)

1 1 + AE

⎛ 1 − α /AE ⎞ ⎟ /ln AE n = ln⎜ ⎝ 1−α ⎠

(11)

AE = L /(K ·V )

α=1−

y = L /V ·x + B /V ·xB

(23)

Transformation of the bottom flow rate B: y = L /V ·x + (L − V )/V ·xB

Introducing the ratio of constant vapor and liquid y = R(x − xB) + xB

(24) 21

gives (25)

Inserting this equation into the definition for the driving force (eq 10) gives

(14) 7135


Industrial & Engineering Chemistry Research y = R(y − xB)/(x − xB)

Article

Step 1.3: The process and product specifications are defined. • Raw materials (IPOH, HOAc) are assumed to be pure. • Production capacity is 50000 t IPAc/year. Step 1.4: The following performance metric PM are selected: • Energy consumption per product formed; • Efficiency represented by the conversion of the reactants; • Task constraints: simple (within this metric the number of repetitive units, complexity of the flowsheet, the number of recycles, etc. are restricted); • Volume; • Waste. Step 1.5: All input (θ and PM) are translated into logical, structural, operational constraints, and performance criteria Ψ. The complete translated problem definition is presented in Table 2.

(26)

In which the y is the equilibrium concentration to x. Inserting the concentrations of the feed and the targets for the separation (distillate D and bottom B) into eq 22 and eq 26 determines a reflux ratio which corresponds to the minimum reflux ratio Rmin. A feed position not lying at the driving force means a deviation from the minimum reflux ratio.15 2.5.6. Additional Tools. Several additional tools are used for • the provision of pure component property data (CAPECDatabase); • the provision of mixture data based on thermodynamic models (CAPEC-Database); • the prediction of pure compound properties in case a compound and/or a property is not found in the database (ProPred); • the generation of molecules for solvent selection (ProCAMD); • the analysis of the base case design for sustainability (SustainPro); • the analysis of physical boundaries within nonreactive and reactive SPBs (PDS); • the systematic model derivation, identification, parameter fitting, and analysis (MOT); model simulation and optimization (MOT, ICASSim); All those tools are available in ICAS (Integrated Computer Aided System) developed in CAPEC, Technical University of Denmark, DTU.19,20

Table 2. Problem Definition of the Case Study: IPAc remarks

Fobj design scenario process scenario logical constraints

yield, volume, and pieces of equipment design of PI equipment for the reaction task continuous L1: reaction 1 necessary (rule 1.1) L2: reaction 1 has to be in the first stage (rule 1.2) Efficiency: S1: Do not integrate units which inhibit each others performance. S2: Add phenomena and stages to the position in the flowsheet in which it has the highest efficiency. S3: Always end the flowsheet with the phenomena giving the highest yield last. Energy: S4: Do not provide energy to streams without purpose S5: Do not connect units with alternating heat addition and heat removal. Task constraints: S6: Do not use repetitive (sequential) units. S7: Do not use prereactors. S8: Do not use recycle streams if not necessary (when efficiency can be reached) S9: Remove options in which stages are redundant raw materials are pure Use raw materials according to the stoichiometry of the reaction(s) (waste) yield = 0.99 (phenomena level)

structural constraints by PI metric

3. CASE STUDY The application of the methodology is highlighted through the production of isopropyl acetate (IPAc). Isopropyl acetate is an important bulk chemical process product used widely as organic solvent. Improvements in terms of its production with respect to operational process costs through PI are of high interest.21 3.1. Base-Case Design. Isopropyl acetate can be produced by a reaction of acetic acid (HOAc) and isopropanol (IPOH) to form isopropyl acetate and water (H2O) as side-product. The stoichiometry of the reaction is given in eq 27 and takes place in the liquid phase. The reaction is catalyzed by a heterogeneous catalyst, which is Amberlyst 15.21 CH3COOH + C3H7OH ↔ C5H10O2 + H2O

parameter

operational constraints (rule 1.6)

(27)

operational constraints (rule 1.7) PI screening criterion for step 5

The base-case design is a single CSTR producing 50000 t/y of IPAc. The CSTR is assumed to run under isothermal conditions (T = 330 K at P = 1 bar). The feed to the reactor is an equimolar mixture of the reactants HOAc and IPOH (nHOAc,in = nIPOH,in = 1430 mol/min). The reaction is not complete (nHOAc,out = nIPOH,out = 499 mol/min). The reaction is exothermic. Hence, the heat of reaction has to be removed, which is in total 16.7 MJ/min. 3.2. Step 1: Define Problem. Step 1.1: The objective is to identify one intensified apparatus for the reaction that achieves a yield of 0.99 and with a corresponding lowest operational costs and capital costs. The operational costs are represented by the yield, thermal energy, and additional utility costs (membrane, solvents), while the capital costs are represented by the volume needed by the reaction-operation. Note: The product purity is not defined since only the reaction task is under investigation. Step 1.2: The design scenario is the development for a new reactor. The process scenario is continuous.

