Multi-fidelity co-simulation management method and system for pneumatic transportation pipe network system

Abstract

The invention discloses a multi-fidelity co-simulation management method and system for a pneumatic transportation pipe network system, and belongs to the technical field of computer simulation. The method comprises the steps of topology analysis and feature extraction, and analysis of a topology structure; simulation dynamic allocation: dynamically allocating a simulation model for each pipe section based on a preset strategy rule base; the simulation tasks are packaged and dispatched, and the continuous pipe sections distributed with the same simulation models are distributed; performing data coupling and consistency verification, and taking a calculation result of the upstream pipe section as an input boundary condition of calculation of the downstream pipe section; global results are synthesized, and results of all parallel computing tasks are automatically collected. According to the method, a one-dimensional theoretical model or a differentiated three-dimensional CFD model is accurately matched for different flow regime areas (a straight pipe flow field development area, a bent pipe separation area and the like) based on pipe section topological characteristics through dynamic model adaptation, the efficiency of a non-key section is guaranteed through the one-dimensional model, the flow regime of a key section is restored through a high-precision turbulence model, and cooperation of precision and efficiency is achieved.

Description

Technical Field

[0001] This invention relates to the field of computer simulation technology, and specifically to a multi-fidelity collaborative simulation management method and system for pneumatic transportation pipeline systems. Background Technology

[0002] As a highly efficient method for conveying particulate materials, pneumatic conveying systems typically involve a large number of straight pipe sections and complex curved pipe sections in their pipeline design. Computer simulation technology plays a crucial role in the design and optimization of such systems, used to predict material transport patterns, assess pressure drop losses, and identify potential blockage risks.

[0003] Currently, the following are the mainstream approaches to simulation analysis of pneumatic transport pipeline networks:

[0004] One approach is to use a high-fidelity computational fluid dynamics (CFD) model to perform a unified three-dimensional numerical simulation of the entire pipeline network. Although this method can obtain detailed data on the flow field and particle phase, it consumes huge computational resources and takes too long to solve, making it difficult to meet the needs of rapid iteration and scheme optimization in engineering design.

[0005] Secondly, simplified theoretical formulas or one-dimensional models are used for system-level calculations. This method is computationally efficient, but it is difficult to accurately capture key physical phenomena such as fluid separation, secondary flow, and particle collisions in complex local structures such as bends and variable diameter pipes. This results in a significant decrease in prediction accuracy at critical parts, which cannot provide effective support for reliability design.

[0006] To balance computational efficiency and local accuracy, existing technologies have proposed a segmented processing approach, employing models of different complexities for straight and curved pipe sections. However, this approach still has significant limitations in practical applications. First, its computational strategy is static and pre-defined, typically allocating models based solely on simple pipe type (straight / curved), lacking refined perception and adaptive decision-making capabilities regarding pipe segment geometry (such as length and curvature), thus failing to achieve intelligent allocation of computational resources. Second, the computational processes between different models are isolated, and data transfer often relies on manual intervention, making it difficult to achieve fully automated execution and closed-loop management. This is especially problematic for large pipe networks with complex topologies, where operations are cumbersome and prone to errors. Existing solutions lack system-level collaborative management and fault tolerance mechanisms, failing to consider the compatibility and consistency verification of data interfaces when multiple models are coupled, making it difficult to guarantee the robustness of the simulation process and the reliability of the results.

[0007] Based on this, the present invention designs a multi-fidelity collaborative simulation management method and system for pneumatic transportation pipeline systems to solve the above problems. Summary of the Invention

[0008] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a multi-fidelity collaborative simulation management method and system for pneumatic transportation pipeline network systems.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] A multi-fidelity collaborative simulation management method for pneumatic transportation pipeline systems includes the following steps:

[0011] S1: Topology analysis and feature extraction: Automatically identify the input 3D model of the pipeline network, analyze its topology, and extract the feature parameters of each pipe segment. The feature parameters include at least the pipe segment type, geometric dimensions, and radius of curvature.

[0012] S2: Simulation dynamic allocation, based on a preset strategy rule base, dynamically allocates a simulation model to each pipe segment. The allocation rules of the strategy rule base are based on the characteristic parameters, fluid property parameters and user-preset system-level objectives of the pipe segment.

[0013] The pre-built policy rule base is constructed as follows:

[0014] Based on the fundamental theories of fluid mechanics, and combined with the development law of the flow field in straight pipes and the flow separation characteristics in bends in pneumatic transportation networks, an initial model allocation rule framework is determined.

[0015] Based on common simulation requirements of pneumatic transport systems (such as efficiency priority and accuracy priority), the judgment dimensions in the initial rule framework are refined to ensure that the rules can adapt to different simulation objectives.

[0016] The rule base supports dynamic updates and can optimize the judgment logic in the rules based on the degree of fit between the model calculation results and physical laws during the actual simulation process. This eliminates the need for repeated trial and error based on human experience and achieves autonomous adaptability of the rules.

[0017] S3: Simulation task encapsulation and scheduling, which dynamically combines continuous pipe segments assigned the same simulation model into a calculation task package, and the task scheduler calls the corresponding solvers in parallel to perform calculations;

[0018] S4: Data Coupling and Consistency Verification: The calculation results of the upstream pipe segment are used as the input boundary conditions for the calculation of the downstream pipe segment, and data consistency verification is performed.

[0019] S5: Global result synthesis, automatically collects the results of all parallel computing tasks, splices and synthesizes them according to the pipeline topology, and generates a system-level simulation report.

[0020] Furthermore, in S2, the specific allocation rule of the strategy rule base is as follows: if the pipe segment type is a straight pipe and its length-to-diameter ratio is greater than or equal to a preset threshold, then a one-dimensional theoretical model is allocated to it.

