A multi-region porous medium filter parameter fast optimization closed-loop simulation method, device, computer equipment and storage medium

Abstract

The application provides a multi-region porous medium filter parameter fast optimization closed-loop simulation method and device, computer equipment and a storage medium, and belongs to the field of nuclear facility ventilation filter design and numerical simulation automation. The application solves the problems of low efficiency of manual trial and error of multi-region parameters, easy error of resistance parameter mapping, non-reproducible simulation results and poor batch operation stability in traditional filter design. The application is aimed at a multi-region porous medium filter, constructs a multi-region parameter space in a closed loop of "simulation-extraction-evaluation-updating", combines adaptive optimization and hierarchical screening, automatically maps resistance parameters, calls CFD solution, extracts indexes and iteratively optimizes, and quickly outputs an optimal design point, so that the number of manual trial and error and simulation times are greatly reduced, the design efficiency is improved, the results are traceable, and the needs of nuclear facilities and other engineering scenes are met.

Description

Technical Field

[0001] This invention relates to the field of nuclear facility ventilation filter design and numerical simulation automation technology, and in particular to a closed-loop simulation method, device, computer equipment and storage medium for rapid optimization of parameters of multi-zone porous media filters. Background Technology

[0002] Spent fuel reprocessing plants and related nuclear facilities may generate aerosols containing radionuclides or heavy metals during processes such as dissolution, extraction, and evaporation concentration. Ventilation and purification systems typically use multi-stage filters to capture and control these pollutants. To meet radiation safety and environmental emission requirements, engineering design must comprehensively balance filtration efficiency, pressure drop, air volume, space requirements, filter media lifespan, and ease of maintenance.

[0003] Current engineering designs commonly employ experience-based selection or single-factor trial calculations: designers perform CFD verification based on a small number of typical parameters (such as average wire diameter, average filler density, and empirical permeability). If the results do not meet requirements, the parameters are manually adjusted and the calculation is repeated. This process has the following problems: (1) The design variables of multi-region / multi-layer structures have high dimensionality, low efficiency of manual trial and error, and difficulty in obtaining a globally optimal solution within a limited period; (2) The resistance parameters of porous media often need to be calculated or inverted based on the filter material structure parameters. Manual writing is prone to errors and it is difficult to ensure traceability. (3) The simulation post-processing indicators are scattered, making it difficult to form a unified evaluation and convergence criterion, resulting in the design process being unreproducible; (4) When there are differences in the CFD solver version, data model interface or script calling method, batch running and result extraction are prone to failure.

[0004] Therefore, there is an urgent need for a closed-loop platform and method that integrates "pore parameter update - batch call of solver - automatic extraction of index - adaptive optimization" to output a single optimal design point in a short time and form traceable and reproducible simulation and evaluation data. Summary of the Invention

[0005] This invention proposes a closed-loop simulation method, device, computer equipment, and storage medium for rapid optimization of parameters in multi-region porous media filters. By constructing independent parameter spaces for multiple regions, and using "simulation-extraction-evaluation-update-resimulation" as the closed-loop main line, combined with an adaptive sampling optimization strategy and a hierarchical screening mechanism, it achieves automatic mapping of porous media parameters, batch calling of CFD solvers, automatic extraction of simulation indicators, and rapid output of optimal design points. This solves the problems of low efficiency of manual trial and error for multi-region parameters, error-prone mapping of resistance parameters, non-reproducible simulation results, and poor stability of batch operation in traditional filter design.

[0006] A closed-loop simulation method for rapid parameter optimization of multi-region porous media filters includes the following steps: S1. Obtain simulation engineering input: The user creates the filter geometry model and mesh, defines at least one fluid region and at least two porous media regions, and sets boundary conditions, turbulence / laminar flow models, discretization schemes, residual convergence criteria, and monitoring quantities. S2. Construct parameter space: Set the design variable range and step size or sampling distribution for each porous media region. The design variables should include at least fiber diameter and filling density. S3. Generate candidate design points: Generate a set of candidate design points from the parameter space according to the preset optimization strategy. The candidate design points include the combination of design variables for each porous medium region. S4. Parameter Mapping and Writing: Automatically map candidate design points to the resistance parameters of each porous medium region and write them into the CFD solver. The resistance parameters include at least the viscous drag coefficient. S5. Coupled Solution: Control the CFD solver to initialize and iteratively calculate the model after the resistance parameters are written in, and obtain the steady-state or quasi-steady-state flow field and pressure field. S6. Index Extraction: Perform automatic post-processing on the solution results to extract simulation indices, including at least representative velocity statistics of the porous media region and pressure difference statistics of key sections. S7. Evaluation and Update: Under constraints, calculate the comprehensive target value based on simulation indicators, and update the optimization strategy accordingly to generate the next set of candidate design points. S8. Termination and Output: When the convergence criterion, the upper limit of the number of simulations, or the time limit criterion is met, the closed-loop optimization is terminated, and the single optimal design point and its corresponding resistance parameters, simulation indicators, configuration files, and result data are output.