thermal energy consumption (phenomena level)

3.3. Step 2: Analyze the Process. Step 2.1: Collect mass and energy data for the base-case design: Step 2.1.1: Energy data is required (rule 2.1).Step 2.1.2: Mass and energy data are collected (see section 3.1). Step 2.2: Transform the flowsheet into a task based flowsheet and a phenomena based flowsheet. Step 2.2.1: Identify for each unit operation of the base-case design the task. Using rules 2.2−2.8, a reaction task is identified (rule 2.2). Step 2.2.2: Identify the split of each separation task. No separation is identified. Step 2.2.3: The list of phenomena involved in the CSTR has been retrieved from the PI knowledge-base which are ideal 7136



Article

been computed using the NRTL model with parameters given by Lai et al.23 In a ternary mixture with isopropanol or acetic acid, the concentration of the third component has to be at least 0.38 mol/mol or 0.27 mol/mol respectively to potentially enter the 2-phase area (see Figure A.1 in Appendix A.1). However, the water phase is almost pure water. Hence, the immiscibility effect on the reaction is not negative and is therefore not identified as a limitation. Step 2.4.3: The reaction is analyzed using AR. The reaction model has been retrieved from Sanz and Gmehling24,25 and is given in the Appendix A.2. The reaction is slightly exothermic and only limited by the equilibrium of the reaction phenomenon. This is added to the list of keywords for targets for PI (KPI). Step 2.4.4: The operating window of the reaction task is identified by using the subalgorithm OPW. It is limited by the operating window of two phenomena: the reaction and the liquid phase mixing: The maximum temperature is T = 403 K to ensure catalyst stability;26 the lowest temperature is the lowest melting point of the component in the system. Concentrations at fixed pressure and temperature have to be below the dew point line. Step 2.5: This step (linking of limitation to a source outside of the task) is not necessary because only one task is under investigation and the raw materials are assumed to be pure. 3.4. Step 3: Identification of Desirable Phenomena. Step 3.1: Collect PI for identified targets (KPI): Step 3.1.1: The list of keywords KPI for identification of PI possibilities includes the following: Unfavorable equilibrium in the reaction. Step 3.1.2: The algorithm KBS to identify replacement of reaction. No potential has been identified through the knowledge base. Step 3.1.3: The algorithm KBS is applied to identify additional tasks for enhancement of the necessary one. To overcome the limitation on the necessary task integration of reaction with a second reaction and/or a separation task is identified (see Table 5). Step 3.2: The algorithm APCP is used to identify potential phenomena blocks for each identified additional desirable task

liquid phase mixing, pseudohomogeneous reaction, and convective cooling. Step 2.3: Identify limitations/bottlenecks LB of the basecase. Step 2.3.1: Collect the limitations of the process by applying the algorithm KBS. Step 2.3.1.1: The list of keywords K contains the following items: K={production of IPAC, esterification reaction, IPOH, HOAc, IPAc, H2O}. Step 2.3.1.2: The algorithm KBS is applied for the knowledge base search. The keyword “esterification reaction” has been found in the knowledge base. For this, the following information about limitations/ bottlenecks is stored: unfavorable equilibrium in the reaction. Step 2.3.2: The algorithm MBS is applied. It confirms that the LB is identified to be in the reaction. Step 2.3.3: The number of limitations/bottlenecks (LB=1) is below 15, no reduction is required (rule 2.11). Step 2.4: Analyze the obtained limitations/bottlenecks LB and their corresponding tasks in the base-case design. Step 2.4.1: The algorithm APCP is applied. The pure component properties are retrieved from the ICAS database (Table 3). The binary ratio of the occurring components for Table 3. List of Pure Component Properties of the System at P = 1 atma component

TB [K]

TM [K]

Log(Kow) [-]

HOAc IPOH IPAc H2O

391.05 355.41 361.65 373.15

289.91 185.28 199.75 273.15

−0.15* 0.53* 1.32* −1.38

Rg [Å]

VM [m3/ kmol]

VVdW [m3/ kmol]

2.61 2.807 3.679 0.615

0.1797 0.22 0.336 0.0559

0.03 0.04 0.06 0.01

“*”: Missing data predicted through the Marrero-Gani-Approach;22 TB: boiling point, TM: melting point, Log(Kow): octanol-water partition coefficient, RG: radius of gyration, VM: molar volume, VVdW: van der Waals volume. a

Table 4. List of the Binary Ratio of Some Pure Components Properties binary mixture

TB

TM

Rg

VM

Log(Kow)