[0021] If the pipe segment type is a bend or a straight pipe with a length-to-diameter ratio less than the preset threshold, then a three-dimensional computational fluid dynamics model is assigned to it;

[0022] The preset threshold is dynamically optimized based on fluid properties and system-level objectives.

[0023] Furthermore, the allocation rules of the policy rule base further include:

[0024] A high-precision turbulence model is assigned to bends with a radius of curvature to diameter ratio less than a critical value;

[0025] The critical value is dynamically adjusted based on the fluid properties and the system-level target.

[0026] Furthermore, the dynamic combination into a computational task package described in S3 specifically includes identifying branch points in the topology and combining continuous pipe segments between branch points that have been assigned the same simulation model into independent task packages.

[0027] Furthermore, the data consistency verification described in S4 specifically includes:

[0028] Monitor the mass flow rate and pressure data at the interface of adjacent pipe sections. If the difference exceeds the allowable tolerance, trigger the adaptive adjustment mechanism.

[0029] The adaptive adjustment mechanism includes recalculating the boundary conditions of the upstream pipe segment or adjusting the boundary conditions of the downstream pipe segment.

[0030] Furthermore, the adaptive adjustment mechanism prioritizes a local fine-tuning strategy, recalculating and coupling only adjacent pipe segments with inconsistent data until the data meets the consistency requirements.

[0031] Furthermore, the task scheduler described in S3 dynamically allocates the number of parallel computing tasks based on the number of available computing cores and memory resources of the current computer system, thereby achieving load balancing.

[0032] Furthermore, a preprocessing step is included before starting S1:

[0033] Receive system-level target parameters input by the user, and convert the system-level target parameters into allocation rule thresholds in the policy rule base;

[0034] The system-level target parameters are the maximum allowable computation time or the expected simulation accuracy level.

[0035] Furthermore, the preprocessing step optimizes the preset threshold and the critical value based on the system-level target parameters.

[0036] A system for implementing any of the above methods, characterized in that it comprises:

[0037] The topology parsing module is configured to parse the 3D model of the pipeline network, identify its topology, and extract the feature parameters of each pipe segment.

[0038] The strategy rule base stores rules for dynamically allocating simulation models based on pipe segment characteristic parameters, fluid property parameters, and user-preset system-level objectives;

[0039] The task scheduler is configured to dynamically combine consecutive pipe segments assigned the same simulation model into computational task packages and perform parallel scheduling computations.

[0040] The data coupling and verification module is configured to manage data transfer between management segments and performs data consistency verification based on the law of conservation of mass.

[0041] The results synthesis module is configured to integrate the results of all parallel computing tasks and splice and synthesize them according to the pipeline topology to generate a system-level simulation report.

[0042] Compared with the prior art, the beneficial effects of this invention are as follows:

[0043] 1. This invention adapts a dynamic model to accurately match a one-dimensional theoretical model or a differentiated three-dimensional CFD model for different flow regime regions (straight pipe flow field development zone, bend separation zone, etc.) based on the topological characteristics of pipe segments (such as the length-to-diameter ratio of straight pipes and the curvature-to-diameter ratio of bends). The one-dimensional model ensures the efficiency of non-critical sections, while the high-precision turbulence model restores the flow regime of critical sections, achieving a synergy between accuracy and efficiency.

[0044] 2. This invention achieves continuous self-consistency of the flow field at the junction of multiple models by using self-consistent coupling and relying on topology to automatically transfer boundary conditions. It also verifies the flow deviation at the junction with mass conservation and adaptively fine-tunes and iterates when deviations exceed the limit, thereby improving the reliability of global simulation.

[0045] 3. This invention automates the entire process, from topology parsing and task encapsulation to parallel scheduling and result synthesis. It automatically parses the pipeline network based on the geometric kernel, intelligently merges tasks and dynamically allocates resources, greatly reducing manual intervention and compressing the iteration cycle. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0047] Figure 1 This is a schematic diagram of the overall process of a multi-fidelity collaborative simulation management method for pneumatic transportation pipeline systems according to the present invention;

[0048] Figure 2 This is a schematic diagram of the dynamic allocation strategy for the simulation model of the present invention; Figure 3 This is a graph showing the relationship between the cumulative pipeline distance and pressure and velocity according to the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0050] Example 1: This example converts the input 3D geometric model without clear engineering semantics into a structured data model containing complete pipe segment connection relationships and key feature parameters. Specifically, the user selects and imports a 3D model file of the pneumatic transportation pipeline network generated by computer-aided design (e.g., CAD) software through a graphical user interface (GUI) or application programming interface (API). The 3D model file is in a standardized neutral format, preferably STEP or IGES format. These two formats can completely preserve the boundary representation data of the model and eliminate compatibility issues caused by using the native format of specific software.

[0051] After successfully loading the model, the automated recognition and classification process begins. This process relies on the geometric kernel library (such as Open CASCADE Technology) integrated behind the method. The specific processing logic is as follows:

[0052] The model analyzes the entity structure to distinguish between pipes and non-pipe entities, classifies pipe segment types (straight pipe segments, bends, fittings) according to spatial morphology, extracts the centerline of each pipe segment, calculates the geometric feature parameters of the pipe segments (length, inner diameter, bending angle, radius of curvature, etc.), and constructs the connection relationship matrix between pipe segments to form a complete topology.

[0053] All steps are executed automatically through the kernel library standard interface without manual intervention, ensuring the accuracy of parsing and extraction. For each entity in the model, an algorithm based on the minimum distance field or skeleton extraction algorithm is used to calculate and extract its centerline, which is composed of a series of continuous spatial points, representing the central path of the pipeline.