[0007] Furthermore, the optimization strategy in S3 is an adaptive sampling strategy, which includes: training a surrogate model based on the evaluated design points to approximate the mapping relationship between "design variables and comprehensive objectives". The surrogate model is any one of Gaussian process regression, random forest regression or neural network regression. The next set of candidate design points is selected under constraints based on the acquisition function, which is any one of expected improvement, upper confidence bound or probabilistic improvement.

[0008] Furthermore, in S4, resistance parameter inversion or calculation is performed for each porous media region: the equivalent permeability is calculated based on the fiber diameter and packing density of the region, and the reciprocal of the equivalent permeability is written into the CFD solver as the viscous resistance coefficient.

[0009] Furthermore, in S6, the volume average velocity modulus is extracted for each porous media region, and the overall average velocity of the filter is obtained by weighting the volume of the porous media region. At the same time, the minimum and maximum values ​​of the volume average velocity modulus of each porous media region are output. At least two monitoring surfaces are set on the outside and inside of the filter, and the area-weighted average static pressure and maximum static pressure of the monitoring surfaces are calculated respectively. The average pressure drop and maximum pressure drop are obtained from the difference.

[0010] Furthermore, in S7, the overall objective is to minimize the pressure drop while satisfying either the filtration efficiency threshold or the penetration threshold, or to maximize the efficiency while satisfying the pressure drop threshold, wherein the filtration efficiency is calculated by a user-pluggable efficiency model module based on representative velocity statistics and particle parameters.

[0011] Furthermore, the closed-loop optimization from S3 to S7 adopts a hierarchical screening mechanism, including: firstly, using a low number of iterations or a simplified model for coarse screening to obtain a candidate subset, and then using a high number of iterations or a high-precision model for fine screening of the candidate subset to output a single optimal design point.

[0012] Furthermore, the simulation engineering input in S1 and the parameter space constructed in S2 are saved in the form of a structured configuration file. The structured configuration file includes: porous medium region name, thickness, design variable range and step size, material name, temperature and pressure conditions, monitoring surface name, number of solution iteration steps and additional solver commands, and automatically records the configuration snapshot corresponding to the candidate design point after each closed-loop iteration.

[0013] A closed-loop simulation device for rapid parameter optimization of multi-region porous media filters, applied to the aforementioned closed-loop simulation method for rapid parameter optimization of multi-region porous media filters, includes: a simulation engineering input acquisition module, a parameter space construction module, a candidate design point generation module, a parameter mapping and writing module, a coupled solution module, an index extraction module, an evaluation and update module, and a termination and output module. The simulation engineering input module is used by the user to create the filter geometry model and mesh, define at least one fluid region and at least two porous media regions, and set boundary conditions, turbulence / laminar flow models, discretization schemes, residual convergence criteria and monitoring quantities; Construct a parameter space module to set the design variable range and step size or sampling distribution for each porous media region. The design variables include at least fiber diameter and packing density. The candidate design point generation module is used to generate a set of candidate design points from the parameter space according to a preset optimization strategy. The candidate design points include the combination of design variables for each porous medium region. The parameter mapping and writing module is used to automatically map candidate design points to the resistance parameters of each porous medium region and write them into the CFD solver, wherein the resistance parameters include at least the viscous drag coefficient. The coupled solution module is used to control the CFD solver to initialize and iteratively calculate the model after the resistance parameters are written in, so as to obtain the steady-state or quasi-steady-state flow field and pressure field. The index extraction module is used to automatically post-process the solution results and extract simulation indices, including at least representative velocity statistics of porous media regions and pressure difference statistics of key sections. The evaluation and update module is used to calculate the comprehensive target value based on simulation indicators under constraints, and update the optimization strategy accordingly to generate the next set of candidate design points. The Termination and Output module is used to terminate the closed-loop optimization when the convergence criterion, the upper limit of the number of simulations, or the time limit criterion is met, and outputs a single optimal design point and its corresponding resistance parameters, simulation indicators, configuration files, and result data.

[0014] A storage medium storing a computer program, which, when executed by a processor, implements the aforementioned closed-loop simulation method for rapid optimization of parameters of a multi-region porous media filter.

[0015] A computer device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method for rapid optimization of parameters of a multi-region porous media filter using closed-loop simulation.

[0016] Compared with the prior art, the present invention achieves significant beneficial effects through the above technical solution: 1. Closed-loop automation: Enables batch updates of resistance parameters in porous media regions, batch calls to solvers, and automatic extraction of post-processing indices, reducing manual operation and errors.

[0017] 2. Rapid optimization: Through strategies such as adaptive sampling, hierarchical screening, or surrogate models, a single optimal design point is output within an hourly cycle, reducing the number of simulations.

