Scribd Logo
Upload

Welcome to Scribd!

  • 31 views

    07 Chemical Process Simulation PDF

    Uploaded by

    Shubham Singh

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content

    You might also like

    07 Chemical Process Simulation PDF

    Uploaded by

    Shubham Singh
    31 views16 pages

    Original Title

    44822675-07-Chemical-Process-Simulation.pdf

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    Download as pdf
    31 views16 pages

    07 Chemical Process Simulation PDF

    Uploaded by

    Shubham Singh

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    Download as pdf
    of 16
    7 Chemical Process Simulation 7.1 INTRODUCTION Inthe preceding chapters, solution of mathematical models of specific unit ‘operations was described. A process plant consists of several processing "units interconnected by process streams. Design of a process plant includes not only solving the material and energy balance equations for all process units, but also equipment sizing subject to design constraints, costing and optimization. A single computer package that can perform all the above tasks for a given process is known as a computer aided process design package or simply a process flowsheeting package. Individual process unit calculations are treated as building blocks or as equation sets that make up the overall process flowsheet. We have seen how individual building blocks could be simulated on the computer. When all the blocks are interconnected the solution becomes more complex. Most process plants contain recycle streams and control loops, and solution of stream properties warrants iterative calculations. Efficient numerical methods are required for convergence. In addition, for the calculation of process unit performance, appropriate physical properties and thermodynamic data are required. Finally, there should be a ‘master program that links al the building blocks, physical property data, thermodynamic calculations package and numerical routines and that supervises information flow. Optimization and economic analysis complete the picture, Although process flowsheeting programs were developed as a design tool, many of the programs operate inthe simulation or performance calculations mode. In these programs, design calculations have to be performed by what is known as iterated simulation, or running the simulation program for dillerent values of system parameters until design conditions are met ‘Computer simulation of steady state chemical processes is over two 00 (CHEMICAL PROCESS SIMULATION iol decades old. A review of the historical development can be found in Kehat land Shacham.' The literature on this subject has grown so rapidly in the last decade that a complete discussion is beyond the scope of this book. Several good reviews have appeared from time to time.'~* The principles of process flowshecting are described by Westerberg eta.” In this chapter, 1 brief outline of the principles behind steady state process simulation is provided. The scope covers the basic concepts used in flowsheeting, decomposition of large systems, package programs available to the process engineer and some future trends. 7.2. PROCESS SIMULATION TECHNIQUES Since the flow diagram of a process consists of a number of process units ‘with interconnections, the mathematical representation of the system includes the following. (a) Interconnection equations; these are the basic material and energy balance equations. (b) Model equations: the model can be that of a mixer. reactor, heat exchanger, distillation column, etc. These have been dealt with in Chapters 3-6. ‘Three main techniques are available for simulating a process.* These are the sequential modular approach, the simultaneous modular approach and the equation oriented approach. 7.2.1 Sequential Modular Approach This method is the most cammonly used technique in process simulation and is similar in principle to the traditional method of hand calculation of unit operations, The computer program that calculates the performance of ‘a particular process is called a unit module. For each process unit in the flow diagram, there is a unit module, Given the values of input stream ‘composition and flow rates, temperature, pressure, enthalpy and the equipment parameters, the unit module will calculate the properties ofits outlet streams, The output stream for a unit will become the input stream for a downstream unit and the calculation proceeds as before until the process is completed. A library of unit modules forms an important part of the simulation software. To solve the model equations and interconnection equations. ‘numerical methods should be used. A library of numerical routines should 502 (CHEMICAL PROCESS COMPUTATIONS bbe provided. These would normally consist of solution procedures for a system of