[0054] For the extracted centerline, a numerical differentiation method (such as the three-point method based on central difference) is used to calculate the curvature value of each discrete point on the centerline. Then, the absolute value of curvature along the entire centerline is integrated to obtain the total curvature integral value. Its physical meaning represents the total degree of curvature of the central path of that section of the pipeline;

[0055] The total curvature integral value of each entity With a preset discrimination threshold Comparison, threshold The setting is based on a simplified engineering criterion for the critical condition of flow separation in fluid mechanics. The principle is that when the pipe's curvature is sufficiently slight, such that the resulting centrifugal force is insufficient to cause significant flow separation between the main flow and the pipe wall, or to generate a strong secondary flow, the flow characteristics of that pipe section can be treated as a straight pipe within the engineering accuracy range. According to Schlichting's boundary layer theory, for a smooth-walled circular pipe, when the following conditions are met... At this time, the additional pressure drop loss caused by bending is negligible compared with the frictional resistance of straight pipe (usually <2%), and will not form a secondary vortex structure that is sufficient to affect the particle transport law;

[0056] Based on this classification rule, it is quantitatively determined that when If so, it is classified as a straight pipe section. If so, it is classified as a bend in the pipe segment. At the same time, special structures (such as tees, crosses, and reducers) that connect three or more pipe segments in the model are identified and uniformly classified as pipe fittings.

[0057] After classifying all entities, their connections need to be reconstructed to form a topological map describing the material transport path (this is done by calculating the spatial geometric relationships of the boundary surfaces of adjacent entities). This involves traversing all pipe segments and fittings. For any two entities, the spatial relationship between their potentially adjacent end faces is calculated. Specific criteria include calculating the ratio of the overlapping area of ​​the two end faces in three-dimensional space to their individual areas. If this ratio is greater than 95%, the two end faces are considered to be substantially coincident in space. The normal vectors of the two coincident end faces are then calculated. If the cosine of the angle between the two normal vectors is... If the included angle is greater than arccos(-0.95)≈162°, then the two end faces are considered to be face-to-face connected. Only when both conditions are met simultaneously, such as the end face overlap > 95% and the cosine value of the included angle of the normal vector < -0.95, is it determined that there is a connection between the two entities. Based on all the determined connection relationships, a directed graph data structure is constructed in memory. It is worth mentioning that in practical applications, the tolerance thresholds of the above judgment criteria (such as 95%, -0.95) can be adaptively adjusted according to the model source and geometric quality, or a fault tolerance mechanism can be introduced to handle non-ideal geometric cases.

[0058] In this diagram, the endpoints of pipe fittings and pipe segments are used as nodes, and the pipe segments themselves are used as edges connecting the nodes. This directed graph fully defines the structure of the pipeline network and the direction of material flow.

[0059] Traverse every edge in the topology graph (i.e., every pipe segment object), accurately calculate and store its geometric feature parameters based on its classification results. For straight pipe segments, calculate its length L, which is obtained by the Euclidean distance between the two endpoints of its centerline (in meters). For curved pipe segments, calculate its bending angle. (i.e., the angle between the starting and ending direction vectors of the centerline, in °) and the radius of curvature R (obtained by fitting the centerline point set into a spatial circular arc, in m). For all pipe segments, the inner diameter D is extracted and obtained by calculating the radial dimension of the internal cavity of the pipe segment (in m).

[0060] Its output is a structured data object, which is no longer a cold geometric model, but an engineering data model that contains a list of all pipe segments, detailed characteristic parameters of each pipe segment, and a complete matrix of connections between them.

[0061] Example 2: This example, based on the pipe segment characteristic parameters of Example 1, dynamically assigns the optimal simulation model to each pipe segment using quantified decision rules. Its core objective is to automatically select a mathematical model that achieves the optimal balance between computational efficiency and simulation accuracy for each pipe segment, based on fluid mechanics principles. This process is implemented by applying a predefined policy rule base containing hierarchical judgment logic. The decisions in this rule base are entirely based on the geometric characteristic parameters of the pipe segment. The complete judgment logic is as follows:

[0062] First, distinguish whether the pipe section is a straight pipe section or a bend;

[0063] The degree of flow field development is determined by combining the length-to-diameter ratio of the straight pipe section with the flow parameters, and then matched with a one-dimensional or three-dimensional model.

[0064] The curvature-to-diameter ratio of the bend section is used to determine the flow separation and secondary flow intensity, and a three-dimensional computational fluid dynamics model is matched with a conventional or high-precision turbulence model.

[0065] For pipe segments classified as straight pipe segments, the main criterion for model selection is whether the flow field has reached a fully developed state. This state depends on the pipe's length-to-diameter ratio (L / D, i.e., the ratio of the pipe segment length L to the inner diameter D) and the Reynolds number of the flow. According to the theory of fully developed pipe flow, the inlet section length required for the flow field to fully develop under laminar flow conditions is... It is proportional to the Reynolds number Re ( ), while in turbulent flow It is proportional to the power of 1 / 6 of Re ( );

[0066] To ensure reliable assessment of flow field development across all possible flow regimes (from laminar to highly turbulent), a conservative and universally applicable threshold must be employed. This embodiment sets a length-to-diameter ratio threshold. The determination of this value is based on the fact that even under extreme turbulent conditions with high Reynolds numbers, Its inlet section length is also less than 50 times the pipe diameter, therefore it adopts... This threshold ensures that when the straight pipe length L ≥ 50D, under the common working flow conditions (usually turbulent) in pneumatic transport systems, the flow field inside the pipe can be considered fully developed and the flow structure stable. Based on this, the rules are as follows:

[0067] If the length-to-diameter ratio L / D of the straight pipe section is greater than or equal to 50, a highly efficient theoretical model is assigned to it. This model is based on the force balance equation and establishes a set of ordinary differential equations, which can complete the calculation in seconds and is used to solve the average gas velocity and pressure at the pipe section outlet.