[0018] 3. Traceability and reproducibility: The input, output and version information of each iteration are fixed through structured configuration files and iterative snapshot mechanism, which facilitates review and comparison.

[0019] 4. Expandable: The parameter mapping module and efficiency evaluation module are pluggable and adaptable to different filter material structure parameters, different particle sizes and different evaluation standards.

[0020] 5. Project Adaptation: The interface adaptation layer shields CFD solver version differences, improving the stability of batch operations. Attached Figure Description

[0021] Figure 1 This is a flowchart of a closed-loop simulation method for rapid optimization of parameters of a multi-region porous media filter according to the present invention. Figure 2 This is a schematic diagram of the CFD solver project file selection interface of the client of the present invention in a specific embodiment. Figure 3 This is a schematic diagram of the solver running parameter setting interface of the client of the present invention in a specific embodiment. Figure 4 A schematic diagram of the porous media region parameter setting interface of the client of the present invention in a specific embodiment; Figure 5 This is a schematic diagram of the export interface for monitoring and displaying the optimization task in the client of the present invention in a specific embodiment. Figure 6 The diagram below shows a closed-loop simulation device for rapid optimization of parameters of a multi-region porous media filter according to the present invention, which can be deployed as a stand-alone software or a client-server architecture. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Reference Figure 1 As shown, a closed-loop simulation method for rapid optimization of parameters of a multi-region porous media filter includes the following steps: S1. Obtain simulation engineering input: The user creates the filter geometry model and mesh, defines at least one fluid region and at least two porous media regions, and sets boundary conditions, turbulence / laminar flow models, discretization schemes, residual convergence criteria, and monitoring quantities. S2. Construct parameter space: Set the design variable range and step size or sampling distribution for each porous media region. The design variables should include at least fiber diameter and filling density. S3. Generate candidate design points: Generate a set of candidate design points from the parameter space according to the preset optimization strategy. The candidate design points include the combination of design variables for each porous medium region. S4. Parameter Mapping and Writing: Automatically map candidate design points to the resistance parameters of each porous medium region and write them into the CFD solver. The resistance parameters include at least the viscous drag coefficient. S5. Coupled Solution: Control the CFD solver to initialize and iteratively calculate the model after the resistance parameters are written in, and obtain the steady-state or quasi-steady-state flow field and pressure field. S6. Index Extraction: Perform automatic post-processing on the solution results to extract simulation indices, including at least representative velocity statistics of the porous media region and pressure difference statistics of key sections. S7. Evaluation and Update: Under constraints, calculate the comprehensive target value based on simulation indicators, and update the optimization strategy accordingly to generate the next set of candidate design points. S8. Termination and Output: When the convergence criterion, the upper limit of the number of simulations, or the time limit criterion is met, the closed-loop optimization is terminated, and the single optimal design point and its corresponding resistance parameters, simulation indicators, configuration files, and result data are output.

[0024] Specifically, this invention constructs a multi-regional independent parameter space to achieve rapid optimization and simulation of parameters for multi-regional porous media filters through a complete closed-loop process. This eliminates the need for repeated manual parameter adjustments and simulation intervention, significantly reducing manual operations and human error. It effectively solves the problems of high-dimensional design variables and low efficiency of manual trial and error in multi-regional / multi-layer structure designs. By automatically generating candidate design points, mapping and writing resistance parameters, calling the CFD solver, extracting simulation indicators, and updating optimization strategies, it can efficiently locate the globally optimal solution within a finite timeframe, avoiding the drawbacks of traditional designs that struggle to simultaneously consider multiple objectives such as filtration efficiency and pressure drop. Furthermore, this method can output a single optimal design point along with its corresponding resistance parameters, simulation indicators, configuration files, and result data, completely solidifying the entire design and simulation process information. This ensures the traceability and reproducibility of the design process and is suitable for the design requirements of multi-regional porous media filtration scenarios such as ventilation filters in nuclear facilities.

[0025] Furthermore, the optimization strategy in S3 is an adaptive sampling strategy, which includes: training a surrogate model based on the evaluated design points to approximate the mapping relationship between "design variables and comprehensive objectives". The surrogate model is any one of Gaussian process regression, random forest regression or neural network regression. The next set of candidate design points is selected under constraints based on the acquisition function, which is any one of expected improvement, upper confidence bound or probabilistic improvement.

