Numerical Singularity Processing Methods and Related Devices for Finite Element Simulation of Electric Arc
By dynamically identifying and correcting local abnormal mesh elements in the finite element simulation of electric arc, and using a volume-weighted average formula for correction, the iterative divergence problem caused by local non-physical anomalies is solved, thus improving the simulation stability and accuracy.
Patent Information
- Application Number
- CN202611295288.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-25
- Publication Date
- 2026-11-10
AI Technical Summary
In existing finite element simulations of electric arcs, local non-physical anomalies are difficult to identify and correct effectively, leading to iterative divergence and affecting the stability and accuracy of the simulation.
By traversing the grid cells in each operation cycle, calculating the local gradient characteristics, identifying local abnormal grid cells by combining the abnormal gradient judgment conditions, and correcting them using the volume-weighted average formula, local conservation is ensured.
The numerical singularities in the finite element simulation of electric arcs are dynamically identified and corrected to avoid outlier propagation, improve simulation stability and accuracy, and reduce iterative divergence.
Smart Images

Figure CN123069769A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of finite element simulation technology for circuit breakers, and in particular to a numerical singularity processing method and related apparatus for finite element simulation of electric arcs. Background Technology
[0002] As a core control and protection device in power systems, circuit breakers involve complex mechanisms such as contact movement, gas compression and expansion, arc heating, electromagnetic force drive, radiative heat dissipation, and high-speed arc blowing from nozzles during the interruption of large currents. To optimize arc-extinguishing structures, shorten product development cycles, and reduce costly interruption tests, CFD and finite element methods are widely used in engineering design to simulate the gas arc within the arc-extinguishing chamber.
[0003] In practical numerical simulation applications, a special type of numerical problem often arises—local non-physical anomalies. These are defined as isolated numerical spikes lacking real physical support, generated by mesh distortion, discretization errors, variable mapping errors, or nonlinear iteration errors. These anomalies typically appear only in a single or a few adjacent mesh cells, significantly deviating from the spatial distribution patterns of surrounding cells or the evolutionary trends of adjacent time steps. They differ from real physical high gradients such as arc boundaries, shock waves, and arc roots in that real high gradients usually have a continuous spatial structure and a clear physical origin, while local non-physical anomalies are characterized by spatial isolation, temporal bursts, and unreasonable physical states.
[0004] Once local non-physical anomalies are generated, they will be transmitted to adjacent cells through convection, diffusion, Joule heating, and physical property feedback. For example, an isolated anomalously high-temperature cell will cause its electrical conductivity to increase abnormally, which in turn will increase the local Joule heat source and form a non-physical positive feedback; abnormal negative density, negative specific heat, or negative thermal conductivity will cause the governing equations to lose their physical meaning; isolated pressure spikes may induce anomalous velocities, pressure waves, and mesh displacements.
[0005] Existing methods for handling such outliers typically involve covering them with the arithmetic mean of adjacent cells or deleting outlier data in the post-processing stage. However, in practical applications, the former fails to consider differences in cell volume, mesh quality, and distance, and cannot automatically guarantee local conservation mathematically. The latter only improves the contour plot display but cannot prevent outliers from propagating during the solution process or solve the iterative divergence problem. Therefore, providing a method that can dynamically identify and correct local non-physical outliers during the solution process has become a pressing technical problem for those skilled in the art. Summary of the Invention
[0006] This application provides a numerical singularity processing method and related apparatus for electric arc finite element simulation, which is used to dynamically identify and correct numerical singularities in electric arc finite element simulation, avoid iterative divergence caused by outlier propagation, and improve the stability and accuracy of numerical simulation.
[0007] To achieve the aforementioned objectives, the first aspect of this application provides a numerical singularity processing method for finite element simulation of electric arcs, comprising: In each computation cycle, the target variable values of each grid cell in the current computation cycle and the spatial gradients corresponding to the target variable values are traversed. For any grid cell to be inspected, a local statistical neighborhood is determined with the grid cell to be inspected as the center, and the local gradient features corresponding to the grid cell to be inspected are calculated based on the spatial gradient of each grid cell in the local statistical neighborhood. Based on the local gradient characteristics and combined with the preset abnormal gradient determination conditions, local abnormal mesh cells that satisfy the abnormal gradient determination conditions are determined. When a local abnormal grid cell is detected, a correction neighborhood is determined with the local abnormal grid cell as the center. Based on the target variable values of the neighboring grid cells in the correction neighborhood, a corrected value of the target variable is calculated using a preset volume-weighted average formula to cover the original value of the target variable of the local abnormal grid cell. The weight term in the volume-weighted average formula is determined by the geometric volume of the grid cell. If all grid cells in the current operation cycle have been traversed, then proceed to the next operation cycle.