[0068] If the length-to-diameter ratio (L / D) of a straight pipe section is less than 50, the pipe section is considered to be in the inlet effect zone or transition zone. The flow field is significantly affected by upstream and downstream local flow disturbances, exhibiting complex velocity distribution evolution. Therefore, a high-fidelity three-dimensional computational fluid dynamics (CFD) model must be used. In this allocation, the Standard model, which is robust in engineering applications and has moderate computational resource consumption, is selected by default. Turbulence model.

[0069] For pipe segments classified as bends, the core criterion for model selection is the severity of flow separation and the intensity of secondary flow, which is determined by the ratio of the bend's radius of curvature to its diameter (R / D). R / D is the Dean number (used to describe the inertial effects in bend flow). A key component of ), directly controlling the intensity of the secondary flow, in which It is the ratio of the pipe diameter to twice the radius of curvature. The square root of this ratio is used to quantify the geometric effects of the bend and is directly related to the secondary flow intensity.

[0070] If the R / D ratio of the bend section is less than 2.0, it is classified as a small radius bend (sharp bend). In such bends, the fluid experiences strong directional changes and centrifugal forces, making it highly susceptible to flow separation on the inner wall, generating strong vortices and backflow zones, accompanied by significant turbulent kinetic energy generation and dissipation. In this case, the model will significantly overestimate the turbulent viscosity, leading to inaccurate predictions of the separation region. Therefore, the following rule applies:

[0071] If the R / D ratio of the bend section is greater than or equal to 2.0, then a three-dimensional CFD model (Standard) should be assigned to it. );

[0072] If the R / D ratio of the bend is less than 2.0, then assign a three-dimensional CFD model (SST) to it. ), SST The model incorporates Wilcox, which performs well in the near-wall region. Model and The model, combined with consideration of turbulent shear stress transmission, achieves a significantly higher accuracy in predicting flow separation under adverse pressure gradients than the Standard model. The model is the preferred model for simulating flow in sharp bends. The above rules are applied sequentially for evaluation, and each pipe segment is ultimately assigned a unique simulation ID. To ensure the computational accuracy of the CFD simulation model, the following requirements apply to its computational mesh:

[0073] For the turbulence model used, the height of the first mesh layer near the wall must meet the requirement of dimensionless walldistance y+. For Standard The model must use standard wall functions, with a requirement of 30 < y + < 300, for SST. The model requires y+≈1 to analyze the viscous sublayer. For the bend section, the inner boundary layer mesh needs to be additionally refined. The number of meshes in the refined region should account for ≥20% of the total number of meshes in the bend to accurately capture the flow separation phenomenon inside the bend.

[0074] In the preprocessing, the mesh independence of typical pipe sections (such as the thinnest pipe and the bend with the greatest curvature) is verified. That is, three sets of meshes (coarse, medium and fine) are used for simulation in sequence. When the rate of change of key parameters (such as pressure drop) under medium and fine meshes is less than 2%, the density of medium mesh is set as the standard and extended to the mesh generation settings of all similar pipe sections.

[0075] The key thresholds in the strategy rule base (such as the straight pipe length-to-diameter ratio threshold and the bend curvature-to-diameter ratio critical value) are not fixed, but can be adaptively adjusted or user-defined according to fluid physical parameters and system-level objectives to ensure their universality under different working conditions.

[0076] For straight pipe sections, the inlet length required for sufficient flow field development is related to the Reynolds number (Re). Therefore, the straight pipe length-to-diameter ratio threshold can be expressed as a function of the Reynolds number, for example... ;

[0077] In a preferred implementation, a conservative fixed value (such as 50) can be used to cover all cases from laminar to turbulent flow;

[0078] In another preferred implementation, for specific applications with a known Reynolds number range, a more precise threshold can be set; for example, for high Reynolds number turbulent conditions, it can be set to... For low Reynolds number laminar flow conditions, it can be set to... ;

[0079] Among them, Reynolds number The definition of , For conveying gas density, For the average gas flow rate, For the inner diameter of the pipe section, This represents the gas dynamic viscosity.

[0080] The system provides a graphical user interface (GUI) or configuration file interface, allowing users to fine-tune the above thresholds based on existing experimental data or high-fidelity simulation results. For example, users can input multiple sets of (operating conditions, optimal threshold) data pairs, and the system will construct the mapping relationship between the threshold and the operating condition parameters through linear interpolation or fitting, and update the strategy rule base.

[0081] As mentioned above, the expected simulation accuracy level or maximum allowable computation time parameter received during preprocessing will be directly used to optimize these thresholds. The high-precision mode will reduce the straight pipe length-to-diameter ratio threshold and the bend curvature-to-diameter ratio critical value, allowing more pipe sections to use high-precision CFD models. The high-efficiency mode will increase these thresholds, allowing more pipe sections to use one-dimensional models.

[0082] As mentioned earlier, the expected simulation accuracy level or maximum allowable computation time parameter received during preprocessing will be directly used to optimize these thresholds. High-precision mode will lower the straight pipe length-to-diameter ratio threshold and the bend curvature-to-diameter ratio critical value, allowing more pipe sections to use high-precision CFD models. High-efficiency mode will raise these thresholds, allowing more pipe sections to use one-dimensional models. For pneumatic transport systems containing particulate phases, the one-dimensional theoretical model needs to be supplemented with particle acceleration calculation equations, specifically:

[0083]

[0084] in For the average velocity of particles, For particle density, For particle size, For particle drag coefficient, For the average gas velocity, For gas density, the three-dimensional CFD model (three-dimensional computational fluid dynamics model) supports the coupling of discrete phase model (DPM) with the flow field to simulate particle motion trajectory and collision behavior.