[0026] Specifically, this invention employs an adaptive sampling optimization strategy, using a surrogate model to approximate the mapping relationship between "design variables and overall objectives." This eliminates the need for full simulation calculations of all design variable combinations within the parameter space, significantly reducing unnecessary simulation overhead and dramatically improving the efficiency of parameter optimization for multi-region porous media filters. Simultaneously, by selecting the next set of candidate design points under constraints based on the acquisition function, it can accurately focus on the parameter regions most likely to improve the overall objective, avoiding blind searches during the optimization process and making the optimization direction more targeted, thus facilitating faster convergence to the globally optimal solution. Furthermore, the flexible selection of various surrogate models such as Gaussian process regression, random forest regression, and neural network regression, as well as various acquisition functions such as expected improvement, upper confidence bound, and probabilistic improvement, can adapt to the design requirements of filters with different multi-region structures and different filter media characteristics, enhancing the robustness and adaptability of the optimization strategy and further ensuring the efficient output of the optimal design point that meets the requirements of filtration efficiency and pressure drop balance within a finite simulation cycle.

[0027] Furthermore, in S4, resistance parameter inversion or calculation is performed for each porous media region: the equivalent permeability is calculated based on the fiber diameter and packing density of the region, and the reciprocal of the equivalent permeability is written into the CFD solver as the viscous resistance coefficient.

[0028] Specifically, this invention automatically calculates the equivalent permeability for each porous media region based on its fiber diameter and packing density, and then uses the reciprocal of the equivalent permeability as the viscous resistance coefficient, writing it into the CFD solver. This achieves a precise correlation between resistance parameters and the core structural parameters of the filter media, completely avoiding the errors that easily occur when manually calculating, inverting, or writing resistance parameters in traditional designs, while also eliminating the tedious process of repeatedly verifying parameters manually. This automated mapping method based on the inherent structural parameters of the filter media ensures the accuracy and rationality of the resistance parameters for each porous media region, providing a reliable foundation for the accurate simulation of the subsequent CFD solver, making the calculated flow field and pressure field results more consistent with actual filtration scenarios. Furthermore, this method can adapt to the personalized structural parameters of different porous media regions, perfectly meeting the design requirements of multi-region filters, further ensuring the effectiveness of parameter iteration during closed-loop optimization, helping to quickly converge to the globally optimal design point, and making the source of resistance parameters clearly traceable.

[0029] Furthermore, in S6, the volume average velocity modulus is extracted for each porous media region, and the overall average velocity of the filter is obtained by weighting the volume of the porous media region. At the same time, the minimum and maximum values ​​of the volume average velocity modulus of each porous media region are output. At least two monitoring surfaces are set on the outside and inside of the filter, and the area-weighted average static pressure and maximum static pressure of the monitoring surfaces are calculated respectively. The average pressure drop and maximum pressure drop are obtained from the difference.

[0030] Specifically, this invention can accurately extract the volumetric average velocity modulus of each porous media region, and obtain the overall average velocity of the filter through volume weighting. It also outputs the minimum and maximum values ​​of the velocity modulus for each region, providing a comprehensive understanding of the flow characteristics of a single porous media region and a clear picture of the overall velocity distribution of the filter. This effectively reflects the velocity uniformity under a multi-region structure, providing crucial data support for the accurate calculation of subsequent comprehensive objectives. Furthermore, by setting monitoring surfaces on the outer and inner sides of the filter, the area-weighted average static pressure and maximum static pressure are calculated, and the average pressure drop and maximum pressure drop are derived. This comprehensively covers the overall pressure drop level and extreme cases, avoiding the limitations of traditional single pressure drop indicators and making the simulation indicators more comprehensive and valuable. This systematic indicator extraction method automates and standardizes the post-processing of simulation results, avoiding the tedious operations and errors of manual indicator extraction, and providing a unified comparison benchmark for indicator data from different iteration cycles. This further ensures the reproducibility of the design process and the rationality of optimization strategy updates.

[0031] Furthermore, in S7, the overall objective is to minimize the pressure drop while satisfying either the filtration efficiency threshold or the penetration threshold, or to maximize the efficiency while satisfying the pressure drop threshold, wherein the filtration efficiency is calculated by a user-pluggable efficiency model module based on representative velocity statistics and particle parameters.

[0032] Specifically, this invention clarifies that the overall objective is to minimize pressure drop while meeting filtration efficiency or penetration threshold constraints, or to maximize efficiency while meeting pressure drop threshold constraints. This precisely aligns with the core requirement of nuclear facility ventilation filters for balancing filtration performance and energy consumption, avoiding the performance imbalance caused by solely pursuing a single indicator in traditional designs. Furthermore, the filtration efficiency is calculated using a user-pluggable efficiency model module based on representative velocity statistics and particle parameters. This allows for flexible efficiency calculations to adapt to application scenarios with different filter material structures and particle spectrum characteristics. Evaluation criteria can be changed without modifying the overall simulation framework, greatly enhancing the method's scalability and practicality.

[0033] Furthermore, the closed-loop optimization from S3 to S7 adopts a hierarchical screening mechanism, including: firstly, using a low number of iterations or a simplified model to perform coarse screening to obtain a candidate subset, and then using a high number of iterations or a high-precision model to perform fine screening on the candidate subset, so as to output the single optimal design point in a shorter time.

