A Method for Constructing a Digital Twin of a Winded Tube Heat Exchanger Based on Computational Fluid Dynamics

Abstract

This invention belongs to the field of digital twins, specifically a method for constructing a digital twin of a wound tube heat exchanger based on computational fluid dynamics. The method includes the following steps: collecting the physical parameters of the wound tube heat exchanger and establishing a tube-unit-scale feature element model based on these parameters; acquiring the scale feature data of the tube-unit-scale feature element model and forming a feature element database; establishing a coarse-grained three-dimensional digital twin model and setting initial and boundary conditions; training the feature element database to obtain training models for momentum and energy source terms in the output feature element database; and obtaining flow field variable values ​​for flow and heat transfer, as well as temperature field distribution, through simulation of the coarse-grained three-dimensional digital twin model. This invention achieves a method for constructing a coarse-grained three-dimensional digital twin model, and the simulation implemented is comparable to the physical time scale of the heat exchanger and has similar accuracy to a global model with fine modeling of the tube-side and shell-side structures.

Description

Technical Field

[0001] This invention belongs to the field of digital twins, specifically a method for constructing a digital twin of a wound tube heat exchanger based on computational fluid dynamics. Background Technology

[0002] Spiral wound tube heat exchangers are increasingly used in LNG transportation and storage, and in heat exchange and reaction processes in petrochemical industries due to their advantages such as compact structure, small temperature difference in heat transfer, ability to combine multiple media for heat exchange, high space utilization, and ease of large-scale manufacturing. The operation of a heat exchanger is a flow-driven transport process, and accurate calculations are determined by the scale of the turbulent structure. However, there is a significant difference in spatial scale between the size of the heat exchanger and the characteristic structure of the turbulence. Even with the most advanced high-performance computing equipment, it is difficult to perform traditional fluid dynamics calculations. Therefore, for many years, a truly applicable design, development, operation, and maintenance platform for industrial-grade spiral wound tube heat exchangers has not yet been established. As process equipment develops towards larger scale, energy efficiency, and higher efficiency, the design and development difficulty increases, prompting a gradual shift in equipment development models from an experience-based scale-up approach towards a comprehensive development direction involving multiple scientific disciplines and intelligent design.

[0003] Recognizing the "multi-level, multi-scale" characteristics of flow and heat transfer processes in wound tube heat exchangers—that is, the essential transformation of flow, heat transfer, and mass transfer characteristics in the heat exchange process is caused by changes in the collective behavior of basic unit vortices, bubbles, solid particles, droplets, or characteristic unit structures—this is difficult to obtain directly in macroscopic measurements. The breakthrough to solving this problem lies in establishing descriptions of macroscopic relationships using these basic units, overcoming the limitations and low precision of correlational measurements at the macroscopic scale, and reconstructing the macroscopic system "bottom-up."

[0004] Meanwhile, digital twin technology has begun to be researched and applied in fields such as aviation, construction, manufacturing, healthcare, and smart cities. In the chemical industry, digital twins are considered a key link in realizing the digitization of chemical equipment manufacturing and production processes. Combining the development trends and needs of chemical process equipment R&D, digital twins present the transport and reaction processes of chemical equipment in a virtualized and digital form, deeply analyzing the decisive role of multi-scale issues in the process, thereby breaking through the traditional experience-based R&D model. Furthermore, digitized process equipment can enhance the control over manufacturing and production processes, monitoring and acquiring key data in the multi-scale transport of process equipment through an IoT platform, thereby enabling equipment operation status prediction and performance evaluation.

[0005] Currently, digital twin technology in the chemical industry is still in its early stages of development. Key unresolved technologies remain for digital twin technology of the heat exchange process in wound tube heat exchangers. One of the core bottlenecks is how to perform rapid and accurate multi-scale modeling, which determines the accuracy of the heat exchange process simulation and whether the simulation process can interact with the measurement system in real time. Multi-scale structures possess complex nonlinear and non-equilibrium characteristics, and their initiation mechanisms vary greatly in different production and reaction transport processes. Traditional human assumptions and empirical correlations cannot generalize their applicability, thus affecting accuracy and practicality. Summary of the Invention

[0006] This invention addresses the challenge of efficient real-time simulation of the heat exchange process during the creation of an industrial-grade digital twin of a wound tube heat exchanger. The aim is to provide a method for constructing a digital twin of a wound tube heat exchanger based on Computational Fluid Dynamics (CFD), enabling rapid and accurate multi-scale modeling, thereby providing an industrial-grade digital platform for the research, development, and maintenance of wound tube heat exchangers.