[0085] It should be noted that for entities classified as pipe fittings, due to their extremely complex flow structures, a high-precision three-dimensional CFD model (SST) is assigned by default. ) to perform calculations.

[0086] Example 3: The input of this example is the list of pipe segments processed in Example 2, where each pipe segment has been assigned a unique simulation ID. By optimizing and reorganizing the computing tasks, the process management overhead is minimized, and the multi-core parallel computing capabilities of modern processors are fully utilized, thereby significantly improving the efficiency of large-scale pipe network simulation analysis.

[0087] The purpose of task encapsulation is to aggregate a large number of fine-grained pipe segment calculation requests into fewer but more balanced coarse-grained calculation task packages, thereby significantly reducing the number of independent processes that need to be scheduled and executed subsequently. The encapsulation process is based on two key pieces of information: the pipe network topology diagram (Example 1) and the simulation ID (Example 2). The specific algorithm is as follows:

[0088] Starting from the entrance of the pipeline system, traverse along the directed edges of the topology graph (e.g., using a depth-first search algorithm).

[0089] During the traversal, the currently processed pipe segment is compared with the next adjacent downstream pipe segment. If both of the following conditions are met, the two pipe segments are merged into the same task package:

[0090] Condition 1 (strategy consistency): The two pipe segments are assigned the exact same simulation ID;

[0091] Condition 2 (topological continuity): The two pipe segments are directly connected in the topological structure, and the connection point is not a branch point (i.e., the out-degree of the connection point is equal to 1).

[0092] Once a branch point (such as a tee or cross, i.e., the out-degree of the connection point is greater than 1) or a pipe segment whose simulation ID changes is encountered, the merging process of the current task package is immediately terminated, and the merged pipe segment sequence is packaged into an independent task package. The pipe segment with the branch point or strategy change will be used as the starting point of the next task package. The above process is repeated until all possible transport paths have been traversed.

[0093] This algorithm intelligently reorganizes the original hundreds or thousands of independent pipe segments into dozens of computational task packages. Each task package contains a continuous pipe with the same physical model, which can be submitted to the solver as an independent computational domain.

[0094] Once the task is packaged, it enters the parallel scheduling and execution phase. The purpose of this phase is to efficiently allocate the packaged computing task packages to available computing cores. The scheduler first automatically detects the hardware configuration of the current computer system, obtains the total number of available CPU logical cores, and estimates the memory usage of each computing task package.

[0095] The scheduler uses a priority-based task queue scheduling algorithm:

[0096] The calculation priority is assigned to the pipe segment corresponding to the task package based on its criticality in the flow regime;

[0097] Arrange all task packages in descending order of priority and place them in a queue;

[0098] Create a thread pool containing (total number of available CPU logical cores - 1) worker threads, and dynamically monitor the status of each thread and the remaining system memory;

[0099] Task packages are assigned to idle threads in order of priority, and the total memory consumption of the assigned task packages is ensured not to exceed the system safety threshold, thereby achieving efficient load balancing and resource utilization.

[0100] The worker thread dynamically calls the corresponding solver executable file based on the simulation ID of the acquired task package to ensure the stability of the system operation and avoid lag or crash due to the complete occupancy of computing resources.

[0101] The scheduler places all generated computational task packages into a first-in, first-out (FIFO) task queue. Idle worker threads in the thread pool retrieve a task package sequentially from the queue. Based on the simulation ID of the retrieved task package, the worker thread dynamically calls the corresponding solver executable file. For example:

[0102] For task packages identified as one-dimensional theoretical models, invoke the pre-compiled fast solver;

[0103] For models identified as 3D CFD models (Standard) The task package calls the CFD solver and runs in batch mode;

[0104] For models identified as 3D CFD models (SST) The task package calls the CFD solver and runs in batch mode.

[0105] The worker thread monitors the solver's execution process and outputs the result to a specified location after the calculation is complete. Then, the thread becomes idle, ready to receive the next task package. Through the above, a maximum of ( The computational task packages are solved simultaneously on different CPU cores. This coarse-grained task-level parallelism greatly shortens the overall computation time, and its speedup is close to ideal linear. The above output is a set of result files generated after all computational task packages have been solved.

[0106] The task scheduler not only manages CPU core allocation but also has advanced resource management capabilities:

[0107] Before scheduling, the task scheduler estimates the memory usage of each CFD task package based on the number of grids (read from the grid file header information) and the solver type. The specific estimation formula is: single task memory usage = number of grids × single grid memory coefficient (1.2KB / grid for 3D CFD models and 0.1KB / grid for 1D theoretical models). The single grid memory coefficient is an empirical value pre-calibrated based on the selected solver and physical model, which can be calibrated through benchmark testing during system deployment.

[0108] The scheduler ensures that the total estimated memory consumption of all parallel task packages does not exceed 90% of the system's physical memory to avoid computational crashes caused by memory overflow. If the total memory requirement exceeds the limit, the scheduler will automatically reduce the number of task packages that are computed in parallel at the same time and use a queue mechanism to schedule them in order.

[0109] In the policy rule base, specific CFD models (such as SST) can be specified. Once a task is marked as recommended for GPU acceleration, the task scheduler will allocate the corresponding task package to a compute node or process configured with a GPU solver after detecting available GPU resources, and will pass the corresponding GPU device parameters in the call command.