linear, non-linear and differential equations. To perform unit computations and solve the continuity equations for mass and energy, physical and thermodynamic property data is available through data banks and estimation routines. Many thermodynamic routines are given in Chapter 2. The heart of the flowshecting package is the executive program that controls the computation sequence. Besides the above routines, there are input and output routines and possibly programs for optimization and economic analysis. The basic components of a simulator are shown in Fig. 7.1 ‘The first step in any simulation is to generate a conceptual flow diagram of the process, showing the models and interconnections. In the conceptual flow diagram, two streams combining is normally represented as a mixer, and stream splitting is represented as a divider. The flow diagram of @ process for isopropyl acetate is shown in Fig. 7.2. The process has four units and eight process streams. Properties of feeds $, and ; must be Physical Property Data_Bak Thermo- yanarric Package Solution Optimi~ saten Econeric Analysis Fig, 7.1. Components of a simulation program. CHEMICAL PROCESS SIMULATION 503 E : oe & i “a & Fig, 7.2, Conceptual low diagram for the isopropyl acetate process. specified as input, All other streams are evaluated by simulation. The process has one recycle stream. The conversion level in the reactor should be specified for the design problem and the reactor size is computed. For performance calculations, the conversion is calculated from the reactor unit module. When recycle streams are present, then iterative calculations are necessary as the calculated value for the recycle may differ from the initial estimation, for all units in the recycle loop. Iterations continue until the stream properties converge to a stable value, within a preset tolerance. Referring to Fig. 7.2, the calculations will proceed as follows. (1). Given streams 8, and S,, guess stream S;. @) Since ll input streams are known, sequentially caleulate the output streams $5, 8,, Ss, Sg. S;- obtaining a new estimate of stream S,, @) If the two stream values for $, are within 2 specified small tolerance, the calculations have converged. Otherwise the executive program estimates another value of stream S, and repeats steps (2) and (3) Various numerical techniques can be used to obtain successive estimates of stream S,. These are discussed in Section 7.4.3, There are several advantages in the sequential modular approach. The flowshect architecture is most easily understood as it closely follows the process flow diagram. Existing individual unit modules may be readily incorporated. Building blocks may be added to or removed from the flowsheet at ease. Besides, unit modules can be used at wo different levels:a simple unit subprogram that does not entail detailed calculations, but performs simple mass and/or energy balances in the early design stage and ‘a more sophisticated unit module either in the final design stage or in the simulation of an operating plant. 504 (CHEMICAL PROCESS COMPUTATIONS ‘The sequential modular approach is not without its drawbacks. In a ‘complex system consisting of a number of recycle streams, establishing the hierarchy of calculation order or precedence ordering and iterative solutions can be ineficient, unless some form of optimal sequencingis built into the executive program. Many widely used modern simulators like DESIGN 2000 (Chemshare, Texas, USA), PROCESS (Simulation Sciences, California, USA) and CONCEPT Ill (CAD Centre, Cambridge, UK) use the sequential modular approach incorporating optimal sequencing methods and efficient convergence promotion methods. “The major disadvantage of the sequential modular approach becomes apparent when itis used for design rather than performance calculations. In design, there are constraints imposed on the stream values and, therefore, model parameters have to be adjusted to meet the design specification. This would normally involv iterated simulations and would become computationally inefficient for large processes, wit the number of iterations approaching 1000 7.2.2 Simultaneous Modular Approach The simultaneous modular approach is similar in principle to sequential modular. The steps involve the following. (1) For the first iteration, make inital estimates of recycle stream values, (2) Solve the problem using the sequential modular approach. G)_ Having obtained the frst estimate of input and output values for each unit, construct a linear relationship between them, ic. linearize the model equations (4) Since the interconnection equations are already linear, solve the whole system of mode! and interconnection equations simul- taneously using matrix methods to obtain a new set of inlet stream values. (5) _Iftwo successive iterates of assumed stream values converge within ‘a preset tolerance, the simulation is complete. Otherwise, go back to step (2). In the simultaneous modular approach, the calculations alternate between sequential modular and solution of linear equations and, thus, the computation time is reduced. Perhaps the greatest advantage of simultaneous modular approach is that it is easier to use for design problems. For example, the feed streams can be left unset and design (CHEMICAL PROCESS SIMULATION 50s constraints on specific intermediate or outlet streams can be imposed during the generation of linearized model equations. Some packages like FLOWPACK Il (ICI) use the simultaneous modular approach, even though, to the user, it would appear sequential modular.” The simultaneous modular approach is described in detail by Umeda and Nishino,!° Westerberg et al.’ and Biegler."” 7.2.3 Equation Oriented Approach In process flowsheeting where design and optimization have to be performed in the same program, because of linear and non-linear constraints imposed by the optimization problem, the modular approach becomes too cumbersome 10 use and iterated simulations may lead to prohibitively long computation times. To overcome this problem, the ‘equation oriented or simultaneous algebraic approach was developed. The basis is that the entire process can be represented by a collection of non- linear algebraic and differential equations, which are solved simultaneously for all the unknowns. The equations consist of model equations, interconnection equations, design specifications and physical property and thermodynamic correlations. In the case of even a medium-sized industrial plant, the system may contain several thousand equations. By suitable Gecomposition techniques, the equations can be solved in modules. The State-of-the-art in equation oriented approach to simulation has been reviewed by Shacham et al.'? and Perkins.'* The major problem is the solution of very large systems of non-linear equations. Many algorithms are described by Sargent.'* ‘The feasibility of using equation oriented simulators for stall steady- state and dynamic problems has been demonstrated successfully, but their implementation to large-scale problems is still currently under develop- ment, Some equation oriented simulators at an advanced stage of development are SPEEDUP (Imperial College, London), ASCEND (Carnegie-Mellon University, USA), FLOWSIM (University of Connecticut, USA) and QUASILIN (CAD Centre). 7.3. PARTITIONING AND TEARING ‘The flowsheet of a large process consists of many process units, control variables and several recycle streams. For ease of computation, the process 506 [CHEMICAL PROCESS COMPUTATIONS flowsheet is decomposed into smaller subsystems. The system decom- position is known as partitioning and consists of two parts. (1) Partitioning; where groups of units that must be solved together are identified in the flowsheet, There are no recycle streams between units in the groups. (2) Precedence ordering; where the sequence of computations of the partitioned subsystems, and sequence within a given partition, are determined. In calculating the process streams sequentially in accordance with the precedence ordering list, all the units constituting a recycle loop have to be Solved first before the sequence can pass on to the next member in the ls. ‘AU least one stream in the loop must be guessed before sequential calculations within the group can proceed. The method of choosing the stream asthe guessed stream s called “Yearing’ and the stream is called the “tear stream Some basic concepts of partitioning and tearing are described below. 7.3.1 Decomposition of Large Systems For partitioning a large system, the conceptual flow diagram of the process may be looked upon as a process flowsheet graph and the principles of graph theory are applied. Referring to Fig. 7.2, all units in boxes are nodes cof the process graph and all the streams are the directed edges of the graph. Since the edges are directed, the graph is called a digraph or directed graph, ‘When tracing a path from one node to another along the directed edges, if any node is encountered only once, then the path is called a simple path. If recycle streams are present, the path originating from a node will erminate atthe same node, Such a path called cyclic and all nodes in the cyclic path will constitute a cyclical net. Partitioning consists of dividing the digraph of the process into the smallest possible blocks or subgraphs. Treating the block or subgraph as a single pseudo-node, the overall digraph, consisting of nodes, pseudo-nodes and the directed edges between them, is made acyclic. Sequential computation would now apply to the acyclic graph, noting that all the nodes and edges within a block have been solved together before proceeding to the next node or block. ‘The underlying structure