H2O/HOAc H2O/IPOH H2O/IPAc HOAc/IPOH HOAc/IPAc IPOH/IPAc

1.05 1.05 1.03 1.10 1.08 1.02

1.06 1.47 1.37 1.56 1.45 1.08

4.24 4.56 5.98 1.08 1.41 1.31

3.21 3.94 6.01 1.22 1.87 1.53

9.20 2.49 −1.05 −3.53 −0.11 0.40

Table 5. Decision Table Regarding PI Solutions to Reported Limitations in the Processa parameter

remarks reaction separation

necessary task desirable task limitation in a phenomenon: reaction is not sufficient contact problems of reactants product decomposes activation problems degradation by T degradation by pH heat supply/removal difficult limiting equilibrium

each property is calculated (see Table 4). The binary ratios for the octanol−water partition coefficient (Log(Kow)) hint that a potential phase split may occur between the binary pairs H2OIPAC, H2O-HOAc, and HOAc-IPOH. Hence, further investigation of mixture properties in step 2.4.2 is necessary. Step 2.4.2: The algorithm AMP is applied to check for azeotropes as well as miscibility gaps between the components. Several low boiling azeotropes of ternary mixtures and binary pairs are formed (see Appendix A.1). Furthermore, a miscibility gap exists between water and the isopropyl acetate that has

reaction heat supply

reaction mixing

reaction reaction

○ ○ ○ ○

○

○ ○

○

○

●

○ ●

“○”indicates intensified option is reported to overcome this limitation, “●” indicates the activation of an option through a search of the knowledge base.

a

7137



Article

Table 6. Identified Phenomena for Each Desirable Task Using the Algorithm APCP and AMP with Pure Component Propertiesa and Result of the Screening of Phenomena for Each Desirable Task Using the Algorithm SoPb task

identified phenomena

pervaporation solid−liquid phase transition liquid−liquid phase transition ... separation IPac/rest pervaporation ... separation IPAc and H2O/ relative volatility rest ... reaction reaction with H2O

reaction with IPAc

how determined? Rg, VM, VVdW TM

separation H2O/rest

screened out not matching operating window with necessary task

Log(KOW)

no additional solvent used when not necessary (Green Chemistry Principle)

Rg, VM, VVdW

suitable membrane not in database

TB, PLV

note: possible until the ternary azeotrope

no additional reaction found in the database not desired

reaction

a

TM: melting point; TB: boiling point; PLV: vapor pressure; Rg: radius of gyration; VM: molar volume; VVdW: van der Waals volume; Log(KOW): octanol-water partition coefficient. bThe remaining phenomena and tasks kept in the search space are written in bold.

Table 7. Conversion of Connected Stages for Different Flow Configurationsa co 1,2 phases

SPBs SPB.1, SPB.3−4, SPB.10−14, SPB.20, SPB.22−23, SPB.29−33, SPB.39,SPB.41−42,SPB.48−52, SPB.58 SPB.2, SPB.21, SPB.40 SPB.5, SPB.24, SPB.43 SPB.6, SPB.25, SPB.44 SPB.7, SPB.10, SPB.13, SPB.15, SPB.18, SPB.26, SPB.29, SPB.32, SPB.34, SPB.37, SPB.45, SPB.48, SPB.51, SPB.53, SPB.56 SPB.8, SPB.16, SPB.19, SPB.27, SPB.46, SPB.35, SPB.54, SPB.38, SPB.57 SPB.9, SPB.17, SPB.28, SPB.47, SPB.36, SPB.55

no reaction 0.58/∞ 0.55/∞ 0.65/∞ 0.57/∞ 0. 67/∞ 0.99/1

crossflow 2 phases no reaction 1-phase 1-phase 1-phase no separate phases 0. 69/∞ 0.99/1

counter 2 phases no reaction 1-phase 1-phase 1-phase no separate phases 0.81/13(∞) 0.99/1

a In the cells, for example, SPB.2 and cocurrent flow, the first number “0.58” indicates the conversion, while the second number “∞” indicates the number of stages necessary to achieve that conversion.

and A.3. The phase transition by relative volatility is described using the NRTL model (Lai et al.23). Step 3.6: All accompanying phenomena are selected by consulting the knowledge base for each identified phenomena. In total 13 phenomena (PB.1-PB.13) are identified: mixing (liquid (L): ideal (Mid), tubular flow (Mtub), rectangular flow (Mrec); vapor (Mv): ideal; 2-phase V-L: ideal (2phM)); dividing (D); convective heating (H) and cooling (C); heterogeneous reaction (R); ideal phase contact of V-L (PC); ideal phase separation of V and L (PS). Step 3.7: The operating windows are determined applying the algorithm OPW. The maximum temperature to avoid catalyst degradation of Amberlyst 15 is 403 K26 which is set to be the maximum limit of the operational window of the liquid phase reaction phenomenon. The membrane is assumed to be stable until T = 347 K or Pmax=2 bar. 3.5. Step 4: Generate Feasible Operation/Flowsheet Options. Step 4.1: Generate all feasible SPBs from the phenomena: Step 4.1.1: The maximum number of phenomena within an SPB is nP,max = 13−1−2−1 = 9. Note: Heating and cooling (subtract 1 from total number of phenomena), liquid mixing patterns flow and perfectly mixed cannot be connected within one SPB (subtract 2). Additionally, the dividing phenomenon is one SPB itself (subtract 1). Step 4.1.2: The total number of phenomena is nP,tot,=13 which gives a total number of 4019 generated SPBs (eq 28).