[0034] Specifically, this invention employs a hierarchical screening mechanism in the closed-loop optimization process. First, candidate design points are coarsely screened using a low number of iterations or a simplified model, quickly eliminating design points that clearly do not meet the constraints or have poor overall objectives. This efficiently selects a subset of potential candidates, avoiding the resource waste and time consumption associated with calculating all candidate points using high-precision models. Based on this, the candidate subset is then finely screened using a high number of iterations or a high-precision model to ensure the accuracy of simulation calculations for the core potential parameter combinations. This guarantees that the final output of a single optimal design point accurately meets the balance requirements of filtration efficiency and pressure drop. This combination of "coarse screening + fine screening" significantly shortens the overall optimization cycle, allowing parameter optimization of multi-region porous media filters to be completed in a shorter time, while effectively ensuring the reliability and engineering applicability of the optimal design point, perfectly balancing optimization efficiency and computational accuracy.

[0035] Furthermore, the simulation engineering input in S1 and the parameter space constructed in S2 are saved in the form of a structured configuration file. The structured configuration file includes: porous medium region name, thickness, design variable range and step size, material name, temperature and pressure conditions, monitoring surface name, number of solution iteration steps and additional solver commands, and automatically records the configuration snapshot corresponding to the candidate design point after each closed-loop iteration.

[0036] Specifically, this invention saves the simulation engineering input and constructed parameter space in the form of a structured configuration file, covering key information such as the name of the porous medium region, thickness, design variable range, and step size. This ensures that all core inputs and parameter settings are clear, organized, and complete, providing a fundamental guarantee for the consistency and accuracy of the simulation process. Simultaneously, after each closed-loop iteration, a configuration snapshot corresponding to the candidate design point is automatically recorded, completely solidifying the input conditions, parameter combinations, and related settings for each iteration. This ensures that every step of the optimization and simulation process is traceable, thoroughly solving the problems of unrecorded parameter adjustments and unreproducible simulation processes in traditional design. This approach not only facilitates subsequent review, comparison, and traceability of the design process but also provides complete data support for possible parameter adjustments, scheme optimization, or troubleshooting.

[0037] A closed-loop simulation device for rapid parameter optimization of multi-region porous media filters, applied to the aforementioned closed-loop simulation method for rapid parameter optimization of multi-region porous media filters, includes: a simulation engineering input acquisition module, a parameter space construction module, a candidate design point generation module, a parameter mapping and writing module, a coupled solution module, an index extraction module, an evaluation and update module, and a termination and output module. The simulation engineering input module is used by the user to create the filter geometry model and mesh, define at least one fluid region and at least two porous media regions, and set boundary conditions, turbulence / laminar flow models, discretization schemes, residual convergence criteria and monitoring quantities; Construct a parameter space module to set the design variable range and step size or sampling distribution for each porous media region. The design variables include at least fiber diameter and packing density. The candidate design point generation module is used to generate a set of candidate design points from the parameter space according to a preset optimization strategy. The candidate design points include the combination of design variables for each porous medium region. The parameter mapping and writing module is used to automatically map candidate design points to the resistance parameters of each porous medium region and write them into the CFD solver, wherein the resistance parameters include at least the viscous drag coefficient. The coupled solution module is used to control the CFD solver to initialize and iteratively calculate the model after the resistance parameters are written in, so as to obtain the steady-state or quasi-steady-state flow field and pressure field. The index extraction module is used to automatically post-process the solution results and extract simulation indices, including at least representative velocity statistics of porous media regions and pressure difference statistics of key sections. The evaluation and update module is used to calculate the comprehensive target value based on simulation indicators under constraints, and update the optimization strategy accordingly to generate the next set of candidate design points. The Termination and Output module is used to terminate the closed-loop optimization when the convergence criterion, the upper limit of the number of simulations, or the time limit criterion is met, and outputs a single optimal design point and its corresponding resistance parameters, simulation indicators, configuration files, and result data.

[0038] Specifically, the device of this invention breaks down the entire process of rapid optimization closed-loop simulation of multi-region porous media filter parameters into eight clearly defined core modules through a modular architecture. Each module performs its own function and is closely connected, forming a complete technical chain from simulation engineering input and parameter space construction to optimal design point output, ensuring the orderliness and efficiency of the simulation and optimization process. The simulation engineering input acquisition module and parameter space construction module are precisely adapted to multi-region scenarios, and can independently complete the basic settings and parameter range definition for each porous media region, providing a solid foundation for subsequent targeted optimization. The collaborative operation of the candidate design point generation module, parameter mapping and writing module, coupled solution module, and index extraction module realizes the automated flow of candidate parameter generation, automatic drag parameter mapping, CFD solver calling, and simulation index extraction, completely eliminating the dependence on manual operation and effectively reducing human error. The evaluation and update module ensures that the optimization direction always focuses on the improvement of the comprehensive goal by dynamically adjusting the optimization strategy, while the termination and output module outputs the optimal design point and the corresponding drag parameters, simulation indexes, and configuration files, ensuring the traceability and reproducibility of the design process. This modular design not only makes the device structure clear and easy to understand, facilitating maintenance and upgrades, but also allows for flexible adaptation to application scenarios with different filter media types and particle size distribution characteristics.