of a digraph can be readily understood by the use ofa process matrix, a stream S, terminates ona node, an entry of +5, is made in the process matrix against that node; ifthe stream originates from that node, then an entry of ~S, is made. As an illustration, consider (CHEMICAL. PROCESS SIMULATION 507 Water | Reale Fy comprtisor Purge Prrve Fic. 7.3. Process for aniline manufactur. the simplified flowsheet in Fig. 7.3 of @ process for manufacture of fniline? The digraph of the process is shown in Fig. 74. There are 10 nodes and 18 streams in the digraph. The process matrixis shown in Table iA Path Tracing Method Perhaps the simplest of available algorithms for partitioning, and precedence ordering ofa digraph is that of Sargent and Westerberg,’ * with its modification by Christiansen and Rudd." This method makes use of the principles of list processing . The tracing starts from an arbitrary node and fr fhe | Fic. 7.4, Process flowgraph of aniline process. 508 (CHEMICAL PROCESS COMPUTATIONS TABLE 7.1 PROCESS MATRIX FOR DIGRAPH SHOWN IN FIG. 7.4 Node Associated stream numbers 1 30 2 3-4 3 4-5 -6 4 6-7 5 7-8 -9 6 208 =10 7 Sei 12) ftw iG 9 B= 10 -18 15-16 traces the path of a stream originating from that node. Each node ‘encountered is placed in alist until a node already in the lst is encountered again. Then all nodes in the list encompassed by the twice encountered node are lumped into @ pseudo-node and the process matrix is updated. The lumped node is treated as a single node. In our example, the pseudo-node formed is [I-2-3-4-5-6]. This node replaces all the nodes in the loop. The individual nodesin the loop are deleted from Table 7.1 and only the streams into and out of the pseudo-node are retained, The exercise of tracing and forming pseudo-nodes is continued for all repeated nodes, until the entire digraph is made acyclic. The final precedence order list becomes [(1-2-3-4-5-6), 7, (8-9-10)}, There are two cyclical nets in the digraph. ‘An alternative to the path tracing method is to use the powers of the adjacency matrix method. The adjacency matrix represents the adjacency of nodes in a digraph, Details ofthis method, which is computationally ess efficient, but easier to program, are discussed by Himmelblaw and Bischoff"* and Kehat and Shacham.’ 7.3.2 Tearing In calculating the process streams sequentially in accordance with the precedence ordering list, all the nodes in a group constituting the recycle loop have to be solved first before the sequence can pass on to the next member in the list, Atleast one stream in the cycle must be guessed or ‘torn’ and the loop made acyclic, before sequential calculations within the loop are possible. The guessed stream values are re-estimated and the calculations proceed iteratively until two successive iterates of the guessed stream converge within a preset tolerance. (CHEMICAL PROCESS SIMULATION 509 Considerable attention has been devoted to tearing of recycle loops in the literature, Various criteria have been suggested as the most suitable to minimize computational efforts. These criteria are: (1) Minimum number of iteration streams. 2). Minimum number of stream variables or iteration stream parameters. (3) Weighting of streams according to user's knowledge of streams. (4) Minimum number of total loops torn. Many tearing algorithms require the formation of a loop/stream matrix and hence information on all individual loops in a digraph. The Joop/stream matrix is one whose rows correspond to individual loops in the digraph and columns correspond to streams in the loop. Ifastream existsin ‘loop, then the matrix element takes on a value of 1. Otherwise, a value of zero is assigned to the element. Weinblatt!” has proposed a path searching method to identify all the individual loops in a directed graph. "An efficient algorithm for criterion 1 was proposed by Barkley and Motard.2° If direct substitution is used for iterative calculations, then criterion (4) is the optimal tearing method. Details of a tearing algorithm according to criterion (4) are described by Upadhye and Grens*" and Westerberg eta.” The latter have shown that the sequence of caleulations when using simple direct substitution is different from the sequence when acceleration methods are used for convergence. Tncriterion (3) a weight is assigned to each stream. The weight can reflect the user’s knowledge about the stream or can incorporate the degree of culty associated with converging that stream. ‘The four tearing criteria mentioned above were combined into a single expression by Pho and Lapidus.?? If w,isthe integer weight associated with stream j and x, is a selection criterion of whether a stream is selected for tearing or not (x,=1 if stream is torn, 0 otherwise), then the optimal tearing criterion is Mine J with the constraints Yeosen 2,...9m streams ay (72) orl 3) 510 CHEMICAL PROCESS COMPUTATIONS TABLE 72 S OF STREAMS FOR OPTIMAL TEARING Criverion Weight Tearing algorithh unity Barkley and Motard?” (2) number of unknowns in the stream Pho and Lapidus™™ @) user defined Pho and Lapidus? (4) number of loops torn by each stream —_Upadhye and Grens?