by analysis of the pure component properties of the components in the process. In total 21 phenomena are identified. An exemplary sample of the identified phenomena is given in Table 6. Step 3.3: The algorithm AMP is used to identify potential phenomena blocks for each identified additional desirable task by analysis of the mixture properties of the components in the process. No potential building blocks are identified. Step 3.4: Use algorithm AR to identify potential reaction phenomena blocks for each task from analysis of pure component properties. Step 3.4.1: An additional reaction task is identified. Step 3.4.2 is entered (rule 3.1). Step 3.4.2: No reactions have been found in the literature/ database. Step 3.4.3: No additional reaction phenomenon is created (rule 3.2). Hence, step 3.5 is entered. Step 3.5: Select potential best phenomena for each identified additional desirable task: A partial result of the screening using the algorithm SoP is also given in Table 6. Here, the mixture data is reused from step 3.3. Azeotropes have been found limiting the operating window of the phase transition phenomenon by relative volatility (see Table A.1 in Appendix A.1). In total, two phenomena (highlighted in bold) are kept in the search space which are phase transition by relative volatility and by pervaporation. Note: The phase transition by pervaporation is described using a selected membrane.27 The models for reaction and phase transition by pervaporation are given in Appendixes A.2 7138


Industrial & Engineering Chemistry Research nP,max

NSPBmax =

∑ k=1

Article

⎛ (nP , tot − 1)! ⎞ ⎜⎜ ⎟⎟ + 1 = 4019 ⎝ (nP , tot − k − 1) ! k! ⎠

Step 4.5: The result of the screening through structural constraints is given in Table 8. Remaining process options are NOOS = 506. 3.6. Step 5: Fast Screening for Process Constraints. Step 5.1: Solve operational constraints (eq 4) and the process model (eq 5) at the phenomena level. The yield is defined as operational constraint (see step 1, Table 2). A selection of the results for some process options is given in Table 9. The number of remaining process options achieving a yield of 0.99 is reduced to NOOO = 118.

(28)

Step 4.1.3: Using connectivity rules and the information of the operating window of each phenomenon, a total number of 58 SPBs are feasible in terms of conditions of the operating windows of the integrated phenomena. All 58 feasible SPBs are listed in Appendix A.4. Step 4.2: Analyze and determine potential configurations: Step 4.2.1−Step 4.2.5: For each reactive SPB to the conversion is calculated for different flow arrangements using the extended Kremser method. Step 4.2.6: The results of the performance calculations are presented in Table 7. The minimum number of stages is 1. Crossflow is selected (Rule 4.7). Step 4.2.7: The number of stages within one process is between 1 and 3 (Rule 4.10). Step 4.3: Generate the number of feasible operation options for each separate task: Step 4.3.1: A crossflow arrangement for 3 stages is retrieved from the model library with seven possible recycle streams for the liquid phase which depends on the position of the dividing phenomenon in the superstructure. Step 4.3.2: The number of generated process options is 218892 (eq 29) using eq 9. The number of forward connections for processes containing 3 stages enabled by the dividing phenomenon are nR,3=6 and for processes containing 2 stages nR,2=3.

Table 9. Excerpt of the Screening by Operating Constraints and Performance Screening by the Thermal Energy Consumption

0.99

0.36

0.99 ... 0.99

0.36 ... >0.36

...

...

Step 5.2: The thermal energy consumption is used as a performance criterion for identification of the most promising options NOOP: The results of the calculation are given in Table 9. In total 22 process options (NOOP = 22) represented by phenomena remain in the search space. Step 5.3: Now, the phenomena based processes are transformed to unit operations using rules 5.3−5.8 and presented in Figure 5. Step 5.4: All options are screened for operational constraints at the unit operation level. That is that the reactor volume of a single unit cannot exceed 50 m3 and that the pressure drop using the membrane cannot exceed the maximum pressure of P = 2 bar. The following assumptions are added: • Temperature of the reactants into the reactor is T = 330 K; • Input flow for each reactant is 941 kmol/min; • Physical limits of the CSTR: A/V=2/H with H/rmax=20; • Maximum amount of catalysts: Flow reactor: 100 kg/m3 catalyst; CSTR: 10 wt.% catalyst. The pressure drop is calculated using eq 30.