[0039] A storage medium storing a computer program, which, when executed by a processor, implements the aforementioned closed-loop simulation method for rapid optimization of parameters of a multi-region porous media filter.

[0040] A computer device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method for rapid optimization of parameters of a multi-region porous media filter using closed-loop simulation.

[0041] Specifically, the storage medium of this invention, by storing the corresponding computer program, can solidify the closed-loop simulation method for rapid optimization of multi-region porous media filter parameters into a reusable digital resource. This facilitates dissemination, deployment, and invocation across different devices, eliminating the need for repeated development of core logic and lowering the barrier to promotion and application of the method. Meanwhile, the computer equipment, by integrating memory, processor, and related programs, provides a stable hardware operating environment for this simulation method. Relying on the processor's computing power and the memory's storage support, it can efficiently execute a series of complex processes, including parameter space construction, candidate design point generation, CFD solver invocation, simulation index extraction, and optimization strategy updates, ensuring the smoothness and efficiency of the closed-loop optimization process. The two work together to enable the rapid optimization closed-loop simulation method for multi-region porous media filters to be easily implemented in engineering scenarios such as the design of ventilation filters for nuclear facilities. This fully leverages the method's advantages of closed-loop automation, rapid optimization, traceability, and scalability, without requiring the construction of a complex operating environment. This ensures both the accuracy and reliability of the simulation results and improves the convenience and efficiency of the design work.

[0042] The following are specific embodiments of the present invention: The closed-loop optimization method of the present invention may include an initialization phase, an iteration phase, and a termination phase. The initialization phase is used to complete the engineering input check and parameter space construction; the iteration phase is used to perform candidate generation, parameter mapping, coupled solution, and index extraction; and the termination phase is used to output the optimal point and trace the data.

[0043] (I) Engineering Input and Parameter Space Construction In one embodiment, the user attaches Figure 2 With appendix Figure 3 The interface shown allows you to select a CFD solver project file (preferably a case file) and set the solution accuracy, number of processors, initialization method, iteration steps, and additional solver commands. Users also need to pre-define the geometry and mesh within the solver, defining at least one fluid region and at least one porous media region, and naming the key monitoring surfaces (e.g., the outer and inner surfaces of a filter). Furthermore, users can add additional... Figure 4The table showing porous region parameters allows you to input the region name, thickness, and search range and step size for fiber diameter and packing density for each porous media region. For multilayer filters or spatial partitioned filters, multiple porous media regions can be configured separately for each layer or partition.

[0044] (II) Candidate Generation and Tiered Screening In one embodiment, the candidate generation module employs a hierarchical screening approach: the first stage uses a coarse step size or simplified solution settings for rapid scanning to obtain a subset of candidate points that meet the basic constraints; the second stage performs fine-grained calculations only on the candidate subset using a fine step size and strict convergence criteria, thereby obtaining a single optimal point within an hourly timeframe. In another preferred embodiment, the candidate generation module employs adaptive sampling: first, an initial sample set is obtained through Latin hypercube or random sampling; then, a surrogate model is trained and combined with a sampling function to progressively select new candidate points, concentrating computational resources on the region most likely to improve the overall objective.

[0045] (III) Parameter Mapping and Pore Resistance Writing As attached Figure 3 As shown, the parameter mapping module converts candidate design points into resistance parameters for porous media regions. To avoid dependence on specific theoretical models, the conversion can be performed in any of the following ways: Method A: Calculate the equivalent permeability based on the filter media structural parameters and obtain the viscous resistance coefficient from its reciprocal; Method B: Directly look up the resistance parameters based on pre-calibrated data or an empirical mapping table; Method C: Inversely calculate the resistance parameters under a given target pressure drop or target flow rate condition to match the macroscopic pressure drop response of the porous region to the target value.

[0046] In this embodiment, the drag parameters are written by the simulation coupling module through the solver interface, and drag coefficients in three directions can be written for the anisotropic porous medium.

[0047] (iv) Coupled solution and stability control As attached Figure 3 As shown, the simulation coupling module includes a material property setting unit, a porous region parameter writing unit, and an initialization and iteration control unit. In one embodiment, the material property setting unit automatically calculates the fluid viscosity and density based on the user-input temperature and pressure and writes them into the solver's material library; in another embodiment, the solver's built-in material and property models can be used directly.