* Where a, are the elements of the loop/stream matrix. Equations (7.1)-(7.3) form a linear integer programming problem for which analytical solutions are feasible only for small size loop/stream matrices. The weights take on the values as shown in Table 7.2. ‘The tearing algorithms proposed by Lee and Rudd”? and Upadhye and Grens?!* require the loop/stream matrix to be determined fist. ‘An entirely different formulation of the tearing problem was proposed by Barkley and Motard?” and Pho and Lapidus.*? The process diagraph is now represented as a signal flow diagram where the streams become the vertices and the dependency between units becomes the edges. These methods do not require loop/stream matrix formulation and therefore computations are greatly simplified. Barkley and Motard*° describe an interval reduction procedure which implicitly imposes a weight of unity (criterion (1)) on the streams. The Basic Tearing Algorithm (BTA) of Pho Fic. 7.5. Signal fowgraph for aniline process. (CHEMICAL PROCESS SIMULATION sil and Lapidus” is more general in that it encompasses all the four criteria in Table 7.2. The Pho-Lapidus method is described here at some length. Signal Flow Graph ‘The signal flow graph is the dual of the traditional flowsheet graph. In the signal flow graph, which is also a digraph, the streams become the nodes and the edges indicate dependency. A directed edge from node (stream) ito node (stream) j indicates that stream jis a physical output from a unit to which stream ‘is the physical input, The signal flow graph representation is very efficient in systems decomposition and to identify the tear streams for specified weights on the streams. Since itis intended to follow through the example of the aniline process (Fig. 7.3) and identify the tear streams, the signal flow graph for the process flow digraph in Fig. 7.4 is shown in Fig, 7.5, A stream/precursor matrix can be written down, similar to the process matrix in Table 7.1. This matrix is shown in Table 7.3 ‘Table 7.3 can also be represented by a stream adjacency matrix. IFthereis a directed edge between stream S, and stream S, inthe signal flow graph, then value of unity is placed in the appropriate element in the stream adjacency matrix. If there is no directed edge between S, and S,, the element is left blank, Ihe stream adjacency matrix is given in Table 7.4. The adjacency matrix can be compacted to exclude all interface streams (streams with either rows or columns with all blanks). The interface streams are 1, 2,9, 12, 17 and 18. The compacted adjacency matrix is shown below. 345678 10 1 13 15 16 My ] Basic Tearing Algorithm ‘The basic tearing algorithm (BTA) was developed by Pho and Lapidu: ‘The algorithm scans the reduced adjacency matrix of the signal low graph | 512 B 4 is 16 7 is (CHEMICAL PROCESS COMPUTATIONS TABLE 73 STREAM/PRECURSOR MATRIX FOR SIGNAL FLOW GRAPH (FIG. 7.) Stream Precursor Stream —-Precursor i = 10 28 2 = nL 5 3 1 10 2 5 4 3 3 M16 5 4 “4 16 6 4 15 B 7 6 16 5 8 7 " 5 9 1 18 6 TABLE 74 STREAM ADIACENCY MATAHG FOR stGHeat FLOW oRAPH (ri. 7.5) ileseaes wey veers 10 1 213 14 15 16 17 18 (CHEMICAL PROCESS SIMULATION 513 and identifies the streams that are ineligible for tearing and tags all tear streams, The steps are as follows: {() Input the weights of all streams and the stream adjacency matrix. ‘The stream/precursor matrix can be input instead of the adjacency matrix and the latter can be formed by the computer program. (2) Compact the adjacency matrix by removing the rows and columns of interface streams. (@) Pick a stream s, from the adjacency matrix. Ifthe matrix is empty, then go to step (9). (4) Check whether the node s, is ineligible. This is done by the following method ‘Sum the weights of all predecessors of sand those of all successors of ‘The immediate successors are given by the non-zero elements in row m (set T) and the immediate predecessors are given by the non-zero elements in column m (set F,) of the adjacency matrix a) W660)" 3) Itw,,2 Myla) oF W,, 2 Ws, then that node stream is ineligible (5) Ifsyisnotineligible, then tag s,, cross out row mand column min the adjacency matrix and go back to step (3). If s, is ineligible, then: (2) _Deletethe row and column corresponding to, from the adjaceney matrix (row m and column m. (0) Isis dominated by its immediate successors, i. #, < W(Sq)> the for each stream s, in T, modify the elements in column j by forminga Boolean sum with the corresponding elements in column (0) Ifsgis dominated by its immediate predecessors, ie. w,,< Wel5a)> then