[0007] The technical solution adopted by this invention to achieve the above objectives is: a method for constructing a digital twin of a wound tube heat exchanger based on computational fluid dynamics (CFD), characterized by comprising the following steps:

[0008] Step S1: Collect the physical parameters of the wound tube heat exchanger;

[0009] Step S2: Establish a tube-scale feature element model based on physical parameters, process it using a CFD solver, obtain the scale feature data of the tube-scale feature element model by changing the boundary conditions of the feature element model, and form a feature element database.

[0010] Step S3: Establish a coarse-grained 3D digital twin model, namely, the coarse-grained model governing equations and the setting of initial and boundary conditions;

[0011] Step S4: Train the feature unit database to obtain the training model of momentum source term and energy source term of output feature unit database;

[0012] The momentum source term R1 and the energy source term R2 serve as closing terms in the coarse-grained three-dimensional digital twin model.

[0013] Step S5: By simulating the coarse-grained three-dimensional digital twin model, obtain the flow field variable values ​​of the flow and heat transfer flow field and temperature field distribution to evaluate the heat transfer performance of the wound tube heat exchanger, realize the prediction of heat transfer performance under a certain working condition, and determine the final coarse-grained digital twin model.

[0014] Step S6: Visualize and map the final coarse-grained digital twin model to visualize the heat exchange process of the wound tube heat exchanger.

[0015] The physical parameters mentioned in step S1 include the number of tube bundles, the diameter of the central cylinder, the outer diameter of the heat exchange tubes, the number of tubes per coil in the first layer, the tube spacing in the first layer, the gasket thickness, the number of layers, the material constant of the working fluid flowing in the tubes, and the material constant of the working fluid flowing in the shell.

[0016] In step S3, training the feature unit database to obtain a multi-scale model of flow and heat transfer in the feature unit database specifically involves:

[0017] Machine learning is used to train the feature unit database to obtain the equation-free model of flow and heat transfer, i.e. the training model of flow and heat transfer. The multi-scale model of flow and heat transfer is saved and packaged into a dynamic link library file.

[0018] The machine learning mentioned therein refers to any one of the following: feedforward neural network, support vector machine, convolutional neural network, recurrent neural network, and random forest.

[0019] Step S4 specifically includes:

[0020] Based on the physical parameters of the actual wound tube heat exchanger collected in step S1, a coarse-grained three-dimensional digital twin model was established using CFD.

[0021] The mesh size of the coarse-grained 3D digital twin model is the same as the region size of the tube-scale feature unit model.

[0022] Step S4 includes the following steps:

[0023] 1) Set initial conditions;

[0024] 2) Substitute the initial conditions into the coarse-grained three-dimensional digital twin model to obtain the numerical solution results: the flow field variable values ​​of the first iteration, namely the velocity u1, temperature T1, and pressure p1 of the flow field;

[0025] 3) Input the flow field variable values ​​into the training model of momentum source term and energy source term to obtain the output feature quantity, and then obtain the momentum source term R1 and energy source term R2;

[0026] 4) Substitute the obtained momentum source term R1 and energy source term R2 into the coarse-grained three-dimensional digital twin model as closed terms to solve for the flow field variables, obtaining the flow field variable values ​​after the nth iteration: the velocity u of the flow field. n Temperature T n Pressure p n ;

[0027] Determine whether the residual Res_m of the flow field variables in the nth iteration is less than the threshold E;

[0028] If the value is less than E, construct a coarse-grained three-dimensional digital twin model based on the flow field variable values ​​output in this iteration; otherwise, take the flow field variable values ​​of this iteration as input and return to step 3) until the residual is less than the threshold E.

[0029] Step 1) specifically involves setting the volume average velocity u0, the volume average pressure T0, and the volume average temperature p0.

[0030] The output characteristic quantities include viscous drag source term, inertial drag source term, momentum dispersion term, energy source term, and energy dispersion term;

[0031] Specifically, the momentum source term R1 is obtained through the viscous resistance source term and the inertial resistance source term; the energy source term R2 is obtained through the energy source term and the energy dispersion term.

[0032] In step 4), the residual Res_m specifically refers to:

[0033]

[0034] Where Res_m is the residual, n is the number of iterations, and m(n) is the value of the flow field variable at the nth step.