[0048] To improve the stability of batch operation, the simulation coupling module can perform an engineering consistency check before each iteration, including at least: whether the porous region name exists, whether the monitoring surface exists, and whether the solver is in an iterable state; and record error logs and execute retry or skip strategies when an anomaly occurs.

[0049] (v) Result Extraction and Feature Construction As attached Figure 5 As shown, the result extraction module performs automatic post-processing after the solution converges: 1) Velocity statistics: Calculate the volume average velocity modulus for each porous medium region and output the minimum, maximum and overall average values ​​of the region, which are weighted by the region's volume. 2) Pressure difference statistics: Calculate the area-weighted average and maximum static pressure on the outer and inner surfaces of the filter, and obtain the average pressure drop and maximum pressure drop from the difference; 3) Optional field characteristics: Further extraction of velocity non-uniformity indices, local backflow ratio, mass flow rate of key sections, and user-specified monitoring quantities can be performed.

[0050] The aforementioned simulation metrics, together with the candidate design points, constitute a data sample for evaluation and learning.

[0051] (vi) Evaluation and Constraints As attached Figure 3 As shown, the evaluation and optimization module maps simulation metrics to a comprehensive objective. To avoid disputes regarding the evaluation model, this invention designs the efficiency calculation as a pluggable module: in one embodiment, the efficiency model calculates the contributions of diffusion, interception, and inertial collisions based on the fiber filtration mechanism; in another embodiment, the efficiency model is directly given by experimental calibration curves, empirical formulas, or machine learning regression models.

[0052] Constraint handling can employ either hard or soft constraints: hard constraints directly classify design points that do not meet the threshold as infeasible; soft constraints incorporate the degree of violation into the overall objective through a penalty function.

[0053] (vii) Adaptive optimization and convergence criteria In a preferred embodiment, the evaluation and optimization module trains the surrogate model based on the current sample set and selects the next set of candidate points using a collection function. To ensure convergence and robustness, any of the following termination conditions can be used: 1) The overall improvement of the target after several consecutive iterations is less than a preset threshold; 2) Reach the maximum number of simulations; 3) Meet the time limit criteria; 4) The feasible region is empty or the solver continues to fail beyond the threshold.

[0054] Upon termination, a single optimal design point is output, along with a list of suboptimal solutions for engineering consideration.

[0055] (viii) Data Traceability and Result Export In one embodiment, the output and traceability module writes the input parameters, porous region resistance parameters, simulation indicators, comprehensive objectives, and solution logs for each candidate point into a structured result file and exports them in tabular form (preferably an Excel file). At the same time, the platform saves the engineering input and parameter space as a configuration file (preferably YAML) and generates a configuration snapshot after each iteration, thereby ensuring that the results are traceable and reproducible.

[0056] (ix) Embodiments of Devices and Computer Equipment As attached Figure 6 As shown, the device of the present invention can be deployed as standalone software or a client-server architecture. The client provides an append-only... Figure 2 To be continued Figure 5 The human-computer interaction interface shown is coupled with a simulation coupling and optimization engine running on the server side to make full use of multi-core CPU or computing cluster resources.

[0057] As attached Figure 6 As shown, the computer device includes a processor, a memory, and a network interface; the memory stores computer-readable instructions, and the processor executes these instructions to implement the aforementioned closed-loop optimization method.

[0058] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A multi-zone porous media filter parameter fast optimization closed-loop simulation method, characterized in that, Includes the following steps: S1. Obtain simulation engineering input: The user creates the filter geometry model and mesh, defines at least one fluid region and at least two porous media regions, and sets boundary conditions, turbulence / laminar flow models, discretization schemes, residual convergence criteria, and monitoring quantities. S2. Construct parameter space: Set the design variable range and step size or sampling distribution for each porous media region. The design variables should include at least fiber diameter and filling density. S3. Generate candidate design points: Generate a set of candidate design points from the parameter space according to the preset optimization strategy. The candidate design points include the combination of design variables for each porous medium region. S4. Parameter Mapping and Writing: Automatically map candidate design points to the resistance parameters of each porous medium region and write them into the CFD solver. The resistance parameters include at least the viscous drag coefficient. S5. Coupled Solution: Control the CFD solver to initialize and iteratively calculate the model after the resistance parameters are written in, and obtain the steady-state or quasi-steady-state flow field and pressure field. S6. Index Extraction: Perform automatic post-processing on the solution results to extract simulation indices, including at least representative velocity statistics of the porous media region and pressure difference statistics of key sections. S7. Evaluation and Update: Under constraints, calculate the comprehensive target value based on simulation indicators, and update the optimization strategy accordingly to generate the next set of candidate design points. S8. Termination and Output: When the convergence criterion, the upper limit of the number of simulations, or the time limit criterion is met, the closed-loop optimization is terminated, and the single optimal design point and its corresponding resistance parameters, simulation indicators, configuration files, and result data are output.

