A complex process simulation computing system and its construction method for process industries

Abstract

This invention relates to the field of process industry computing, specifically to a complex process simulation computing system and its construction method for process industries. Addressing the problems encountered in the mathematical simulation of complex process industries, the process simulation computing system is divided into two main layers: internal iterative computation and external iterative computation. Internal iterative computation focuses on the numerical solution of industrial unit models, while external iterative computation emphasizes the modeling and analysis of the flow diagram to reveal the topological structure of the process flow. In the overall method, external iteration continuously calls internal iteration, decoupling model parameters and process structure. Internal iterative computation is nested within external iterative computation, achieving a holistic solution of the entire external process and each internal model. This solves the problems of unclear description of complex processes, slow iteration convergence speed, and low algorithm efficiency in external iterative computation.

Description

Technical Field

[0001] This invention relates to the field of process industry computing, and more specifically to a complex process simulation computing system and its construction method for process industries. Background Technology

[0002] Process simulation technology describes process systems in the process industry using mathematical methods. By combining it with computer technology, it enables the measurement and calculation of material flow, energy flow, and other parameters within the process system. A process simulation system consists of various industrial equipment units arranged in a specific topological relationship. These units can all be mathematically described by a system of nonlinear equations. The entire process system is a very complex and large-scale system of nonlinear equations obtained by combining the mathematical models of each unit according to certain relationships. Solving this system of equations is the process of realizing the mathematical simulation of the process system.

[0003] The mathematical model is the core component of a process simulation system, primarily consisting of two parts: the industrial unit model and the structural model. The industrial unit model describes the specific industrial units within the system, while the structural model describes the complex network relationships between these units. The process simulation platform comprises a basic physical property database, a thermodynamic method library, a unit mechanism model library, and a computational system. The computational system, after obtaining the flow chart from the platform's drawing interface, first establishes the mathematical model for the corresponding unit, then retrieves the physical property data and thermodynamic methods related to each module from the underlying database, identifies the process topology, determines the calculation order of each module and the values ​​of each material, and finally calls appropriate computational methods from the computational method library to solve the overall mathematical model of the process. A mature process mathematical model often involves a large amount of computation and is difficult to solve. Furthermore, the equipment, materials, and structures of the process flow are often complex and variable, frequently requiring adjustments to the process during system design and optimization. Furthermore, the selection and replacement of equipment, adjustment of logistics data, or changes in process structure make it more difficult to calculate and solve the overall mathematical model, and the accuracy of the calculation results will directly affect the reliability of the overall process simulation.

[0004] Existing process simulation systems often employ the sequential modular approach (such as Aspen Plus and PRO / II), which compiles the mathematical model of each process unit into a solver program and then performs simulations sequentially from front to back according to the connection order of modules in the process system, iterating and converging when encountering cyclic material or energy flows. In contrast, process simulation platforms like UniSim and gPROMS utilize a simultaneous equations approach. The basic idea of ​​the simultaneous equations approach is to connect all equations describing the process system (including model equations within units) into a single, mathematically vast nonlinear equation system. This system includes all internal and external variables: physical property variables, unit model variables (internal variables), flow variables describing the system structure, and process parameters (external variables). Once this nonlinear equation system containing all internal and external variables is established, the steady-state simulation can be considered a purely mathematical problem that is, in principle, independent of the process system structure and even the chemical engineering background. Sequential modular approach systems offer advantages such as aligning with engineer intuition, ease of generalization, and ease of error diagnosis and debugging; however, their nested iterative computational efficiency is relatively low. The simultaneous equation method is detached from the process system structure and technological background, simply simplifying the problem into a mathematical solution. Therefore, this method does not conform to the production environment and the intuition of engineers. When the process flow fails or the simulation is incorrect, it is difficult to diagnose the error. When performing iterative calculations, due to the detachment from actual production, it is more difficult to estimate the initial values ​​of variables, which increases the difficulty of computational convergence. Furthermore, when the process changes, the simultaneous equation method is difficult to inherit existing modules and cannot simplify the calculation process and improve computational efficiency by migrating the original method.

[0005] Chen Huidong proposed a computational framework for a process flow simulation platform based on the sequential modular method and graph theory. By developing configuration tools, he transformed directed graphs into adjacency matrices and employed graph theory methods to perform topological decomposition and processing of the process system. This enabled the implementation of a method for segmenting, cutting off, and sequencing the chemical system, achieving a complete solution for the entire chemical system process. After calculating and sequencing the process flow and cutting off the backflow streams, he used a hybrid computational method combining direct iteration and the Wegstein method for calculating the convergence of the cut-off streams. The Wegstein method, similar to the secant method, often exhibits oscillatory convergence when solving multivariable nonlinear equations, limiting its application to weak crosslinks and thus its narrow scope. Furthermore, this method suffers from slow iterative convergence and low algorithm efficiency.