[0008] Preferably, the local gradient features include: the spatial gradient magnitude of the target variable, the local average gradient, and the local gradient standard deviation.
[0009] Preferably, the expression for the outlier gradient determination condition is: ; In the formula, For local average gradient, For the local gradient standard deviation, The preset sensitivity coefficient, This represents the spatial gradient magnitude.
[0010] Preferably, the volume-weighted average formula is as follows: ; In the formula, Let i be the set of neighboring mesh cells corresponding to the local abnormal mesh cell i. The geometric volume of the neighboring mesh element j in the corrected neighborhood. The target variable value of grid cell j in the corrected neighborhood is... This is the correction value of the target variable corresponding to the local abnormal mesh element i.
[0011] Preferably, the range of the local statistical neighborhood is larger than the modified neighborhood.
[0012] Preferably, it further includes: When multiple local abnormal grid cells are detected in the correction neighborhood, the local abnormal grid cells are excluded when calculating the correction value of the target variable using a preset volume-weighted average formula. When the number of remaining valid neighboring grid cells is less than the preset minimum number of valid neighbors, the range of the corrected neighborhood is expanded until the number of valid neighboring grid cells in the corrected neighborhood meets the minimum number of valid neighbors.
[0013] A second aspect of this application provides a numerical singularity processing device for finite element simulation of electric arcs, comprising: The grid traversal module is used to traverse the target variable values of each grid cell in each operation cycle, as well as the spatial gradients corresponding to the target variable values; The local gradient feature determination module is used to determine a local statistical neighborhood centered on any grid cell to be inspected, and to calculate the local gradient features corresponding to the grid cell to be inspected based on the spatial gradient of each grid cell in the local statistical neighborhood. An abnormal mesh identification module is used to determine local abnormal mesh cells that satisfy the abnormal gradient determination conditions based on the local gradient features and in combination with preset abnormal gradient determination conditions. An abnormal mesh correction module is used to determine a correction neighborhood centered on the local abnormal mesh cell when a local abnormal mesh cell is detected. Based on the target variable values of the neighboring mesh cells in the correction neighborhood, a corrected value of the target variable is calculated using a preset volume-weighted average formula to cover the original value of the target variable of the local abnormal mesh cell. The weight term in the volume-weighted average formula is determined by the geometric volume of the mesh cell. The cycle transformation determination module is used to proceed to the next operation cycle if all grid cells in the current operation cycle have been traversed.
[0014] A third aspect of this application provides an electronic device, including: a memory and a processor; The memory is used to store program code, which corresponds to the numerical singularity processing method for arc finite element simulation as provided in the first aspect of this application. The processor is used to read and execute the program code to implement the numerical singularity processing method for electric arc finite element simulation.
[0015] The fourth aspect of this application provides a computer-readable storage medium storing program code, which is read and executed by a processor to implement the numerical singularity processing method for arc finite element simulation provided in the first aspect of this application.
[0016] The fifth aspect of this application provides a computer program product consisting of program code, which corresponds to the numerical singularity processing method for arc finite element simulation provided in the first aspect of this application, and is used to be read and executed by a processor to implement the numerical singularity processing method for arc finite element simulation.
[0017] As can be seen from the above technical solutions, this application has the following advantages: The solution provided in this application dynamically identifies local abnormal mesh elements, traverses the mesh elements and performs real-time detection in each computation cycle, and combines an anomaly judgment mechanism based on local gradient features to promptly detect and process outliers. It also calculates correction values by considering the geometric volume of neighboring mesh elements to ensure local conservation, making the correction results more consistent with physical reality and effectively avoiding outlier propagation. It has the advantages of dynamically identifying and correcting numerical singularities in electric arc finite element simulation, avoiding iterative divergence caused by outlier propagation, and improving the stability and accuracy of numerical simulation. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is a schematic diagram of the core process of an embodiment of a numerical singularity processing method for finite element simulation of electric arc provided in this application.