[0035] The present invention has the following beneficial effects and advantages:

[0036] 1. This invention employs machine learning to accurately and rapidly extract complex nonlinear multi-scale models from a database of feature units representing local information on flow and heat transfer at the tube unit scale of a wound tube heat exchanger. This enables the construction of a coarse-grained three-dimensional digital twin model, allowing the simulation to be comparable to the physical time scale of the heat exchanger and to have similar accuracy to a global model with fine modeling of the tube side and shell side structures.

[0037] 2. The implementation of the digital twin construction method for wound tube heat exchangers based on computational fluid dynamics of the present invention can provide an industrial-grade digital platform for the research, development and maintenance of wound tube heat exchangers.

[0038] 3. The coarse-grained three-dimensional digital twin model of the present invention simplifies the complex geometric structure between unit tubes and shell-side tubes, and describes the influence of complex flow field and temperature field on the overall performance of heat exchanger with a multi-scale model.

[0039] 4. The data-driven paradigm of computational fluid dynamics in this invention can improve the generalization of digital twin systems by expanding the feature unit database, and can be applied to the heat exchange process of wound tube heat exchangers with different parameters. Attached Figure Description

[0040] Figure 1This is a schematic diagram illustrating the technical process of the present invention;

[0041] Figure 2 This is a construction diagram of the tube unit scale feature unit model of the present invention;

[0042] Figure 3 This is a simulation flowchart of the coarse-grained three-dimensional digital twin model of the present invention. Detailed Implementation

[0043] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. The illustrative embodiments and descriptions of the present invention are for illustrative purposes only and are not intended to limit the scope of the invention.

[0044] This invention utilizes a data-driven approach to obtain a multi-scale model of flow and heat transfer at characteristic scales of associated tube units, which serves as the closing term of the flow and heat transfer model for a coarse-grained wound tube heat exchanger digital twin, thereby realizing the construction of a digital twin of a wound tube heat exchanger for convective heat transfer processes. Furthermore, it employs a data-driven approach based on Computational Fluid Dynamics (CFD) to achieve the construction method of the digital twin, and this invention is based on the implementation of a digital twin system.

[0045] like Figure 1 The diagram shown illustrates the technical process of this invention. The method of this invention includes the following steps:

[0046] Step S1: Collect the physical parameters of a real wound tube heat exchanger;

[0047] Step S2: Establish a pipe unit scale feature element model based on physical parameters, perform numerical calculations using a CFD solver, and collect pipe unit scale and equipment-level scale feature quantity data by changing the boundary conditions of the feature element model to form a feature element database.

[0048] Step S3: Use machine learning to train the feature unit database to obtain a multi-scale model of feature unit flow and heat transfer, which serves as the closed term of the control equation for the heat transfer process of the coarse-grained wound tube heat exchanger.

[0049] Step S4: Use CFD technology to establish a coarse-grained three-dimensional digital twin model of the heat exchange process of the wound tube heat exchanger;

[0050] Step S5: The flow field and temperature field distribution of the flow and heat transfer are obtained by simulation calculation of the digital twin system of the wound tube heat exchanger, the heat transfer performance of the wound tube heat exchanger is evaluated, and the heat transfer performance under a certain operating condition is predicted.

[0051] Step S6: Render the simulation data using computer graphics technology to form a virtual digital twin visualization mapping of the heat exchange process.

[0052] The physical parameters used in step S1 include the number of tube bundles, the diameter of the central cylinder, the outer diameter of the heat exchange tubes, the number of tubes per coil in the first layer, the tube spacing in the first layer, the gasket thickness, the number of layers, the material constant of the working medium flowing in the tubes, and the material constant of the working medium flowing in the shell.

[0053] like Figure 2 As shown, the tube unit scale feature unit model in step S2 is a three-dimensional local flow and heat transfer model constructed based on CFD technology, and its region size is approximately 3-5 times the outer diameter of the tube.

[0054] A coarse-grained three-dimensional digital twin model is established based on the volume average theory, namely, the governing equations of the coarse-grained model:

[0055]

[0056]

[0057]

[0058] Where ρ, u, and T are the density, velocity, and temperature of the flowing working fluid, respectively; p is the pressure; μ is the viscosity; g is the acceleration due to gravity; and c is the acceleration due to gravity. p λ and φ represent specific heat capacity and thermal conductivity, respectively. φ is the porosity of the characteristic element model. Represents the intrinsic mean. Representative characteristics are small in quantity. Represents velocity, pressure, or temperature in the flow field, i.e., u, p, or T, where t is time. R1 is a vector differential operator, and R2 is the momentum source term and energy source term, respectively.