Step 4.3.3: The screening in terms of feasible connections of SPBs, that is, a two-phase outlet cannot enter a one-phase SPB and by checking the correctness of the potential additional forward connections through the dividing phenomenon gives NOO = 121610 formally feasible options remaining in the search space. Step 4.4: The result of the screening through logical constraints is presented in Table 8. Remaining options are NOOL = 24142. Table 8. Overview of the Search Space Reduction through Screening by Logical and Structural Constraints

energy redundant operations are removed ; e.g. −H−C− (S5) remove all options with redundant phenomena not improving the yield after reaction phenomenon/phenomena (S1, S2, S4, and S9) phenomena blocks with highest effect on improving the yield are last (S3) process in 1 unit operation possible (S6 and S7) no external recycle (S8)

yield

-MFl,tub=MV=R=H=2phM=PC=PT(PVL) =PS(VL)MFl,tub=MV=R=H=2phM=PC=PT(PVL) =PS(VL)MFl,tub=MV=R=H=2phM=PC=PT(PVL) =PS(VL)-MId=MV=R=H=2phM=PC=PT(PVL)=PS(VL)... -MId,tub=MV=R=H=2phM=PC=PT(PVL) =PS(VL)-M=R=C-D ...

...

(29)

constraint

phenomena based representation

#1

#1 ... #23

NOOmax = 583 + 582 + 581 + 6·582 + 3·581 = 218892

formally feasible operations: product is formed (L1) reaction phenomenon before or simultaneously with phase transition phenomenon (L2) process potentially acts within operating window

rank

thermal energy [MW]

no. of remaining options

Δp = ζ /Re ·ρ /2w 2L /r

121610 102424 64179

(30)

The Reynolds number is calculated with Re = wd /ν

(31)

For the rectangular channel the corresponding radius is determined via the calculation of the hydraulic diameter (eq 32) A dh = 4 (32) P with A being the cross sectional area of the channel, and P being the perimeter of the cross-section. The resistance factor ζ in the channel depends on the Reynolds number and on the shape of the channel. For

NPOL = 24142 12244 11153 7619 518 NPOS = 506 7139



Article

Figure 5. Identified unit operations.

Figure 6. Detailed design of the rectangular plate-frame-flow reactor-pervaporator (width/height = W/H = 1000).

example, for a laminar flow (Re < 2300), the resistance factor is ζ = 64 for a tube and ζ = 80 for a rectangular channel.28 In the simulation, each stage runs until 90% of the maximum limit and the last reactive stage reaches until the desired yield. In total two options, option 10 (Vtotal = 141 m3, Vsingle = 1.41 m3) and option 22 (Vtotal = 242 m3, Vsingle = 2.42 m3) are kept in the search space. All other options are screened out because the

necessary volume did not match unit operational constraint (volume of a single reactor below 50 m3) to reach the yield of 0.99 at a pressure drop below Δp = 1 bar to ensure membrane stability. 3.7. Step 6: Solve the Reduced Optimization Problem. Steps 6.1−6.2: For each process option, the reduced optimization problem is now solved. 7140



Article

Table 10. Simulation Results for Both Remaining Process Options #10 and #22a rectangular channel

a

option

T [K]

parallel units nparallel [-]

W/H

H [m]

W [m]

Amembrane/V [m−1]

Vsingle [m3]

r or rh [m]

Δp [bar]

#10 #22

373 373

100 20

1000*

0.0021

2.1

100 476

1.46 0.75

0.01 0.0021

1.0 1.0

W: width; H: height; “*”: at boundary constraint.

ical) synthesis problem is tackled by a detailed workflow in an efficient and systematic manner. Within this workflow, the links to subalgorithms, rules, supporting tools, and methods are established. The phenomena based approach presented in this paper accompanies the unit operation based approach for process intensification published earlier.4 It establishes that within the PI synthesis/design framework, intensified solutions can be obtained at all three levels (scales): process, unit operation, and phenomena. The developed methodology depends on the reliability and availability of the models and the certainty of the parameters. It is for sure that the optimal solution may not be the same if uncertainties in the model parameters were included. This, however, is beyond the scope of this paper. The integration or enlargement of this method which take the uncertainty of the model parameters into account would be highly beneficial to increase the confidence into the theoretical obtained results. Through several case studies, it has been confirmed that a limitation in a process is often occurring in or caused by the reaction step. Therefore, it may be beneficial to couple the methodology with a tool which identifies all possible reaction routes. Additionally, up to now, the processes have been evaluated using the performance metric developed for PI and single criterion objective functions. However, important more complex methods to evaluate sustainability such as life cycle assessment (LCA) in which trade-offs between different criteria are taken into account as well as methods predicting capital costs of novel equipment in order to incorporate capital costs may be useful for evaluation of different generated process options.