[0110] Example 4: The input of this example is all the task package result files output after parallel computing in Example 3. Its core purpose is to solve the core challenges in multi-model collaborative simulation, and to ensure that the physical quantities at the task package interface strictly satisfy the conservation law, thereby ensuring that the results of distributed computing can be spliced ​​into a physically real global flow field.

[0111] Data transfer is a prerequisite for coupling. The process is automatically executed based entirely on the topological relationship reconstructed in Example 1. According to the topological directed graph, the upstream dependencies of each task package are automatically determined. For a downstream task package B, all its direct upstream task packages (e.g., task package A) are located.

[0112] After task package A completes its calculation, its output result file (such as the monitoring surface data file of CFD calculation or the result file of a one-dimensional model) is automatically parsed. The result file is output in a predefined standardized data interface format, which contains the necessary physical quantity fields and their units to facilitate automatic parsing.

[0113] The core elements of a standardized data interface format include:

[0114] Pipe segment association information is used to identify the pipe segment to which the data belongs and the upstream and downstream positional relationship of the pipe segment in the topology;

[0115] Core information about physical quantities, including the name of the physical quantity, its corresponding numerical expression, and its unit, to ensure that upstream and downstream pipe sections have a consistent understanding of the physical quantities;

[0116] Data validity information, including the timestamp of data generation and the corresponding calculation step size, is used to determine whether the data is suitable for the boundary conditions of the next level of calculation; all content is organized in a structured manner to ensure that different simulation model solvers can automatically recognize and read it;

[0117] From the results file, the key flow field physical quantities at the exit interface of task package A are precisely extracted. These physical quantities include at least:

[0118] Mean static pressure of gas ( );

[0119] Gas average velocity vector ( );

[0120] The average velocity of the granular phase ( );

[0121] Gas density ( ).

[0122] The extracted physical quantities are automatically written into the inlet boundary condition definition file of task package B according to the format required by the downstream solver. For example, for the CFD task package, it is set as the pressure inlet or velocity inlet boundary condition, and for the one-dimensional model task package, it is set as the inlet initial condition. This process ensures the seamless transfer of physical quantities between computational domains.

[0123] After all task packages have been calculated, a verification is performed. The verification is based on the most fundamental law of mass conservation in fluid mechanics: for any interface connecting upstream task package A and downstream task package B, the mass flow rate into that interface must be equal to the mass flow rate out of that interface. Two independent calculations are performed at the interface:

[0124] Calculate the mass flow rate from the result of upstream task package A. ,in, For the fluid density at upstream task package A, The gas phase velocity vector at upstream mission packet A, The unit normal vector of the interface. This is an area integral performed on the area of ​​the interface;

[0125] Calculate the mass flow rate implied by the inlet boundary conditions from downstream task package B. For CFD models, the values ​​are directly calculated from the inlet boundary conditions; for one-dimensional models... This refers to the inlet mass flow rate; the relative deviation between the two is calculated. ,in, The relative deviation is used to quantify the degree of non-conservation of mass flow rate at the interface. The absolute deviation of upstream and downstream mass flow rates, Upstream mass flow rate as normalized baseline;

[0126] The relative deviation By quantifying the degree of quality conservation at the interface, it is possible to rigorously verify whether the mass flow between each task package satisfies the law of physical conservation, thereby ensuring the physical authenticity of the calculation results.

[0127] Set a maximum permissible relative deviation tolerance The tolerance value is set to 0.01 (i.e., 1%). This value is set based on the fact that in most engineering applications, the impact of a mass flow rate deviation within 1% on the overall system performance evaluation is negligible. Therefore, this value is selected, and the verification logic is as follows:

[0128] like If the data satisfies the mass conservation law at the interface, the verification is passed, and the result is reliable.

[0129] like If the verification fails, the adaptive adjustment mechanism is immediately triggered. The design principle of this mechanism is to achieve convergence with minimal computational cost, and its execution flow is as follows:

[0130] Local iterative initialization keeps all calculation results of upstream task package A completely unchanged, and only adjusts downstream task package B;

[0131] Boundary condition fine-tuning based on mass flow rate deviation The size and orientation are adjusted proportionally to fine-tune the entry boundary conditions of task package B, for example, if... Then according to a certain coefficient (such as Increase the ingress speed or pressure setting for task package B;

[0132] Recalculate, using the fine-tuned boundary conditions, only recalculate downstream task package B;

[0133] Recalibrate and calculate the new mass flow rate deviation;

[0134] The process involves iterative checks, repeated boundary condition adjustments, recalculation, and re-verification until a new mass flow rate deviation is identified. For problems with smooth convergence, 1-2 iterations are sufficient to meet the requirements.

[0135] If the iterations exceed the maximum number (e.g., 5 times) and still do not converge, the calculation is terminated and an error is reported, prompting the user to check the model settings or mesh quality, thereby avoiding meaningless consumption of computational resources, and outputting the final results of all physically reliable task packages that have passed the consistency check.

[0136] In addition to mass flow rate, the data consistency verification mechanism also monitors key interface physical quantities. For example, its tolerance standard is set as follows (the tolerance standard can be customized according to the user's requirements for the accuracy of the simulation results):

[0137] Mass flow tolerance ;

[0138] Total pressure tolerance (total pressure = static pressure + dynamic pressure), relative deviation of total pressure on both sides of the interface. ;

[0139] Flow direction tolerance, for CFD models, refers to the deviation of the airflow direction angle at the interface. ;

[0140] The fine-tuning coefficients in the adaptive adjustment employ an algorithm based on the proportional-integral (PI) control concept to achieve stable and rapid convergence, specifically as follows:

[0141] Let the first The mass flow deviation of the next iteration is The adjustment amount for the inlet boundary conditions of downstream task package B. for:

[0142]

[0143] in, It is the first The amount of adjustment to the boundary conditions in each iteration. It is the first The mass flow rate deviation of the second time. and These are the proportional and integral coefficients; the engineering feasibility range is... , It is usually set according to experience. and 1. The object to be adjusted can be the speed value at the velocity inlet or the pressure value at the pressure inlet;

[0144] If the deviation occurs after three consecutive iterations If the value is not reduced, it may trigger a more conservative adjustment strategy, such as... Halve the amount to prevent system instability.