[0006] As can be seen from the above, structural complexity is a significant factor contributing to the complexity of mathematical models in process industries. In system process design or optimization, adjustments to the process are frequently necessary, i.e., changes to the system's topology. Even small structural changes can lead to the re-segmentation of the entire system, significantly reducing computational efficiency. Therefore, constructing an accurate, feasible, and universally applicable computational system, and improving algorithm efficiency, are paramount in process simulation technology and are of great importance for process development, design, and optimization. Summary of the Invention

[0007] To address the aforementioned issues, this application aims to develop a general and effective method for simulating complex process flows in industrial applications, based on theories of mathematical optimization and system topology decomposition. This method divides the process simulation system into two main layers: internal iterative computation and external iterative computation. Internal iterative computation focuses on the numerical solution of industrial unit models, while external iterative computation emphasizes the modeling and analysis of the flow diagram to reveal the topological structure of the process. In the overall method, internal iteration is continuously invoked through external iteration, decoupling model parameters and process structure. Internal iterative computation is nested within external iterative computation, achieving a holistic solution of the entire external process and each internal model. This solves the problems of unclear description of complex processes, slow iteration convergence, and low algorithm efficiency in external iterative computation. Through this strategy, we can accurately solve the mathematical models of each unit and effectively address the problem of the overall computational sequence in complex process simulation, as well as the problem of determining the values ​​of material flows between upstream and downstream equipment. Therefore, this method provides a computational system suitable for complex process flow structures, providing a computational and algorithmic foundation for developing interactive simulation technologies for process industries.

[0008] To achieve the above objectives, this application provides a method for constructing a complex process simulation and calculation system for process industries, comprising two computational layers: internal iteration and external iteration. Internal iteration is used to calculate the basic model, while external iteration solves the entire process flow by calling the parameters of the internal iteration model. Specifically, the method includes the following steps:

[0009] (1) Identify the target topology diagram and generate an adjacency matrix: Based on the target topology diagram, the user automatically numbers the units and directed lines according to the module creation order when drawing, obtains the initial unit and directed line calculation sort, and sets connection nodes at the midpoint of the edge of each unit and the endpoint of the directed line, associating the basic modules in the internal iterative calculation, and establishing a two-dimensional structure data table of different units and directed lines n*3 accordingly. Then, the unit and directed line structure data tables are identified and an adjacency matrix is ​​generated.

[0010] (2) Loop identification and sorting algorithm: Loop identification and sorting algorithm: Input the adjacency matrix according to the order of the initial unit calculation and sorting, process and update the adjacency matrix row by row through loop identification, and then process the updated adjacency matrix in reverse to obtain the processed unit calculation and sorting and the processed adjacency matrix. According to the processed unit calculation and sorting, traverse the directed line structure data table to obtain the set of reverse flow streams of the directed line. Then, according to the processed adjacency matrix, use the upstream and downstream backflow identification method to determine the final calculation and sorting and the set of flow streams to be cut.

[0011] (3) Solution of the entire flow stream: Create a virtual flow stream to replace the set of flow streams that need to be cut and perform the cutting operation. Then, input the initial variable data and the initial value of the virtual flow stream variable in the unit structure data table to obtain the initial flow stream data table. In sequence, call the basic model of each unit with other variable values ​​or internal iteration calculations to perform the full process calculation and solution. After the initial solution, use the quasi-Newton method to iteratively converge the cut flow streams. After completing the iterative calculation of all cut flow streams, output the final flow stream variable data table to complete the full process solution.

[0012] Further, the loop identification process in step (2) involves forward row-by-row processing and updating the adjacency matrix as follows: The cell corresponding to the row with the nearest non-zero element starting from the i-th element in the i-th column is denoted as K, and the cell corresponding to the column with the farthest non-zero element starting from the i-th element in the i-th row is denoted as P. The adjacency matrix is ​​checked row-by-row according to the initial cell calculation sorting order to determine if there is a backflow. If the adjacency matrix has downstream cells sorted before cell I, that is, the sum of all elements n from the i-th element in the i-th row to the left of the adjacency matrix is ​​not 0, then there is a backflow. At this time, the sum of all elements m from the i-th element in the i-th column upwards is determined. If m < n, then the cell I corresponding to the cell calculation sorting and the adjacency matrix is ​​inserted before cell P; otherwise, it is inserted after cell K. All adjacency matrices are traversed to obtain the forward-processed adjacency matrix.

[0013] Furthermore, the method for constructing the set of reverse flow segments of directed lines in step (2) is as follows: traverse the data table of directed line structures. If the direction from the starting unit to the final unit of a certain directed line is in reverse order in the unit numbering, then it is a reverse flow segment of the directed line. All reverse flow segments of directed lines form a set of reverse flow segments of directed lines.