Additional design constraints: • Re > 20 • Maximum amount of catalyst: 100 kg/m3 • The ratio of width over height in the rectangular channel maximum W/H = 1000. Note: The width of the rectangular channel is the one which is parallel to the membrane and the heat exchange plate, while the height is vertical to membrane and heat exchange plate (see Figure 6). Hence, the membrane area is the product of width multiplied with the length of the channel. • The maximum number of flow channels is nparallel = 100 channels. The objective function is minimized with respect to the diameter (and/or volume) of the tube/channel to achieve theoretically a conversion of 0.99, a purity of 98 mol.% IPAc in the outlet, and a pressure drop Δp < 1 bar for membrane stability reasons. The results of the optimization are given in Table 10. The best option is the plate-and-frame heat exchanger-reactor-pervaporator with a rectangular channel (option #22, Figure 6). 3.8. Comparison to Conventional and Known Intensified Solutions. The new design is benchmarked against data obtained for a reactive distillation system23 as well as the data for a conventional base case process.29 The reactive distillation system proposed by Lai et al.23 consists of one reactive distillation column with a decanter at the top of the column and an external stripping column. The distillate of the reactive distillation is near the ternary azeotrope composition (see Appendix A.1). The final product is obtained as the bottom product from the stripping column. Detailed data of the process are known.23 The conventional base case process is more complex in terms of the number of unit operations because the reaction is incomplete, and the separation is difficult due to the number of azeotropes in the system. The base case design found in Corrigan and Stichweh29 consists of one reactor, six distillation systems, one extractor using water as a solvent, and one decanter. With the phenomena based synthesis/design methodology, a conversion of 99% of IPOH is achieved compared to around 94% for base case as well as 93% for the reactive distillation system. The thermal heat requirement of the novel design is only 1/10th of the thermal heat requirement of the conventional process and around 1/5th of the reactive distillation system. The amount of catalyst in the novel design can be reduced by 40% compared to the reactive distillation system.

■

APPENDIX Within this Appendix, additional information for the case study is given. A.1. Azeotropes and Miscibility Gaps

In step 2, one ternary azeotrope, four binary azeotropes (Table A.1) and miscibility gaps (Figure A.1) are identified for the system. A.2. Reaction Kinetics

The reaction rate for the production of isopropyl acetate from isopropanol and acetic acid can be expressed through eq A1 with parameters taken from Sanz and Gmehling:24,25 Table A.1. List of the Identified Azeotropes in the Mixture at P = 1 atma

4. CONCLUSIONS A phenomena based methodology for synthesis/design to achieve PI has been developed and tested through a case study. With this approach it is possible to generate potentially novel process options. Use of truly and reliable predictive models has provided predictive solutions. Using the decomposition approach as the solution procedure, the complex (mathemat-

ternary/binary pair

TB [K]

x1

x2

x3

IPOH/IPAc/W IPAc/W IPOH/IPAc IPOH/W W/HOAc

347.37 349.72 351.69 355.65 371.97

0.2377 0.5981 0.5984 0.6875 0.8384

0.4092 0.4019 0.4016 0.3125 0.1616

0.3531

Min.BP Min.BP Min.BP Min.BP Min.BP

a

Min.BP = low boiling azeotrope. Compositions are given in mol. Data are from Lai et al.23

7141



Article

Figure A.1. Ternary LLE diagrams of this system at P = 1 atm using an NRTL model.23

Table A.2. List of Feasible SPBsa SPB SPB.1 SPB.2 SPB.3 SPB.4 SPB.5 SPB.6 SPB.7 SPB.8 SPB.9 SPB.10 SPB.11 SPB.12 SPB.13 SPB.14 SPB.15 SPB.16 SPB.17 SPB.18 SPB.19 SPB.20, SPB.21, SPB.22, SPB.23, SPB.24, SPB.25, SPB.26, SPB.27, SPB.28, SPB.29, SPB.30, SPB.31, SPB.32, SPB.33, SPB.34, SPB.35, SPB.36, SPB.37, SPB.38, SPB.58 a

SPB.39 SPB.40 SPB.41 SPB.42 SPB.43 SPB.44 SPB.45 SPB.46 SPB.47 SPB.48 SPB.49 SPB.50 SPB.51 SPB.52 SPB.53 SPB.54 SPB.55 SPB.56 SPB.57