[0020] Figure 2 This is a schematic diagram of the overall process of an embodiment of a numerical singularity processing method for finite element simulation of electric arc provided in this application.
[0021] Figure 3 This is a schematic diagram of the local statistical neighborhood and the modified neighborhood in the numerical singularity processing method for electric arc finite element simulation provided in this application.
[0022] Figure 4This is a schematic diagram of the architecture of a numerical singularity processing device embodiment for finite element simulation of electric arc provided in this application. Detailed Implementation
[0023] In traditional finite element method (FEM) simulations of electric arcs, the generation of local non-physical anomalies leads to numerical instability. These anomalies, caused by mesh distortion, discretization errors, variable mapping errors, or nonlinear iteration errors, manifest as isolated numerical spikes lacking real physical mechanism support in single or a few adjacent mesh elements. These anomalies are characterized by spatial isolation, temporal suddenness, and unreasonable physical states, fundamentally different from real physical high gradients. Furthermore, once generated, these anomalies are transmitted to adjacent elements through convection, diffusion, Joule heating, and physical property feedback. For example, abnormally high-temperature elements cause an abnormal increase in electrical conductivity, leading to an increase in local Joule heat sources; abnormal negative density, negative specific heat, or negative thermal conductivity renders the governing equations meaningless; isolated pressure spikes may induce abnormal velocities, pressure waves, and mesh displacements. Therefore, this problem affects the accuracy, stability, and convergence of the simulation, making the calculation results unreliable.
[0024] For example, in the numerical simulation of gas arc in the arc-extinguishing chamber of a circuit breaker, an isolated anomalous high-temperature point was detected in the grid cells of the high-speed arc-blowing region of the nozzle at the moment of contact separation. This anomalous point exists only in a specific grid cell, its temperature value deviates from the spatial distribution pattern of the surrounding cells, and there is no physical mechanism to explain it. Due to this anomalous high temperature, the calculated conductivity value increases non-physically, leading to an enhancement of the local Joule heat source, which in turn causes pressure field distortion and grid displacement. This phenomenon makes the calculation process unable to converge in subsequent iterations, requiring the simulation to be interrupted and parameters readjusted, affecting the efficiency of arc-extinguishing structure optimization. If the above problems are not solved, local non-physical anomalies will continue to propagate and amplify during the solution process, causing the simulation results to deviate from physical reality. The governing equations may fail due to negative physical property parameters, making the entire calculation process impossible to complete. In addition, the iterative divergence problem will increase the consumption of computational resources, prolong the product development cycle, and force engineers to rely on costly breaking tests for verification, failing to fully utilize the role of numerical simulation in reducing the number of tests and optimizing the design.
[0025] This application provides a numerical singularity processing method and related apparatus for finite element simulation of electric arcs, used to dynamically identify and correct numerical singularities in finite element simulation of electric arcs, avoid iterative divergence caused by outlier propagation, and improve the stability and accuracy of numerical simulation.
[0026] To make the inventive objectives, features, and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0027] First, this application provides a detailed description of an embodiment of a numerical singularity processing method for finite element simulation of electric arcs, as follows: Please see Figure 1 This application provides a numerical singularity processing method for finite element simulation of electric arcs, the steps of which include: Step 101: In each operation cycle, traverse the target variable value of each grid cell in the current operation cycle, and the spatial gradient corresponding to the target variable value; Step 102: For any grid cell to be inspected, determine a local statistical neighborhood centered on the grid cell to be inspected, and calculate the local gradient features corresponding to the grid cell to be inspected based on the spatial gradient of each grid cell in the local statistical neighborhood. Step 103: Based on the local gradient characteristics and combined with the preset abnormal gradient determination conditions, determine the local abnormal mesh cells that satisfy the abnormal gradient determination conditions. Step 104: When a local abnormal grid cell is detected, a correction neighborhood is determined with the local abnormal grid cell as the center. Based on the target variable values of the neighboring grid cells in the correction neighborhood, the target variable correction value is calculated using a preset volume weighted average formula to cover the original target variable value of the local abnormal grid cell. The weight term in the volume-weighted average formula is determined by the geometric volume of the grid cell; Step 105: If all grid cells in the current operation cycle have been traversed, proceed to the next operation cycle.