[0014] Furthermore, in step (2), the upstream and downstream backflow identification method is used to confirm the final calculation sort and the set of streams that need to be cut. Specifically, each directed line reverse-order stream is found in the final adjacency matrix and processed one by one. Assuming there are n streams, a set of streams that can be cut is created with a length of n. The first stream is E→D. Then, the cyclic flow containing the reverse-order stream is found through the upstream and downstream backflow identification method. If there is only one L1 in the cyclic flow containing the reverse-order stream, all streams in the cyclic flow are added to the set of streams that can be cut (there are n sets in total). If there are multiple L1, L2, etc. in the cyclic flow containing the reverse-order stream, it is determined whether the streams that these cyclic flows share are unique except for E→D. If they are not unique, these shared streams are added to the set of streams that can be cut. Then, the judgment of the next stream is entered. After all streams are judged, the set of streams that can be cut is output.

[0015] Furthermore, the full-process calculation and solution described in step (3) is as follows: when calculating each unit sequentially, the mathematical model of the corresponding unit is called in turn, the flow streams flowing into the unit are determined through the structure data table, and the variable values ​​required for solving the model are obtained from the flow stream variable data table. After the solution is completed, the variable values ​​obtained from the solution are filled into the flow stream variable data table, the flow stream variable data table is updated and used for subsequent calculations.

[0016] Furthermore, the iterative convergence of the cut-off flow stream using the quasi-Newton method in step (3) specifically involves: after completing the initial full-process calculation, ensuring that the values ​​of the two virtual flow streams obtained at the cut-off point are equal, resulting in a set of nonlinear equations F(X) = X, which is then transformed into G(X) = F(X) – X = 0. The initial values ​​of the virtual flow streams are obtained from the flow stream variable data table and substituted into the quasi-Newton solution to finally obtain the value X of the returning flow stream. The cut-off flow stream is then restored in the flow stream variable data table, the corresponding virtual flow stream is deleted, and the value of the returning flow stream is entered to update the table.

[0017] Furthermore, in step (3), after iterative calculation of all cut-off streams, when there is an iterative loop inside the model and the calculation requires multi-level iterative interaction between the inside and outside, the whole-flow program solution method is integrated into the iterative solution method at the broken stream, and the residual is used as the objective function to realize the iterative convergence process at different levels.

[0018] And, a complex process simulation computing system for process industries, comprising:

[0019] Internal Iteration Module: This module relies on the amount of physical property data, including thermodynamic data and transport property data of the material, and is used for the calculation and solution of the basic model.

[0020] The outer iteration module is used to perform mathematical modeling and in-depth analysis on the input flowchart to achieve the above steps. It can call the calculation results of the inner iteration module as needed.

[0021] A computer device comprising:

[0022] Memory; and

[0023] The processor is configured to execute the aforementioned method for constructing complex process simulation calculations for the process industry, based on instructions stored in the memory.

[0024] A computer-readable storage medium storing computer instructions that, when executed by a processor, can implement the aforementioned method for constructing complex process simulation calculations for process industries.

[0025] The beneficial effects of this application are as follows: This invention divides the process simulation calculation system into two major layers: internal iterative calculation and external iterative calculation. Internal iterative calculation is mainly responsible for calculating and solving basic physical properties, thermodynamic methods, and unit mechanism models. It integrates more than 20 process industry unit mechanism models for calculating material flow and energy flow, which is sufficient to cover commonly used unit models in process simulation. External iterative calculation, based on topology recognition theory and adjacency matrix model, mainly performs mathematical modeling and in-depth analysis on the input process flow. It can quickly respond to complex variable topology structures and make accurate process cutting and calculation sorting judgments. This method overcomes the problems of unclear description of complex processes, slow iterative convergence speed, and low algorithm efficiency in existing process simulation calculation frameworks by decoupling the model parameters and process structure of the process simulation calculation system, achieving the calculation and solution effects of current mainstream process simulation software. This method provides engineers and researchers engaged in process industry simulation with a powerful automated calculation tool and provides a computational technology and algorithmic foundation for developing interactive simulation technology for the process industry. Attached Figure Description

[0026] Figure 1 This is a structural design diagram of a complex process simulation calculation system according to the present invention;

[0027] Figure 2 This is a structural diagram of the directed graph recognition and matrix reading algorithm of the external iteration calculation system of this invention;

[0028] Figure 3 This is a forward step diagram of the loop identification and sorting algorithm based on the adjacency matrix of this invention;

[0029] Figure 4 This is a reverse step diagram of the loop identification and sorting algorithm based on the adjacency matrix of this invention;

[0030] Figure 5This is a diagram illustrating the steps of the flow stream cutting process in the loop identification and sorting algorithm based on the adjacency matrix of this invention.