connected phenomena

in

out

MId MId=R MId=R MId=C MId=R=H MId=R=C MId=MV=R=2phM=PC=PT(VL) MId=MV=R=2phM=PC=PT(VL)=PS(VL) MId=MV=R=2phM=PC=PT(PVL)=PS(VL) MId=MV=H=2phM=PC=PT(VL) MId=MV=H=2phM=PC=PT(VL)=PS(VL) MId=MV=H=2phM=PC=PT(PVL)=PS(VL) MId=MV=C=2phM=PC=PT(VL) MId=MV=C=2phM=PC=PT(VL)=PS(VL) MId=MV=R=H=2phM=PC=PT(VL) MId=MV=R=H=2phM=PC=PT(VL)=PS(VL) MId=MV=R=H=2phM=PC=PT(PVL)=PS(VL) MId=MV=R=C=2phM=PC=PT(VL) MId=MV=R=C=2phM=PC=PT(VL)=PS(VL) MFl MFl=R MFl=H MFl=C MFl=R=H MFl=R=C MFl=MV=R=2phM=PC=PT(VL) MFl=MV=R=2phM=PC=PT(VL)=PS(VL) MFl=MV=R=2phM=PC=PT(PVL)=PS(VL) MFl=MV=H=2phM=PC=PT(VL) MFl=MV=H=2phM=PC=PT(VL)=PS(VL) MFl=MV=H=2phM=PC=PT(PVL)=PS(VL) MFl=MV=C=2phM=PC=PT(VL) MFl=MV=C=2phM=PC=PT(VL)=PS(VL) MFl=MV=R=H=2phM=PC=PT(VL) MFl=MV=R=H=2phM=PC=PT(VL)=PS(VL) MFl=MV=R=H=2phM=PC=PT(PVL)=PS(VL) MFl=MV=R=C=2phM=PC=PT(VL) MFl=MV=R=C=2phM=PC=PT(VL)=PS(VL) D

1..n(L) 1..n(L) 1..n(L) 1..n(L) 1..n(L) 1..n(L) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1..n(VL) 1..n(VL) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1..n(L) 1..n(L) 1..n(L) 1..n(L) 1..n(L) 1..n(L) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1..n(VL) 1..n(VL) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1..n(L,VL) 1(L;VL,V)

1(L) 1(L) 1(L) 1(L) 1(L) 1(L) 1(V/L) 2(V;L) 2(V;L) 1(VL) 2(V;L) 2(V;L) 1(VL) 2(V;L) 1(V/L) 2(V;L) 2(V;L) 1(V/L) 2(V;L) 1(L) 1(L) 1(L) 1(L) 1(L) 1(L) 1(V/L) 2(V;L) 2(V;L) 1(VL) 2(V;L) 2(V;L) 1(VL) 2(V;L) 1(V/L) 2(V;L) 2(V;L) 1(V/L) 2(V;L) 1..n(L;V; VL)

The flow mixing MFl can be a flow within a tubular or a rectangular channel flow pattern; hence, those represent two SPBs.

7142



Article

⎛ −64.59 ⎞ ⎟a a r = 1.02 × 107exp⎜ ⎝ RT ⎠ HOAc IPOH ⎛ −73.63 ⎞ ⎟a a − 1.90 × 107exp⎜ ⎝ RT ⎠ IPAC w

The activity of each component is ai = xiγi

(5) Papalexandri, K. P.; Pistikopoulos, E. N. Generalized modular representation framework for processs synthesis. AIChE J. 1996, 42, 1010. (6) Peschel, A.; Freund, H.; Sundmacher, K. Methodology for the design of optimal chemical reactors based on the concept of elementary process functions. Ind. Eng. Chem. Res. 2010, 49 (21), 10535. (7) Siirola, J. J. Strategic process synthesis: advances in the hierarchical approach. Comput. Chem. Eng. 1996, 20(S), s1637. (8) Rong, B. G.; Kolehmainen, E.; Turunen, I. Methodology of conceptual process synthesis for process intensification. Comput.-Aided Chem. Eng. 2008, 25, 283. (9) Arizmendi-Sánchez, J.; Sharratt, P. Phenomena-based modularisation of chemical process models to approach intensive options. Chem. Eng. J. 2008, 135, 83. (10) Algusane, T. Y.; Proios, P.; Georgiadis, M. C.; Pistikopoulos, E. N. A framework for the synthesis of reactive absorption columns. Chem. Eng. Process. 2006, 45, 276. (11) Lutze, P.; Babi, D. K.; Woodley, J. M.; Gani, R. Phenomenabased synthesis and design to achieve process intensification. Comput.Aided Chem. Eng. 2012, 31, 1697. (12) Karunanithi, A. T.; Achenie, L. E. K.; Gani, R. A new decomposition-based computer-aided molecular/mixture design methodology for the design of optimal solvents and solvent mixtures. Ind. Eng. Chem. Res. 2005, 44, 4785. (13) D’Anterroches, L.; Gani, R. Group contribution based process flowsheet synthesis, design and modeling. Fluid Phase Equilib. 2005, 228−229, 141. (14) Harper, P. M.; Gani, R. A multi-step and multi-level approach for computer aided molecular design. Comput. Chem. Eng. 2000, 24, 677. (15) Bek-Pedersen, E.; Gani, R. Design and synthesis of distillation systems using a driving-force-based approach. Chem. Eng. Process. 2004, 43, 251. (16) Jaksland, C. A.; Gani, R.; Lien, K. M. Separation process design and synthesis based on thermodynamic insights. Chem. Eng. Sci. 1995, 50, 511. (17) Seader, J. D.; Henley, E. J. Separation process principles; John Wiley & Sons Inc.: New York, 1998. (18) Perez-Cisneros, E. S.; Gani, R.; Michelsen, M. L. Reactive separation system - I. Computation of physical and chemical equilibrium. Chem. Eng. Sci. 1997, 52 (4), 527. (19) Gani, R.; Hytoft, G.; Jaksland, C.; Jensen, A. K. An integrated computer aided system for integrated design of chemical processes. Comput. Chem. Eng. 1997, 21 (10), 1135. (20) Heitzig, M.; Sin, G.; Sales Cruz, M.; Glarborg, P.; Gani, R. Computer-aided model framework for efficient model development, analysis and identification: combustion and reactor modeling. Ind. Eng. Chem. Res. 2011, 50 (9), 5253. (21) Tang, Y. T.; Chen, Y. W.; Huang, H. P.; Yu, C. C.; Hung, S. B.; Lee, M. J. Design of reactive distillation for acetic acid esterification. AIChE J. 2005, 51 (6), 1683. (22) Marrero, J.; Gani, R. Group-contribution based estimation of pure component properties. Fluid Phase Equilib. 2001, 1−2, 183. (23) Lai, I. K.; Hung, S. B.; Hung, W. J.; Yu, C. C.; Lee, M. J.; Huang, H. P. Design and control of reactive distillation or ethyl and isopropyl acetates production with azeotropic feeds. Chem. Eng. Sci. 2007, 62, 878. (24) Sanz, M. T.; Gmehling, J. Esterification of acetic acid with isopropanol coupled with pervaporation. Part I: Kinetics and pervaporation studies. Chem. Eng. J. 2006, 123, 1. (25) Sanz, M. T.; Gmehling, J. Esterification of acetic acid with isopropanol coupled with pervaporation. Part II. Study of a pervaporation reactor. Chem. Eng. J. 2006, 123, 9. (26) Dow. Technical data sheet of Amberlyst 15. http://www.dow. com (accessed 15.08.2011). (27) Van Hoof, V.; Dotremont, C.; Buekenhoudt, A. Performance of Mitsui NaA type zeolite membranes for the dehydrogenation of