for each stream s, in T,, modify the elements in row j by forming a Boolean sum with the corresponding elements in row m. (@) Goto sep (6) I no ineligible streams can be found, then go to step (8) 514 (CHEMICAL PROCESS COMPUTATIONS (6) Examine the nodes for self-loop. A non-zero element in the diagonal of the adjacency matrix indicates a sef-loop. Ifa self-loop is present, then declare the node as a torn stream and remove the corresponding row and column from the adjacency matrix. (1) Remove all rows and columas that have all zero entries and goto step @ (8) The signal flow graph is not reducible by the BTA. Print out an appropriate error message, the most recent status of adjacency matrix and stop. (9) Print out all the tagged (essential) stream nodes. These are the tear streams for the given weights Most moderately sized signal flow graphs can be reduced by the above algorithm. If none of the stream nodes can be identified as ineligible, then the algorithm fails. However, by the time this stage is reached, the graph is reduced to such a level that loops formed by two-way edges can appear One of the streams of the two-way edges is selected for tearing (usually the stream with a lower weight) and after the adjacency matrix is reduced further, the BTA can be applied to the rest of the computations. Thus, the BTA and two-way edge reduction can together reduce most signal flow zraphs. It should be noted that the optimal torn set obtained by Pho- Lapidus method is not unique. A different optimal set can be obtained by following other algorithms. ‘As an exercise, we can now apply the BTA to the reduced stream adjacency matrix of the signal flow graph shown in Fig. 7.5. To start with, assume that all weights are equal to unity. Stream S, is ineligible because the sum of the weights of its predecessors or successors is equal toils own weight. Removing the row and column for s, and modifying column s,, we find that stream s, is not ineligible. Tagging s, and removing row and column s,, e are left with a number of rows and columns with only zeros. ‘These are also deleted (55, 54453, 5 0,54). The adjacency matrix reduces toa 3x 3 matrix with streams s,,, 5,5 and & 13 15 16 weights 130 fo 1 0 1 is jo o 1 i we Li o 0 1 Following our argument, we have two choices. If s,5 is taken to be dominated by its immediate successors, then we are left with two-way edge (514, 5,4) from which either can be torn, both being of the same weight. (CHEMICAL PROCESS SIMULATION 515 However, ifs, is assumed to be dominated by its predecessors, then we have s,, as non-ineligible and, thus, the final tear streams set is {545 515] Let us consider a situation where the weights are the number of components in the streams, based on the process in Fig. 7.3. The vector of weights is listed in Table 7.5. Let us apply the BTA to the adjacency matrix in Table 7.4, after removing the interface streams and using the weights in Table 7.5. Since Wy Wo» its immediate predecessor, and therefore 5 is ineligible. ‘The matrix reduces to 15 16 weight 1s ° H] 3 w lo 1 2 Removing s, 5 because ofall zero column, we are left with s,, as the last tear stream. Therefore the tear set is now [55,6]: tis interesting to compare the two sets of tear streams obtained with different vectors of weights. In the first case with all unit weights, we have to tear s, and 51, which are the product streams from the reactor and the 516 CHEMICAL PROCESS COMPUTATIONS crude still condenser, respectively. In the second case, we have totear s, and yoy Which are the feed stream to the reactor and a recycle stream to the crude still, respectively. In the design problem, where the products are to meet design specifications, it is better to tear product streams and impose the design constraints as initial guesses. In the simulation problem, where the performance of an operating plant is evaluated, itis preferable to tear the feed streams and calculate the product streams. If we decide beforehand, from a knowledge of the processes, that a certain stream cannot be torn, then it can be made an ineligible stream by ascribing a higher weight. Alternatively, if a stream is required to be torn because of ‘our knowledge of the process, then it can be declared a torn stream and removed from the adjacency matrix. ‘The basic tearing algorithm, together with the two-way edge reduction, is simple and easy to implement, fast to execute and does not require a knowledge of the loop/stream matrix. In addition, it has the flexibility of ascribing different weights to the streams. 