[0031] Figure 6 This is a technical schematic diagram of Scheme 1 of the whole-process flow strand solution method of the present invention;

[0032] Figure 7 This is a technical schematic diagram of Scheme 2 of the whole-process flow strand solution method of the present invention;

[0033] Figure 8 This invention uses complex Sargent and Westerberry directed graphs as the input case for the computational system.

[0034] Figure 9 This is a schematic diagram of the overall case of Experiment Example 2. Detailed Implementation

[0035] The present invention will now be described in further detail with reference to the accompanying drawings. The specific implementation steps described below will enable those skilled in the art to gain a more comprehensive understanding of the present invention, but do not limit the invention in any way.

[0036] First, it should be noted that the use of computer simulation to generate a three-dimensional rock mass fracture network model and three-dimensional block identification are existing technologies. The technical solution of this invention is based on the block data obtained from three-dimensional block identification and performs small block processing, which involves a large number of existing technologies, the definitions or concepts of which are already known or well-known in the field. Therefore, unless otherwise specified by this invention, other meanings consistent with existing known meanings will not be described in detail here.

[0037] Example 1: A method for constructing a complex process simulation computing system for process industries.

[0038] like Figure 1 As shown, the system includes internal iteration calculation and external iteration calculation. The internal iteration calculation is used to calculate the basic model, including conventional basic physical property databases in the chemical industry, thermodynamic method libraries, and unit mechanism model libraries. The external iteration realizes the solution of the entire flow stream by calling the parameters of the internal iteration model, which specifically includes the identification of the topology diagram and the generation of the adjacency matrix, the identification of loops and the calculation of the sorting algorithm, and the solution of the entire flow stream.

[0039] 1. Identify the target topology diagram and generate an adjacency matrix: Based on the target topology diagram, the user automatically numbers the units and directed lines according to the module creation order when drawing, obtains the initial unit and directed line calculation sort, and sets connection nodes at the midpoint of the edge of each unit and the endpoint of the directed line, associating the basic modules in the internal iterative calculation. Based on this, a two-dimensional structure data table of different units and directed lines n*3 is established, and then the adjacency matrix is ​​generated by identifying the unit and directed line structure data tables respectively.

[0040] The number is automatically assigned by the program when the interface is drawn.

[0041] Among them, the target topology diagram is a diagram formed in the actual operation of the process industry. It is a module created by the user on the drawing interface based on the target flow diagram, such as Sargent and Westerberry directed graphs.

[0042] It should be noted that the cell structure table and the directed line structure table are generated adjacency matrices (e.g., ...). Figure 1 The output matrix N and output matrix M shown are the same matrix, generated simultaneously for comparison to verify the correctness of the data; that is... Figure 1 In this case, N = M.

[0043] Its construction process is as follows Figure 2 As shown, it includes two parts: identifying a directed graph to obtain a structure data table and identifying the structure data table to obtain an adjacency matrix. Specifically:

[0044] Step (1) Construct cells and number directed lines and create connection nodes:

[0045] When creating unit modules, they are numbered sequentially using uppercase letters A through Z. Modules exceeding this range are numbered AA through AZ, and so on, with no upper limit. Directed line modules are also numbered sequentially, starting with the number 1, again with no upper limit. Hovering the mouse over any module will instantly display its corresponding number. Furthermore, each unit's edge midpoints and the endpoints of directed lines have connection nodes; the modules associated with these nodes are their parent classes.

[0046] Step (2) Obtain the data table of the unit and directed line structure:

[0047] 1) Cell Structure Data Table: Create a two-dimensional data structure table with different numbers of cells (n*3). The header fields are the cell number "name", the directed line that inputs the cell "linein", and the directed line that outputs the cell "lineout". First, fill the cell numbers into the "name" column sequentially. Traverse the connection nodes, find the directed line number that connects the arrow end to the cell and fill it into "linein". Find the directed line number that connects the starting end to the cell and fill it into "lineout". Output the table after completion.

[0048] 2) Directed Line Structure Data Table: Create a two-dimensional data structure table with different numbers of units (n*3). The header fields are the directed line number "name", the starting unit "start", and the ending unit "end". First, fill in the directed line numbers sequentially into the "name" column. Traverse the connection nodes, find the unit number connected to the starting end of the directed line and fill it into the "start" column. Find the unit number connected to the arrow end of the directed line and fill it into the "end" column. Output the table after completion.