[0145] Example 5: The input of this example is the final result data of all task packages that have passed the consistency verification of Example 4. The local calculation results stored in multiple independent files and divided by task package are seamlessly spliced ​​and fused according to their real physical spatial location to reconstruct a continuous physical field covering the entire pipeline network system and automatically generate a comprehensive analysis report containing multi-dimensional visualization and key performance indicators.

[0146] Read the global coordinate system of the original 3D CAD model of the pipeline network and use it as the spatial reference for all data mapping.

[0147] The result files of each task package are read sequentially. For each data point in the file (which is the center of the grid cell or a node in the CFD result, and the pipe section in the one-dimensional model result), the precise coordinates (X,Y,Z) of the data point in the three-dimensional space of the overall pipe network are calculated according to the global coordinate transformation relationship of the pipe segment corresponding to the task package in the topology diagram established in Example 1.

[0148] The coordinates of all data points and their physical quantities (pressure, velocity vector, etc.) are injected into a unified global data field structure. For the non-overlapping problem that may exist in the junction area of ​​data fields from different task packages, a grid-type-based interpolation algorithm (such as trilinear interpolation for hexahedral grids and shape function-based interpolation for tetrahedral grids) is used for smoothing to ensure the spatial continuity of physical quantities in the entire computational domain and eliminate artificial data steps caused by model switching. Finally, a unified data field containing the continuous distribution of physical quantities at every spatial point from the entrance to the exit is generated in memory.

[0149] Based on the reconstructed global data field, a visualization engine is invoked to generate charts that visually represent the flow characteristics, such as:

[0150] Generate static and dynamic visualization results for the entire pipeline network system, including:

[0151] Pressure cloud map, which displays the static pressure distribution on the pipe wall in a color gradient, intuitively identifies high-pressure and low-pressure areas;

[0152] Velocity cloud map shows the distribution of airflow velocity along the duct wall.

[0153] Velocity vector animation displays the velocity vector and direction of the airflow on the central cross-section of the pipe, dynamically demonstrating the airflow path, vortices, and separation zones.

[0154] The critical path plot creates a sampling line from the inlet to the outlet along the main material transport path. It automatically plots the total pressure (dynamic pressure + static pressure) change curve along this path, with the X-axis representing the cumulative distance from the inlet and the Y-axis representing the total pressure value. This curve directly reflects the overall pressure loss of the pipeline system.

[0155] Based on visualization, the data is analyzed, quantitative indicators are extracted, and structured reports are generated. Key performance parameters extracted include:

[0156] Minimum velocity point location: Scan the global velocity field to automatically find the minimum gas flow velocity. And its three-dimensional spatial location, which is usually the location with the highest risk of blockage;

[0157] Locate the point of maximum pressure loss, scan the pressure field, and find the pipe section with the largest pressure gradient;

[0158] Performance evaluation includes:

[0159] Total pressure drop calculation: Calculate the total static pressure difference between the inlet and outlet. ;

[0160] Performance evaluation, based on inlet conditions and high-risk location, comprehensively assesses the reliability and efficiency of the transport.

[0161] The report compilation and output will include all the above content, including visual images, videos, graphs, and key parameter tables (including...). and its location, The data, including the results of the simulation and diagnostic conclusions, are automatically compiled into a comprehensive simulation analysis report (usually output in PDF format).

[0162] Example 6: A system for implementing Examples 1-5 above, comprising:

[0163] The topology parsing module is configured to parse the 3D model of the pipeline network, identify its topology, and extract the feature parameters of each pipe segment.

[0164] The strategy rule base stores rules for dynamically allocating simulation models based on pipe segment characteristic parameters, fluid property parameters, and user-preset system-level objectives;

[0165] The task scheduler is configured to dynamically combine consecutive pipe segments assigned the same simulation model into computational task packages and perform parallel scheduling computations.

[0166] The data coupling and verification module is configured to manage data transfer between management segments and performs data consistency verification based on the law of conservation of mass.

[0167] The results synthesis module is configured to integrate the results of all parallel computing tasks and splice and synthesize them according to the pipeline topology to generate a system-level simulation report.

[0168] Example 7: Figure 3 As shown, this embodiment takes an L-shaped waste pneumatic conveying pipeline system as an example to demonstrate the full-process application of the present invention. The pipeline system consists of a straight pipe with a length of 10m and an inner diameter of 0.3m and a 90° bend with a curvature radius of 0.5m.

[0169] Preprocessing and topology analysis, importing CAD models, automatic extraction of centerlines from geometric kernels, and calculation of the total curvature integral of straight pipe sections. Classified into straight pipes and bends The pipes were classified as bends, and the following feature parameters were extracted: straight pipe L=10m, D=0.3m, L / D≈33.3; bend pipe R=0.5m, D=0.3m, R / D≈1.67.