2. The closed-loop simulation method for rapid optimization of parameters of multi-region porous media filters according to claim 1, characterized in that, The optimization strategy in S3 is an adaptive sampling strategy, which includes: training a surrogate model based on the evaluated design points to approximate the mapping relationship between "design variables and comprehensive objectives". The surrogate model is any one of Gaussian process regression, random forest regression or neural network regression. The next set of candidate design points is selected under constraints based on the acquisition function, which is any one of expected improvement, upper confidence bound or probabilistic improvement.

3. The closed-loop simulation method for rapid optimization of parameters of multi-region porous media filters according to claim 2, characterized in that, In S4, resistance parameter inversion or calculation is performed for each porous media region: the equivalent permeability is calculated based on the fiber diameter and packing density of the region, and the reciprocal of the equivalent permeability is written into the CFD solver as the viscous resistance coefficient.

4. The closed-loop simulation method for rapid optimization of parameters of multi-region porous media filters according to claim 3, characterized in that, In S6, the volume average velocity modulus is extracted for each porous media region, and the overall average velocity of the filter is obtained by weighting the volume of the porous media region. At the same time, the minimum and maximum values ​​of the volume average velocity modulus of each porous media region are output. At least two monitoring surfaces are set on the outside and inside of the filter, and the area-weighted average static pressure and maximum static pressure of the monitoring surfaces are calculated respectively. The average pressure drop and maximum pressure drop are obtained from the difference.

5. The closed-loop simulation method for rapid optimization of parameters of a multi-region porous media filter according to claim 4, characterized in that, In S7, the overall objective is to minimize the pressure drop while satisfying either the filtration efficiency threshold or the penetration threshold, or to maximize the efficiency while satisfying the pressure drop threshold. The filtration efficiency is calculated by a user-pluggable efficiency model module based on representative velocity statistics and particle parameters.

6. The closed-loop simulation method for rapid optimization of parameters of a multi-region porous media filter according to claim 5, characterized in that, The closed-loop optimization from S3 to S7 adopts a hierarchical screening mechanism, which includes: first, using a low number of iterations or a simplified model to perform coarse screening to obtain a candidate subset, and then using a high number of iterations or a high-precision model to perform fine screening on the candidate subset to output a single optimal design point.

7. The closed-loop simulation method for rapid optimization of parameters of a multi-region porous media filter according to claim 6, characterized in that, The simulation engineering input in S1 and the parameter space constructed in S2 are saved in the form of a structured configuration file. The structured configuration file includes: porous medium region name, thickness, design variable range and step size, material name, temperature and pressure conditions, monitoring surface name, number of solution iteration steps and additional solver commands, and automatically records the configuration snapshot corresponding to the candidate design point after each closed-loop iteration.

8. A closed-loop simulation device for rapid optimization of parameters of a multi-region porous media filter, applied to the closed-loop simulation method for rapid optimization of parameters of a multi-region porous media filter as described in any one of claims 1-7, characterized in that, It includes: a module for acquiring simulation engineering input, a module for constructing a parameter space, a module for generating candidate design points, a module for parameter mapping and writing, a coupled solution module, a module for index extraction, a module for evaluation and updating, and a module for termination and output. The simulation engineering input module is used by the user to create the filter geometry model and mesh, define at least one fluid region and at least two porous media regions, and set boundary conditions, turbulence / laminar flow models, discretization schemes, residual convergence criteria and monitoring quantities; Construct a parameter space module to set the design variable range and step size or sampling distribution for each porous media region. The design variables include at least fiber diameter and packing density. The candidate design point generation module is used to generate a set of candidate design points from the parameter space according to a preset optimization strategy. The candidate design points include the combination of design variables for each porous medium region. The parameter mapping and writing module is used to automatically map candidate design points to the resistance parameters of each porous medium region and write them into the CFD solver, wherein the resistance parameters include at least the viscous drag coefficient. The coupled solution module is used to control the CFD solver to initialize and iteratively calculate the model after the resistance parameters are written in, so as to obtain the steady-state or quasi-steady-state flow field and pressure field. The index extraction module is used to automatically post-process the solution results and extract simulation indices, including at least representative velocity statistics of porous media regions and pressure difference statistics of key sections. The evaluation and update module is used to calculate the comprehensive target value based on simulation indicators under constraints, and update the optimization strategy accordingly to generate the next set of candidate design points. The Termination and Output module is used to terminate the closed-loop optimization when the convergence criterion, the upper limit of the number of simulations, or the time limit criterion is met, and outputs a single optimal design point and its corresponding resistance parameters, simulation indicators, configuration files, and result data.

9. A storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the closed-loop simulation method for rapid optimization of parameters of multi-region porous media filters as described in any one of claims 1-7.

10. A computer device, characterized in that, include: The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the closed-loop simulation method for rapid optimization of parameters of a multi-region porous media filter as described in any one of claims 1-7.

Images