[0049] Step (3) Identify the structured data table and generate the adjacency matrix:

[0050] 1) Identify the unit structure data table: First, determine the number of elements and names in name. If the number is n, create an n*n matrix with all zeros. Then, extract the numbers from each element in lineout and record the index x of the corresponding name element (x indicates which name element this element is). For each number, iterate through linein, find the same number, and record the index y of the corresponding name element (which has the same meaning as x). Then, fill the (x, y) position of the created matrix with the number 1. Continue until all elements in lineout are traversed, then end the loop and finally output the matrix.

[0051] 2) Identify directed linear data tables: First, determine the number m of distinct elements in start and end, and number them sequentially to create an m*m matrix filled with 0s. Then, extract the corresponding elements in start and end row by row, find the number x corresponding to the start element and the number y corresponding to the end element in the same row, and change the (x,y) position of the matrix to 1. If either start or end element in a row is empty, skip that row. After traversing all rows, end the loop and output the matrix.

[0052] 2. Loop Identification and Sorting Algorithm: The adjacency matrix is ​​input according to the initial unit calculation and sorting order. The adjacency matrix is ​​updated row by row through loop identification in the forward direction. Then, the updated adjacency matrix after forward processing is processed in reverse to obtain the final unit calculation and sorting order and the final adjacency matrix. The directed line structure data table is traversed according to the final unit calculation and sorting order to obtain the set of reverse-order flow segments of the directed lines. Finally, the upstream and downstream backflow identification method is used to determine the final calculation and sorting order and the set of flow segments that need to be cut according to the final adjacency matrix.

[0053] like Figure 3 The diagram shows the forward row-by-row processing and updating of the adjacency matrix: specifically:

[0054] Step (1) is to number the adjacency matrix:

[0055] The numbers, sorted by cell number, are labeled with each row from top to bottom and each column from left to right in the adjacency matrix. The label is the cell name corresponding to each row or column.

[0056] Step (2) Identify the adjacency matrix and determine whether there is backflow by identifying whether there are downstream units before sorting: If the critical matrix has downstream units sorted before unit I, that is, judge whether the sum of all elements to the left of the i-th element in the i-th row (i.e. the number of downstream units) is 0 according to the label order. If it is 0, skip the row. If it is not 0 (there is backflow), proceed to step (3) and traverse all rows to proceed to step (6).

[0057] Step (3) Compare the number of upstream and downstream units before each unit is sorted:

[0058] Sum all elements upwards from the i-th element in the i-th column of the adjacency matrix, denoted as m (number of upstream units). Compare n with m. If n is less than or equal to m, proceed to step (4). If n is greater than m, proceed to step (5).

[0059] Step (4) handles the case where the number of downstream units is not greater than the number of upstream units:

[0060] The labels added in step (1) record the row corresponding to the nearest non-zero element starting from the i-th element of the i-th column of the adjacency matrix, cell K, and cell I corresponding to the i-th row. Cell I is inserted after K in the 'Item' column of the cell numbering. Cells following it are automatically shifted. The 'ID' column of the cell numbering remains unchanged. The reordered cell numbering replaces the previous cell numbering. The initial labels of each row and column of the adjacency matrix do not change with their position. They are transformed in two steps: first, the row corresponding to cell I is inserted below the row corresponding to cell K, and the rows below them are automatically moved to the next row; second, the column corresponding to cell I is inserted to the right of the column corresponding to cell K, and the columns to the right of the column corresponding to cell I are automatically moved to the right. After the transformation of the adjacency matrix is ​​completed, the current adjacency matrix is ​​replaced with the transformed adjacency matrix. Step (2) is returned and the judgment continues from the (i+1)-th row.

[0061] Step (5) handles the case where the number of downstream units is greater than the number of upstream units:

[0062] The labels added in step (1) record the column corresponding to the first non-zero element from left to right in the i-th row of the adjacency matrix (the column with the furthest non-zero element starting from the i-th element in the i-th row), the cell P corresponding to the i-th row, and the cell I corresponding to the i-th row. In the 'Item' sorting of cell numbers, cell I is inserted before cell P. Cell P and the cells following it are automatically shifted, and 'ID' remains unchanged. The initial labels of each row and column of the adjacency matrix do not change with their position. They are transformed in two steps: first, the row corresponding to cell I is inserted above the row corresponding to cell P, and the rows below them are automatically moved down one row; then, the column corresponding to cell I is inserted to the left of the column corresponding to cell P, and the columns to their right are automatically moved one column to the right. After the adjacency matrix transformation is completed, the transformed adjacency matrix is ​​output, and the current adjacency matrix is ​​replaced. Step (2) is returned, and the judgment continues from the (i+1)-th row; finally, the adjacency matrix after positive processing is formed.