[0059] In step S4), the training model for the momentum source term and the energy source term is specifically as follows:

[0060]

[0061]

[0062] The viscous resistance source term is:

[0063] The inertial drag source term is:

[0064] The momentum dispersion term is:

[0065] The energy source term is:

[0066] The energy diffusion term is:

[0067] Where V is the average volume, A i The interface area, n represents the area between A and B. i A vertical unit vector.

[0068] The feature unit database established in step S2 has input feature quantities including feature unit volume, volume average velocity, volume average pressure, and volume average temperature, and output feature quantities including viscous drag source term, inertial drag source term, momentum dispersion term, energy source term, and energy dispersion term.

[0069] The multi-scale model of flow and heat transfer of the feature units in step S3 is obtained by training the feature unit database using machine learning techniques, including feedforward neural networks, support vector machines, convolutional neural networks, recurrent neural networks, and random forests. The multi-scale model in step S3 is an equation-free model and is encapsulated as a DLL (Dynamic Link Library) file.

[0070] The coarse-grained three-dimensional digital twin model in step S4 is constructed using CFD technology based on the physical parameters, wherein the mesh size of the coarse-grained three-dimensional digital twin model is the same as the size of the region of the tube-scale feature unit model.

[0071] like Figure 3 The diagram shown is a simulation flowchart of the coarse-grained three-dimensional digital twin model of the present invention, which specifically includes steps T1 to T4.

[0072] Step T1: Set the initial conditions for the coarse-grained 3D digital twin model, including volume-averaged velocity. i Volume average pressure i Volume average temperature <t> i That is: u0, T0, p0;

[0073] Step T2: Perform the first iteration of the model and read the volume, volume-average velocity, volume-average pressure, and volume-average temperature of each coarse-grained unit; obtain the flow field variable values, namely the velocity u1, temperature T1, and pressure p1 of the flow field.

[0074] Step T3: Call the DLL file to read the multi-scale model, and calculate the viscous drag source term, inertial drag source term, momentum dispersion term, energy source term, and energy dispersion term using the above coarse-grained unit parameters, i.e., formulas (6) to (10); obtain the momentum source term R1 and energy source term 22 using formulas (4) to (5).

[0075] The momentum source term R1 and energy source term R2 are read into the coarse-grained three-dimensional digital twin model and the second step of iterative calculation is performed.

[0076] After step T4, the second iterative calculation, repeat step T2 until the calculation reaches the convergence requirement.

[0077] The governing equations of the coarse-grained CFD model can be solved using numerical methods such as the finite difference method, the finite volume method, and the finite element method. For example, the convergence requirement of the finite volume method is that the residuals reach a specified value, i.e.:

[0078]

[0079] Where Res_m is the residual, n is the number of iterations, and m(n) is the value of the flow field variable at the nth step.

[0080] The coarse-grained three-dimensional digital twin model of the present invention simplifies the complex geometric structure between unit tubes and shell-side tubes, and describes the influence of complex flow field and temperature field on the overall performance of heat exchanger with a multi-scale model.

[0081] In this invention, the simulation time of the coarse-grained three-dimensional digital twin model is comparable to the physical time scale of the heat exchanger; the calculation accuracy of the coarse-grained three-dimensional digital twin model is similar to that of the global model of a wound tube heat exchanger that finely describes the tube-side structure and shell-side structure.

[0082] The data-driven paradigm can improve the generalization of digital twin systems by expanding the feature unit database, and can be applied to the heat exchange process of wound tube heat exchangers with different parameters.

[0083] The aforementioned coarse-grained 3D digital twin model simplifies the complex geometry between unit tubes and shell-side tubes, and describes the impact of complex flow and temperature fields on the overall performance of the heat exchanger using a multi-scale model. The simulation time of the coarse-grained 3D digital twin model is comparable to the physical timescale of the heat exchanger, and its computational accuracy is similar to that of a finely detailed global model of a wound-tube heat exchanger that describes both the tube-side and shell-side structures. The data-driven paradigm can improve the generalization of the digital twin system by expanding the feature unit database, enabling its application to heat exchange processes in wound-tube heat exchangers with different parameters.< / t>