(A1)

(A2)

The molar fractions of the liquid can be determined through the change of moles of the component i over the overall moles in the batch reactor (eq A3). n xi = i ∑ ni (A3) A.3. Model of the Phase Transition by Pervaporation

The pervaporation membrane (Celfa CMC-VS-11 V) is described by the flux in [g m−2 h−1] (eq A4) and a separation factor of water from other components in the system to be 99.5 (eq A5) based on experimental data.27 It is assumed that the flux of water through the membrane only depends on the concentration of water in the feed side. JW = a ·xW 2 + b·xW

(A4)

With the parameters a = 12.331 and b = 185.85. αi , w = 99.5

(A5)

A.4. List of Feasible SPBs

In step 4, a large set of SPBs is screened for feasibility. In total 58 feasible SPBs are remaining in the search space which are presented in Table A.2. The notation is similar to the SMILES code (see chapter 2.2).

■

ASSOCIATED CONTENT

S Supporting Information *

Additional information for the application of the workflow is given which includes the total list of phenomena in the phenomena library, the list of rules used in the workflow, and the algorithm for applying the Kremser method (AKM). This material is available free of charge via the Internet at http:// pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Phone: +49 (0) 231 755-2670. Fax: +49 (0) 231 755-3035. Email: [email protected]. Notes

The authors declare no competing financial interest.

■

REFERENCES

(1) Lutze, P.; Gani, R.; Woodley, J. M. Process intensification: a perspective on process synthesis. Chem. Eng. Process. 2010, 49, 547. (2) Van Gerven, T.; Stankiewicz, A. Structure, energy, synergy, timethe fundamentals of process intensification. Ind. Eng. Chem. Res. 2009, 48, 2465. (3) Harmsen, J. Process intensification in the petrochemicals industry: drivers and hurdles for commercial implementation. Chem. Eng. Process. 2010, 49, 70. (4) Lutze, P.; Roman-Martinez, A.; Woodley, J. M.; Gani, R. A systematic synthesis and design methodology to achieve process intensification in (bio) chemical processes. Comput. Chem. Eng. 2012, 36, 189. 7143



Article

organic solvents in comparison with commercial polymeric pervaporation membranes. Sep. Purif. Technol. 2006, 48, 304. (28) Käst, W. VDI-Wärmeatlas. Kapitel 9.5.1 Druckverlust bei der Strö mung durch Rohre und Kanäle: Springer Verlag: Berlin, Heidelberg, New York, 2002. (29) Corrigan, T. E.; Stichweh, L. A. Esterification process development. Chem. Eng. Sci. 1968, 23, 991.

7144


Phenomena Based Methodology for Process Synthesis Incorporating

Recommend Documents