[0063] The reverse processing of the adjacency matrix updated after the forward processing is as follows: Figure 4 As shown, it is the same as the forward processing steps (1)-(5), except that the processing order is reversed. After the reverse processing, the final unit calculation sort and the final adjacency matrix are obtained. Based on the final unit calculation sort, the directed line structure data table is traversed to obtain the set of reverse-order flow segments of the directed line. Then, based on the final adjacency matrix, the upstream and downstream backflow identification method is used to determine the final calculation sort and the set of flow segments that need to be cut, as shown. Figure 5 As shown, specifically:

[0064] Step (1) Obtain the final cell calculation sort and identify the reverse flow:

[0065] Traverse all directed lines in the directed line structure data table. If the direction from the starting cell to the ending cell of a directed line is in reverse order in the cell numbering, then add the directed line to the reverse flow set.

[0066] Step (2) Find each reverse flow in the final adjacency matrix and process them one by one. Assuming there are n flows, create a cuttable flow set with a length of n. The first flow is E→D. Then, find the cyclic flow containing the reverse flow by identifying upstream and downstream backflow. If there is only one cyclic flow L1 containing the reverse flow, add all flows in the cyclic flow to the cuttable flow set 1 (there are n sets in total). If there are multiple cyclic flows L1, L2, etc. containing the reverse flow, determine whether the common flow of these cyclic flows is only E→D. If not, add the common flow to the cuttable flow set 1. Then proceed to the judgment of the next flow. After judging all flows, output the cuttable flow set.

[0067] Step (3) pops up a dialog box, where the user can select the specific stream to be cut from the provided set of streams that can be cut. Then the user returns the result to the program, which determines whether to change the stream to be cut. If the stream is changed, the corresponding calculation sort is changed, and the final calculation sort and the set of streams to be cut are output.

[0068] 3. Full-process flow stream solution: Create virtual flow streams to replace the set of flow streams to be cut for the cutting operation. Then, input the initial variable data and the initial values ​​of the virtual flow stream variables into the unit structure data table to obtain the initial flow stream data table. In sequence, call the basic model of each unit with other variable values ​​or internal iteration calculations to perform full-process calculation and solution. After the initial solution, use the quasi-Newton method to iteratively converge the cut flow streams. After completing the iterative calculation of all cut flow streams, output the final flow stream variable data table to complete the full-process solution.

[0069] like Figure 6 As shown, the method takes the set of streams to be cut, the structured data table, the stream variable data, the initial values ​​of the stream variables to be cut, the environment variables, the constraint variables, the mathematical models of each unit (externally called), and the overall calculation order as input, and outputs the values ​​of each stream variable. The steps are as follows:

[0070] Step (1) Obtain the set of streams to be cut obtained by the above method, and perform stream cutting operation:

[0071] Traverse the cell structure data table and the directed line structure data table, delete the streams contained in the set of streams to be cut, and replace them with the corresponding virtual streams. For example, if stream 1 is found in the set of streams to be cut in the cell structure data table, delete it and replace stream 1 in lineout with virtual stream 1a, and replace stream 1 in linein with virtual stream 1b. If stream 1 is found in the set of streams to be cut in the directed line structure data table, delete the corresponding row and add rows corresponding to virtual streams 1a and 1b. The start unit of 1a is the original start unit of stream 1, and the end unit is empty. The start unit of 1b is empty, and the end unit is the original end unit of stream 1.

[0072] Step (2) Obtain the flow variable data and initialize the flow variable data table, including the virtual flow:

[0073] The table has n rows and m columns, where n is the number of streams, each row contains all variable information for its corresponding stream, and m is the number of variables in the stream with the most variables. Users need to enter known variable data and, based on the manual or experience, set the initial values ​​for the virtual streams generated after cutting, and add them to the corresponding positions in the stream variable data table.

[0074] Step (3) Obtain the values ​​of other variables, such as the values ​​of environmental variables needed for calculation and the constraint variables that meet the design requirements.

[0075] Step (4) Obtain the processed full-process calculation sequence and the mathematical model of each unit, and perform full-process calculation and solution:

[0076] When performing calculations unit by unit in sequence, the mathematical model of the corresponding unit is called in turn. The flow streams flowing into the unit are determined through the structural data table, and the variable values ​​required to solve the model are obtained from the flow stream variable data table. After the solution is completed, the variable values ​​obtained are filled into the flow stream variable data table, the flow stream variable data table is updated, and it is used for subsequent calculations.

[0077] After completing the initial full-process calculation in step (5), the quasi-Newton method is used to iteratively converge the cut-off stream:

[0078] After completing the entire calculation process, the values ​​of the two virtual streams obtained at the point of stream cutoff are made equal to each other, resulting in a set of nonlinear equations F(X) = X. This is transformed into G(X) = F(X) – X = 0. The initial values ​​of the virtual streams are obtained from the stream variable data table and substituted into the quasi-Newton solution to finally obtain the value X of the return stream. The cut stream is then restored in the stream variable data table, the corresponding virtual stream is deleted, and the return stream value is entered to update the table.