[0028] For ease of understanding, the following explains some key terms in this embodiment: Arc finite element simulation refers to the process of numerically simulating the phenomenon of electric arcs using the finite element method. This simulation aims to approximate the physical field equations of the arc region using discretized mesh elements, thereby predicting the distribution of temperature, pressure, velocity, and other parameters of the arc to optimize the performance of related equipment.
[0029] Numerical singularities, in this embodiment, specifically refer to local non-physical anomalies that occur during finite element simulations. These anomalies typically manifest as significant deviations of the target variable value from the spatial distribution of surrounding elements within a single or a few adjacent mesh cells, lacking any real physical mechanism to support their existence. Such anomalies may be caused by mesh distortion, discretization errors, variable mapping errors, or nonlinear iteration errors.
[0030] A computation cycle refers to the time step or iteration step in the iterative solution process of a finite element simulation, during which a complete calculation and update of the entire computational domain is performed. In each computation cycle, the physical quantities of the mesh elements are updated, and corresponding checks and processing are carried out.
[0031] In the finite element method, a mesh element is a unit of data that discretizes a continuous physical domain into a finite number of small regions. These mesh elements together form the computational mesh for the simulation.
[0032] The target variable value refers to the physical quantity represented by each mesh element in a finite element simulation. This embodiment uses characters... The variables referred to can specifically include: temperature, pressure, electrical conductivity, fluid density, etc. The values of these variables are the core of the simulation solution.
[0033] Spatial gradient refers to the rate of change of the target variable value in space, and can be expressed as: By calculating the spatial gradient, we can reflect the degree of difference in the value of the target variable between different grid cells. A high gradient usually indicates a region of drastic change in the physical quantity.
[0034] The grid cell to be inspected refers to the grid cell that is being checked by the system in the current operation cycle to determine whether there are any numerical anomalies.
[0035] A local statistical neighborhood refers to a local area defined by extending outwards from the grid cell to be inspected within a certain range. This area typically includes the grid cell to be inspected and several surrounding neighboring grid cells, and is used to obtain broader local information for anomaly detection.
[0036] Local gradient features refer to one or more numerical indices calculated by analyzing the spatial gradient of each grid cell in a local statistical neighborhood. These features are used to quantify the gradient distribution around the grid cell under investigation, thereby helping to determine whether the grid cell is an outlier.
[0037] Anomaly gradient determination criteria refer to preset rules or mathematical expressions used to determine whether local gradient features satisfy anomaly conditions. When a local gradient feature satisfies this condition, the corresponding mesh cell is considered to potentially contain a numerical singularity.
[0038] Locally anomalous mesh cells are those identified as containing numerical singularities after being judged by the anomalous gradient determination criteria. The target variable value of this mesh cell needs to be corrected.
[0039] The correction neighborhood refers to a local region defined by extending outwards from the local anomalous grid cell. This region is usually smaller than the local statistical neighborhood and is mainly used to collect the target variable values of the neighboring grid cells for correction calculations.
[0040] Neighborhood grid cells refer to the effective grid cells surrounding the local anomalous grid cells in the correction neighborhood. The target variable values of these cells will be used to calculate the correction value.
[0041] The volume-weighted average formula refers to a method for calculating an average value, in which the target variable value of each grid cell participating in the averaging is assigned a different weight based on its own geometric volume. Cells with larger geometric volumes have a greater impact on the average value.
[0042] The target variable correction value refers to the new value obtained by calculating the volume weighted average formula, which is used to replace the original target variable value of the local abnormal grid cell.
[0043] The original value of the target variable refers to the original target variable value that the local abnormal mesh cell has before it is corrected.
[0044] Geometric volume refers to the actual volume occupied by each mesh cell in three-dimensional space. In volume-weighted averaging, this volume is used as a weighting factor.
[0045] This embodiment provides a numerical singularity handling method for finite element simulation of electric arcs, and its main technical features are described in detail below: First, in each computation cycle, the system is configured to traverse the target variable value and its corresponding spatial gradient for each grid cell within the current computation cycle. This traversal process can be implemented in several ways. For example, a sequential scan can be used, accessing the data of each grid cell one by one according to a preset grid cell index order. Alternatively, parallel processing can be used, dividing the grid cells into multiple subsets, with different processing units simultaneously accessing and reading the data. The purpose of this traversal is to obtain the physical state information of all grid cells at the current moment, providing basic data for subsequent anomaly detection.