Claims

1. A method for constructing a digital twin of a wound tube heat exchanger based on computational fluid dynamics, characterized in that, Includes the following steps: Step S1: Collect the physical parameters of the wound tube heat exchanger; Step S2: Establish a tube-scale feature element model based on physical parameters, process it using a CFD solver, obtain the scale feature data of the tube-scale feature element model by changing the boundary conditions of the feature element model, and form a feature element database. Step S3: Establish a coarse-grained 3D digital twin model, namely, the coarse-grained model governing equations and the setting of initial and boundary conditions; Step S4: Train the feature unit database to obtain the training model of momentum source term and energy source term of output feature unit database; The momentum source term R1 and the energy source term R2 serve as closing terms in the coarse-grained three-dimensional digital twin model. Step S5: By simulating the coarse-grained three-dimensional digital twin model, obtain the flow field variable values ​​of the flow and heat transfer flow field and temperature field distribution to evaluate the heat transfer performance of the wound tube heat exchanger, realize the prediction of heat transfer performance under a certain working condition, and determine the final coarse-grained digital twin model. Step S6: Visualize and map the final coarse-grained digital twin model to visualize the heat exchange process of the wound tube heat exchanger.

2. The method for constructing a digital twin of a wound tube heat exchanger based on computational fluid dynamics according to claim 1, characterized in that, The physical parameters mentioned in step S1 include the number of tube bundles, the diameter of the central cylinder, the outer diameter of the heat exchange tubes, the number of tubes per coil in the first layer, the tube spacing in the first layer, the gasket thickness, the number of layers, the material constant of the working fluid flowing in the tubes, and the material constant of the working fluid flowing in the shell.

3. The method for constructing a digital twin of a wound tube heat exchanger based on computational fluid dynamics according to claim 1, characterized in that, In step S3, training the feature unit database to obtain a multi-scale model of flow and heat transfer in the feature unit database specifically involves: Machine learning is used to train the feature unit database to obtain the equation-free model of flow and heat transfer, i.e. the training model of flow and heat transfer. The multi-scale model of flow and heat transfer is saved and packaged into a dynamic link library file. The machine learning mentioned therein refers to any one of the following: feedforward neural network, support vector machine, convolutional neural network, recurrent neural network, and random forest.

4. The method for constructing a digital twin of a wound tube heat exchanger based on computational fluid dynamics according to claim 1, characterized in that, Step S4 specifically includes: Based on the physical parameters of the actual wound tube heat exchanger collected in step S1, a coarse-grained three-dimensional digital twin model was established using CFD. The mesh size of the coarse-grained 3D digital twin model is the same as the region size of the tube-scale feature unit model.

5. The method for constructing a digital twin of a wound tube heat exchanger based on computational fluid dynamics according to claim 1, characterized in that, Step S4 includes the following steps: 1) Set initial conditions; 2) Substitute the initial conditions into the coarse-grained three-dimensional digital twin model to obtain the numerical solution results: the flow field variable values ​​of the first iteration, namely the velocity u1, temperature T1, and pressure p1 of the flow field; 3) Input the flow field variable values ​​into the training model of momentum source term and energy source term to obtain the output feature quantity, and then obtain the momentum source term R1 and energy source term R2; 4) Substitute the obtained momentum source term R1 and energy source term R2 into the coarse-grained three-dimensional digital twin model as closed terms to solve for the flow field variables, obtaining the flow field variable values ​​after the nth iteration: the velocity u of the flow field. n Temperature T n Pressure p n ; Determine whether the residual Res_m of the flow field variables in the nth iteration is less than the threshold E; If the value is less than E, construct a coarse-grained three-dimensional digital twin model based on the flow field variable values ​​output in this iteration; otherwise, take the flow field variable values ​​of this iteration as input and return to step 3) until the residual is less than the threshold E.

6. The method for constructing a digital twin of a wound tube heat exchanger based on computational fluid dynamics according to claim 5, characterized in that, Step 1) specifically involves setting the volume average velocity u0, the volume average pressure T0, and the volume average temperature p0.

7. The method for constructing a digital twin of a wound tube heat exchanger based on computational fluid dynamics according to claim 5, characterized in that, The output characteristic quantities include viscous drag source term, inertial drag source term, momentum dispersion term, energy source term, and energy dispersion term; Specifically, the momentum source term R1 is obtained through the viscous resistance source term and the inertial resistance source term; the energy source term R2 is obtained through the energy source term and the energy dispersion term.

8. The method for constructing a digital twin of a wound tube heat exchanger based on computational fluid dynamics according to claim 5, characterized in that, In step 4), the residual Res_m specifically refers to: Where Res_m is the residual, n is the number of iterations, and m(n) is the value of the flow field variable at the nth step.

Images