[0079] It should be noted that when the model contains iterative loops and the calculation requires multi-level iterative interaction between internal and external processes, the full-flow procedural solution method is integrated into the iterative solution method at the fractured flow stream, with the residual used as the objective function, such as... Figure 7 As shown, iterative convergence processes at different levels are implemented.

[0080] Experimental Example 1

[0081] The complex process simulation calculation system constructed in this invention uses data in a fixed format as the basic input of the inner iterative calculation system and data tables in the form of cells and directed lines as the basic input of the outer iterative calculation system. It can quickly construct process simulation networks of different complexity by dragging and dropping, and return the correct output results. Figure 8 This invention uses complex Sargent and Westerberry directed graphs as input cases for its computational system and has achieved excellent verification results.

[0082] Experimental Example 2

[0083] The complex process simulation calculation system constructed in this invention is applied to a three-stage flash evaporation process in chemical engineering. This calculation system is based on an externally constructed mathematical model of the flash tank and mixer, along with related physical and thermodynamic databases. Using the structured process system data shown in Table 1 as input, a three-stage flash evaporation directed graph is constructed on the drawing interface using a drag-and-drop method. By identifying the topology of the directed graph, a structural data table is obtained, followed by the corresponding adjacency matrix. The final calculation order and cut-off streams are obtained by identifying the adjacency matrix. Based on the calculation order and cut-off streams, sequential and iterative calculations are performed according to the full-process calculation framework provided by the calculation system. Finally, the results of various variables for each stream in the entire process and the specific vaporization rate of each flash tank are obtained. The results are shown in Tables 2 and 3, verifying the effectiveness of the calculation system. A schematic diagram of the overall case is shown below. Figure 9 As shown, this case demonstrates that the complex process simulation calculation system constructed by the present invention has good application effects when facing the calculation of actual process systems, and is a feasible technical solution for solving complex process simulation calculation.

[0084] Table 1. Structured Process Data

[0085]

[0086] Table 2 shows the results of the specific vaporization rates for each flash evaporator.

[0087] Solution results Flash evaporator B2 Flash evaporator B3 Flash evaporator B4 Gasification rate 0.522 0.617 0.331

[0088] Table 3 Results of various variables for each stream throughout the entire process.

[0089] Flowing Variables Molar flow rate (mol / h) Butane mole fraction pentane mole fraction hexane mole fraction 1 453.6 0.3 0.4 0.3 2 706.451 0.283 0.426 0.291 3 368.502 0.398 0.426 0.176 4 337.949 0.157 0.425 0.417 5 141.085 0.231 0.474 0.295 6 111.766 0.279 0.470 0.250 7 227.417 0.502 0.397 0.101 8 226.183 0.097 0.403 0.500 9 252.851 0.252 0.472 0.275

[0090] Note: The actual scope of this invention includes not only the specific embodiments disclosed above, but also all equivalent solutions implemented or performed under the claims.

Claims

1. A method for constructing a complex process simulation computing system for process industries, characterized in that, The computation includes two levels: inner iteration and outer iteration. The inner iteration is used to calculate the basic model, while the outer iteration solves the entire flow path by calling the parameters of the inner iteration model. Specifically, it includes the following steps: (1) Identify the target topology graph and generate an adjacency matrix: The cells and directed lines of the target topology graph are automatically numbered according to the module creation order to obtain the initial cells and directed lines. The initial cells and directed lines are sorted and connected nodes containing connection information are set at the midpoint of the edge of each cell and the endpoint of the directed line. Based on this, a two-dimensional structure data table of different cells and directed lines n*3 is established. Then, the adjacency matrix is ​​generated by identifying the cell and directed line structure data tables respectively. (2) Loop identification and sorting algorithm: The adjacency matrix is ​​input according to the order of the initial unit calculation and sorting. The adjacency matrix is ​​updated by processing row by row in the forward direction through loop identification. Then, the adjacency matrix updated after the forward processing is processed in reverse to obtain the processed unit calculation and sorting and the processed adjacency matrix. The directed line structure data table is traversed according to the processed unit calculation and sorting to obtain the set of reverse flow streams of the directed line. Then, the upstream and downstream backflow identification method is used to determine the final calculation and sorting and the set of flow streams to be cut according to the processed adjacency matrix. (3) Solution of the entire flow stream: Create a virtual flow stream to replace the set of flow streams that need to be cut and perform the cutting operation. Then, input the initial variable data and the initial value of the virtual flow stream variable in the structure data table to obtain the initial flow stream data table. In sequence, call the basic model of each unit calculated by other variable values ​​or internal iteration to perform the full process calculation and solution. After the initial solution, use the quasi-Newton method to iteratively converge the cut flow streams. After completing the iterative calculation of all cut flow streams, output the final flow stream variable data table to complete the full process solution.