[0046] Secondly, for any grid cell to be inspected, a local statistical neighborhood is determined centered on that grid cell. Based on the spatial gradients of each grid cell within this local statistical neighborhood, the local gradient characteristics corresponding to that grid cell are calculated. The determination of this local statistical neighborhood can employ various strategies. For example, it can simply select the directly adjacent grid cells of the grid cell to be inspected as the neighborhood, or select grid cells a certain number of layers away from the grid cell to be inspected as the neighborhood. After obtaining the spatial gradients of all grid cells within this local statistical neighborhood, various forms of local gradient characteristics can be calculated. For example, the arithmetic mean of all spatial gradients within the neighborhood can be calculated, or the maximum or minimum value of the spatial gradients within the neighborhood can be calculated. These calculation results will serve as indicators characterizing the gradient distribution around the grid cell to be inspected.
[0047] Next, based on the local gradient characteristics and a preset anomaly gradient determination criterion, local anomaly mesh cells that meet the criterion are identified. This anomaly gradient determination criterion can be a simple threshold comparison. For example, if the calculated local gradient characteristic (such as the average gradient) exceeds a preset fixed threshold, the mesh cell under investigation is considered potentially an anomaly. Alternatively, the criterion can be based on statistical principles; for example, if the deviation of the local gradient characteristic from the surrounding environment exceeds a certain standard deviation, it is judged as an anomaly; otherwise, it is judged as normal. This step aims to distinguish mesh cells with non-physically drastic changes by analyzing local gradient characteristics.
[0048] When a local anomalous grid cell is detected, a correction neighborhood is determined centered on that anomalous grid cell. Based on the target variable values of the neighboring grid cells within this correction neighborhood, a corrected value for the target variable is calculated using a pre-defined volume-weighted average formula. This corrected value is then used to cover the original target variable value of the local anomalous grid cell. The correction neighborhood is typically smaller than the local statistical neighborhood; for example, it may only include the directly adjacent grid cells of the local anomalous grid cell. After determining the correction neighborhood, the target variable values of all neighboring grid cells within it are collected. The weights in the volume-weighted average formula are determined by the geometric volume of the grid cells. Specifically, a weighted average of the target variable values of all neighboring grid cells in the correction neighborhood can be performed, where the weight of each neighboring grid cell is proportional to its own geometric volume. For example, a larger neighboring grid cell will have a greater impact on the corrected value. This corrected value is then used to replace the original target variable value of the local anomalous grid cell, thereby eliminating numerical singularities.
[0049] Finally, if all mesh cells in the current computation cycle have been traversed, the system is configured to proceed to the next computation cycle. This step ensures that the numerical singularity handling method continues throughout the simulation process, thereby dynamically identifying and correcting any outliers that may occur, maintaining the stability and physical plausibility of the simulation.
[0050] The following example will provide a more detailed explanation of the above technical solution: Suppose we are simulating the arc behavior inside the arc-extinguishing chamber of a circuit breaker in a finite element simulation scenario. During a specific computation cycle, due to the instability of numerical computation, a non-physically abnormally high temperature suddenly appears at mesh cell A, far exceeding the temperature of the surrounding mesh cells, forming a numerical singularity.
[0051] Can be combined Figure 2 As shown in the flowchart, firstly, in the current computation cycle, the system initiates a traversal process of the grid cells. For each grid cell, including grid cell A and its surrounding grid cells, the current target variable value (e.g., temperature value) and the corresponding spatial gradient are read and recorded one by one.
[0052] Next, when the system traverses to grid cell A, it determines a local statistical neighborhood centered on grid cell A. This local statistical neighborhood may include grid cell A and the two surrounding grid cells. The system collects the temperature spatial gradient information of all grid cells within this local statistical neighborhood. Based on this gradient information, it calculates the local gradient characteristics corresponding to grid cell A. For example, it can calculate the average of all gradient values in this region, or calculate the variance of the gradient values, to reflect the severity and uniformity of temperature changes around grid cell A.
[0053] Subsequently, the calculated local gradient characteristics of grid cell A are compared with a preset abnormal gradient determination criterion. This criterion might be a simple rule; for example, if the local gradient characteristics (such as the average gradient) of grid cell A are significantly higher than the average level of its surrounding grid cells, or if its gradient variance is far beyond the normal range, then grid cell A is determined to be a locally abnormal grid cell. In this example, because the temperature of grid cell A is abnormally high, its gradient with that of its surrounding grid cells will be very large, thus satisfying the abnormal gradient determination criterion.