[0170] Simulation dynamic allocation, querying the rule base, for straight pipe sections, because L / D=33.3<50, allocates a 3D CFD model (Standard). For the bend section, since R / D = 1.67 < 2.0, a high-precision three-dimensional CFD model (SST) is assigned. );

[0171] Task encapsulation and scheduling: Since the two pipeline models are different, they are encapsulated into two independent task packages. The scheduler detects an 8-core CPU and creates 7 worker threads to compute these two tasks in parallel.

[0172] Data coupling and verification: After the calculation of the straight pipe (upstream) is completed, the average pressure and velocity vectors at its outlet face are extracted as the inlet boundary conditions for the bend pipe (downstream). After the calculation is completed, the mass flow rate deviation at the interface is verified. The adaptive adjustment was triggered, and the inlet speed of the bend was fine-tuned by 2% before recalculation. The new deviation was 0.7%, and the verification passed.

[0173] Global result synthesis involves stitching and interpolating the pressure and velocity fields of the two pipe sections to generate a system-level pressure contour map and velocity vector map. The report shows that the total system pressure drop ΔP = 1250 Pa, and the lowest velocity point is located on the outside of the bend. The speed is higher than the critical speed for blockage, indicating a reasonable system design.

[0174] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention 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 will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for managing multi-fidelity co-simulation of a pneumatic transport pipe network system, characterized in that, Includes the following steps: S1: Topology analysis and feature extraction: Automatically identify the input 3D model of the pipeline network, analyze its topology, and extract the feature parameters of each pipe segment. The feature parameters include at least the pipe segment type, geometric dimensions, and radius of curvature. S2: Simulation dynamic allocation, based on a preset strategy rule base, dynamically allocates simulation models to each pipe segment. The allocation rules of the strategy rule base are based on the characteristic parameters of the pipe segment, fluid property parameters and user-preset system-level objectives. S3: Simulation task encapsulation and scheduling, which dynamically combines continuous pipe segments assigned the same simulation model into a calculation task package, and the task scheduler calls the corresponding solvers in parallel to perform calculations; S4: Data Coupling and Consistency Verification: The calculation results of the upstream pipe segment are used as the input boundary conditions for the calculation of the downstream pipe segment, and data consistency verification is performed. S5: Global result synthesis, automatically collects the results of all parallel computing tasks, splices and synthesizes them according to the pipeline topology, and generates a system-level simulation report.

2. A method for multi-fidelity co-simulation management of a pneumatic transport pipe network system according to claim 1, characterized in that, In S2, the specific allocation rules for the policy rule base are as follows: If the pipe segment type is a straight pipe and its length-to-diameter ratio is greater than or equal to a preset threshold, then a one-dimensional theoretical model is assigned to it. If the pipe segment type is a bend or a straight pipe with a length-to-diameter ratio less than a preset threshold, then a three-dimensional computational fluid dynamics model will be assigned to it. The preset threshold is dynamically optimized based on fluid properties and system-level objectives.

3. A method for managing multi-fidelity co-simulation of a pneumatic transport pipe network system according to claim 2, characterized in that, The allocation rules of the policy rule base further include: A high-precision turbulence model is assigned to bends with a radius of curvature to diameter ratio less than a critical value; The critical value is dynamically adjusted based on fluid properties and system-level objectives.

4. A method for multi-fidelity co-simulation management of a pneumatic transport pipe network system according to claim 1, characterized in that, In S3, the dynamic combination into a computational task package specifically includes identifying branch points in the topology and combining continuous pipe segments between branch points that have been assigned the same simulation model into independent task packages.

5. A method for multi-fidelity co-simulation management of a pneumatic transport pipe network system according to claim 1, characterized in that, The specific steps involved in data consistency verification in S4 are: Monitor the mass flow rate and pressure data at the interface of adjacent pipe sections. If the difference exceeds the allowable tolerance, trigger the adaptive adjustment mechanism. The adaptive adjustment mechanism includes recalculating the boundary conditions of the upstream pipe segment or adjusting the boundary conditions of the downstream pipe segment.

6. A method for multi-fidelity co-simulation management of a pneumatic transport pipe network system according to claim 5, characterized in that, The adaptive adjustment mechanism prioritizes a local fine-tuning strategy, recalculating and coupling only adjacent pipe segments with inconsistent data until the data meets the consistency requirements.

7. A method for multi-fidelity co-simulation management of a pneumatic transport pipe network system according to claim 1, characterized in that, In S3, the task scheduler dynamically allocates the number of parallel computing tasks based on the number of available computing cores and memory resources of the current computer system, thereby achieving load balancing.

8. A method for multi-fidelity co-simulation management of a pneumatic transport pipe network system according to claim 1, characterized in that, Before starting S1, a preprocessing step is also included: Receive system-level target parameters input by the user and convert them into allocation rule thresholds in the policy rule base; The system-level target parameters are the maximum allowable computation time or the expected simulation accuracy level.

9. A method for managing multi-fidelity co-simulation of a pneumatic transport pipe network system according to claim 8, characterized in that, The preprocessing step optimizes preset thresholds and critical values ​​based on system-level target parameters.

10. A co-simulation management system for implementing any of the methods of claims 1 to 9, characterized by include: The topology parsing module is configured to parse the 3D model of the pipeline network, identify its topology, and extract the feature parameters of each pipe segment. The strategy rule base stores rules for dynamically allocating simulation models based on pipe segment characteristic parameters, fluid property parameters, and user-preset system-level objectives; The task scheduler is configured to dynamically combine consecutive pipe segments assigned the same simulation model into computational task packages and perform parallel scheduling computations. The data coupling and verification module is configured to manage data transfer between management segments and performs data consistency verification based on the law of conservation of mass. The results synthesis module is configured to integrate the results of all parallel computing tasks and splice and synthesize them according to the pipeline topology to generate a system-level simulation report.

Images