2. The construction method according to claim 1, characterized in that, The loop identification process in step (2) involves forward row-by-row processing and updating the adjacency matrix as follows: The cell corresponding to the row with the nearest non-zero element starting from the i-th element in the i-th column is denoted as K, and the cell corresponding to the column with the farthest non-zero element starting from the i-th element in the i-th row is denoted as P. The adjacency matrix is ​​checked row-by-row according to the initial cell calculation sorting order to determine if there is a backflow. If the adjacency matrix has downstream cells sorted before cell I, that is, the sum of all elements n from the i-th element in the i-th row to the left of the adjacency matrix is ​​not 0, then there is a backflow. At this time, the sum of all elements m from the i-th element in the i-th column upwards is determined. If m < n, then the cell I corresponding to the cell calculation sorting and the adjacency matrix is ​​inserted before cell P; otherwise, it is inserted after cell K. All adjacency matrices are traversed to obtain the forward-processed adjacency matrix.

3. The construction method according to claim 1, characterized in that, The method for constructing the set of reverse flow segments of directed lines in step (2) is as follows: traverse the data table of directed line structure. If the direction from the starting unit to the final unit of a certain directed line is reversed in the unit numbering order, then it is a reverse flow segment of directed lines. All reverse flow segments of directed lines form a set of reverse flow segments of directed lines.

4. The construction method according to claim 1, characterized in that, In step (2), the upstream and downstream backflow identification method is used to confirm the final calculation sort and the set of streams that need to be cut. Specifically, each directed line reverse-order stream is found in the final adjacency matrix and processed one by one. There are n streams. A set of streams that can be cut is created with a length of n. The first stream is E→D. Then, the circulating stream containing the reverse-order stream is found by the upstream and downstream backflow identification method. If there is only one L1 in the circulating stream containing the reverse-order stream, all streams in the circulating stream are added to the set of streams that can be cut. There are a total of n sets. If there are multiple L1, L2, etc. in the circulating stream containing the reverse-order stream, it is determined whether the streams that these circulating streams share are unique except for E→D. If they are not unique, these shared streams are added to the set of streams that can be cut. Then, the judgment of the next stream is entered. After all streams are judged, the set of streams that can be cut is output.

5. The construction method according to claim 1, characterized in that, The full-process calculation and solution described in step (3) is as follows: when calculating each unit sequentially, the mathematical model of the corresponding unit is called in turn, the flow streams flowing into the unit are determined through the structure data table, and the variable values ​​required for solving the model are obtained from the flow stream variable data table. After the solution is completed, the variable values ​​obtained from the solution are filled into the flow stream variable data table, the flow stream variable data table is updated and used for subsequent calculations.

6. The construction method according to claim 1, characterized in that, The step (3) of iteratively converging the cut-off stream using the quasi-Newton method is as follows: After completing the initial full-process calculation and solution, the values ​​of the two virtual streams obtained at the cut-off point are made to correspond one-to-one and equal, resulting in a set of nonlinear equations F(X) = X. This set is then transformed into G(X) = F(X) – X = 0. The initial values ​​of the virtual streams are obtained from the stream variable data table and substituted into the quasi-Newton solution to finally obtain the value X of the return stream. Then restore the severed flow in the flow variable data table, delete the corresponding virtual flow, and enter the return flow value to update the table.

7. The construction method according to claim 1, characterized in that, In step (3), after iterative calculation of all cut-off streams, when there is an iterative loop inside the model and the calculation requires multi-level iterative interaction between the inside and outside, the whole-flow program solution method is integrated into the iterative solution method at the broken stream, and the residual is used as the objective function to realize the iterative convergence process at different levels.

8. A complex process simulation calculation system for process industries, characterized in that, include: Internal Iteration Module: This module relies on the amount of physical property data, including thermodynamic data and transport property data of the material, and is used for the calculation and solution of the basic model; External iteration module: used for mathematical modeling and in-depth analysis of the input flowchart, to implement the method for constructing a complex process simulation calculation system for process industries as described in any one of claims 1-7, wherein it can call the calculation results of the internal iteration module as needed.

9. A computer device, comprising: Memory; as well as A processor configured to execute, based on instructions stored in the memory, the method for constructing a complex process simulation computing system for process industries as described in any one of claims 1-7.

10. A computer-readable storage medium storing computer instructions that, when executed by a processor, enable the method for constructing a complex process simulation computing system for process industries as described in any one of claims 1-7.

Images