[0054] Once grid cell A is detected as a locally anomalous grid cell, the system immediately determines a correction neighborhood centered on grid cell A. This correction neighborhood is typically smaller than the local statistical neighborhood; for example, it may only include grid cells directly adjacent to grid cell A. The system collects the target temperature variable values of all neighboring grid cells within this correction neighborhood. Then, a corrected target variable value is calculated using a pre-defined volume-weighted average formula. In this formula, the temperature value of each neighboring grid cell is assigned a different weight based on its geometric volume; larger neighboring grid cells contribute more to the correction value. For example, if the volume of a neighboring grid cell B of grid cell A is twice that of grid cell C, then the temperature value of grid cell B will receive twice the weight of grid cell C when calculating the correction value. The final calculated corrected target variable value is used to cover the original anomalous temperature value of grid cell A, thereby adjusting the temperature of grid cell A to a more reasonable level that is more in line with its surrounding physical environment.
[0055] After the correction of mesh cell A is completed, the system continues to traverse other mesh cells within the current computation cycle, repeating the above detection and correction process. Once all mesh cells have been traversed and all detected outliers have been processed, the system enters the next computation cycle to continue the simulation calculation. Through this dynamic, real-time processing mechanism, this method can effectively suppress the generation and propagation of numerical singularities, ensuring the stability of the electric arc finite element simulation and the physical rationality of the calculation results.
[0056] In the above basic implementation, this application proposes a numerical singularity processing method for finite element simulation of electric arcs. Based on this, this application further proposes that the local gradient features include: the spatial gradient magnitude of the target variable, the local average gradient, and the local gradient standard deviation.
[0057] The spatial gradient magnitude refers to the intensity of the rate of change of the target variable in space. Its function is to directly quantify the degree of change of the target variable within a local region; the larger the gradient magnitude, the more significant the spatial change of the solution variable near the grid cell. This magnitude can be calculated in several ways. For example, by differentiating the shape function within the grid cell to obtain the partial derivatives of the target variable in each coordinate direction, and then calculating the square root of the sum of the squares of these partial derivatives; or, by using the finite difference method to approximate the gradient by the difference in the target variable values of adjacent grid cells, and then obtaining its magnitude.
[0058] The local average gradient refers to the average trend of the spatial gradient of the target variable across all grid cells within a local statistical neighborhood. Its function is to provide a benchmark for the overall gradient level within a local region, helping to assess the gradient of an individual grid cell relative to its neighborhood. This average gradient can be obtained by arithmetically averaging the gradient vectors of all grid cells within the local statistical neighborhood; alternatively, it can be a weighted average of the gradients based on the geometric characteristics of the grid cells (such as volume or area) to more accurately reflect the overall gradient distribution within the local region.
[0059] The local gradient standard deviation refers to the dispersion of the spatial gradient of the target variable within a local statistical neighborhood. Its function is to reflect the uniformity or volatility of gradient changes within a local region, effectively identifying anomalous concentrations or dispersions in the local gradient distribution. This standard deviation can be obtained by calculating the square root of the average of the squares of the deviations of the gradient magnitudes (or their components) of all grid cells within the local statistical neighborhood from the local average gradient; alternatively, it can be obtained by performing an analysis of variance on the gradient data within the local region using statistical methods.
[0060] The further approach proposed in this embodiment refines the local gradient features into the spatial gradient magnitude of the target variable, the local average gradient, and the local gradient standard deviation, making the characterization of the local gradient behavior of mesh elements more comprehensive and accurate. In finite element simulations of electric arcs, numerical singularities often manifest as abnormally drastic changes in the target variable within a very small region. These changes are not only reflected in the gradient magnitude but also lead to non-uniformity in the local gradient distribution. By introducing the local average gradient, a benchmark for gradient changes within a local region can be established, allowing the gradient change of a single mesh element to be compared with the overall trend of its neighborhood. Simultaneously, the local gradient standard deviation quantifies the dispersion of this local gradient distribution, i.e., determining whether the gradient within a local region is generally high and uniform, or whether there are individual mesh elements with abnormally prominent gradients. This multi-dimensional and comprehensive gradient feature description can effectively distinguish between normal steep gradients caused by physical phenomena (e.g., the temperature gradient in the core region of an electric arc) and true numerical singularities caused by numerical computation instability or mesh quality issues. When these three features are combined for anomaly gradient determination, the gradient patterns unique to numerical singularities can be captured more accurately, thus avoiding misjudging normal physical gradient changes as singularities or missing real singularities. This significantly improves the accuracy and robustness of singularity identification, thereby ensuring the reliability of the finite element simulation results of electric arcs.