7.4 THE FLOWSHEET SIMULATOR {Use a simulation package for paces design or process simulation has ecome a commonplace activity with the process engineer. The progr: are used in the following situations. Hy _—— (1) In the initial stages of design, the program is used to calculate material and energy balances for alternative process flowsheets. At this stage, simple block diagrams of the process flowsheet are adequate (2) Inthe final stages of design, the simulator produces specifications for all process units and streams in the flowsheet. A detailed flowsheet of the process is required at this stage. (3) The simulator applied to an operating plant helps to evaluate the performance of the plant and to compare against design specifications in order to identify areas for improvement (4) Changes in operating conditions can be simulated and bottlenecks in the plant can be identified for improving process efficiency. The benefits of using a simulator at the design stage are capital saving through improved designs, engineering manpower savings via improved efficiency and productivity, a significant reduction in process engineering time on routine calculations, consistency of data and techniques used CHEMICAL PROCESS SIMULATION 317 throughout the design, and improved and quicker communication among, process engineers at all company locations. ‘The number of steady-state process simulation packages have grown so rapidly in the lst ten years that itis beyond the scope of this chapter to describe them in detail. Kehat and Shacham,? Flower and Whitehead,” Motard et al,° Hlavacek* and Rosen® have reviewed the state-of-the-art feom time to ime. A number of available simulation programs have been compiled by Peterson et al.®* and Chen and Evans." 7.4.1 Structure of a Simulator ‘The basic function of all simulators is the calculation of mass and energy balances for the entire process, including recycle streams. From the early stage of having specific programs for simulating one process orone plant, the art of steady-state process flowsheeting has developed into an advanced form, where versatile programs can simulate a large number of processes and plants. Some of the details specific to a process that are not available in 1 general package may be supplied by the user, because of the flexible structure of many simulators, ‘The essential components of a fowsheeting package are shown in Fig, 7.1. The package can be a rigid structure or a flexible structure. Programs swith a rigid structure are simple and straightforward. All that the user has to supply is the data and the executive handles the program in the same way, irrespective of the nature of the process simulated. Unfortunately, all unit modules have to be loaded into the computer main memory whether or rot many of the routines are relevant for the process being simulated. As these programs grow to include more unit modules, physical and thermodynamic property estimations and wider output options, the complexity increases without the offsetting feature of flexibility Fortunately, most modern simulators use a flexible or variable structure Ina flexible structure simulator, a problem oriented language (POL) is used. The process topology can be explained by sentences. For instance, if a stream is split into two streams in the process without physical or chemical transformation, then the sentence for such an operation may read ‘MODEL UNIT X BY DIVIDER’. The tag numbers of input and output streams for the unit will also be supplied to complete the topology: DIVIDER would be a unit module which performs the stream splitting function during the calculation stage. The POL normally generates a procedure oriented language (often FORTRAN) which is compiled, linked ‘and executed, Only those modules that are actually needed are loaded. Data storage space may be allocated only as required by the size of the 518 (CHEMICAL PROCESS COMPUTATIONS problem. User supplied routines can be interfaced with the package. The variable structure simulator thus provides greater flexibility and modu- larity. An alternative to the POL is the user-written main program linking all unit modules through a series of CALL statements. From Fig. 7.1, it is seen that the structure of a simulator essentially consists of five phases. () Input 2) Preprocessing G) Calculation (4) Output (5) Postprocessing ‘The input should consist of process topology, information on all input streams, design parameters ofthe units and convergence criteria, Process topology is defined through POL. Feed stream information consists of flow rate, number of components and composition, temperature and pressure. The components can be identified either through @ code number or by name, in a form that the physical property data bank and thermodynamic package can recognize and supply all necessary information for the calculations. The design parameters of the units may be heat exchanger area, height of packing in an absorber or numberof theoretical tages in a Meio opees "ou papa Kies sn ‘owen NVMLMOT uso) poumpy YSIS one NOISE onvuruxs inno pet fupssoon oneieond wey sqapousanboe vsa

    You might also like