[0061] In some of the above embodiments, this application proposes to determine local abnormal mesh elements that satisfy the abnormal gradient determination condition based on local gradient characteristics, and further refines the specific local gradient characteristics. Based on this, this application further proposes the following expression for the outlier gradient determination condition: ; In the formula, For local average gradient, For the local gradient standard deviation, The preset sensitivity coefficient, The magnitude of the spatial gradient; When the gradient magnitude of the grid cell i to be tested satisfies the following formula, it can be preliminarily determined that the solution variable in the grid cell is an outlier.
[0062] Through the above technical solution, this application provides an anomaly gradient determination method based on statistical principles. This method dynamically sets the anomaly determination threshold according to the gradient distribution characteristics of local regions, rather than using a fixed threshold. This makes the identification process of anomaly mesh cells more adaptive and robust, effectively distinguishing between anomaly gradients caused by real physical singularities or numerical instabilities and normal physical changes or numerical noise. Therefore, in electric arc finite element simulation, it can more accurately locate numerical singularities, reducing the risk of misjudgment and missed judgment, thus providing a more reliable basis for subsequent target variable correction, and ultimately helping to improve the stability and accuracy of simulation results.
[0063] In some of the above embodiments, this application proposes to calculate the corrected value of the target variable using a volume-weighted average formula to cover the original value of the target variable in local abnormal mesh cells. Based on this, this application further proposes that the volume-weighted average formula is specifically as follows: ; In the formula, Let i be the set of neighboring mesh cells corresponding to the local abnormal mesh cell i. The geometric volume of the neighboring mesh element j in the corrected neighborhood. The target variable value of grid cell j in the corrected neighborhood is... This is the correction value of the target variable corresponding to the local abnormal mesh element i.
[0064] It should be noted that when a monitored variable in a certain grid cell i is determined to be numerically abnormal, this method does not delete the grid cell or set the corresponding variable to a null value. The finite element grid cell is the fundamental carrier for discretizing the governing equations, interpolating shape functions, numerical integration, and assembling the global equations. Directly deleting the grid cell would disrupt the grid topology and the structure of the discrete equations in the computational domain; directly deleting variable data would result in a lack of effective state variables at that location, preventing the governing equations from closing and potentially disrupting the continuity of local mass, momentum, or energy transfer. Therefore, this method preserves the original grid cell and its topological connections, only performing local reconstruction on the variable values determined to be abnormal.
[0065] Specifically, the corrected neighborhood N(i) is formed by selecting directly adjacent grid cells around the anomalous cell that have not been identified as anomalous. Preferably, a 3x3 grid is used with the anomalous cell as the center. 3. Grid Region. Based on the variable values and geometric volumes of each effective neighboring cell within the correction neighborhood, the reconstructed value of the anomalous cell, i.e., the target variable correction value, is calculated using a volume-weighted average method. The original abnormal variable value is replaced with the reconstructed value.
[0066] The proposed solution explicitly defines a volume-weighted average formula. When a local anomalous grid cell is detected, a correction neighborhood is determined centered on that anomalous grid cell. The correction value of the target variable is then accurately calculated based on the target variable values of the neighboring grid cells within this correction neighborhood. This formula multiplies the target variable value of each neighboring grid cell by its geometric volume, sums these products, and divides by the sum of the geometric volumes of all neighboring grid cells, resulting in a physically meaningful weighted average. This calculation method ensures that the correction value fully considers the physical size of the neighboring grid cells and the actual distribution of their target variables, allowing larger neighboring grid cells to contribute more to the correction result, thus improving the physical rationality and accuracy of the correction. In this way, the method effectively utilizes reliable surrounding grid information to smooth and correct outliers, avoiding correction deviations or the introduction of new numerical oscillations caused by improper weight allocation. Through the aforementioned further optimization scheme, the specific form of the volume-weighted average formula was clarified, enabling more accurate and physically reasonable correction values to be obtained when correcting local anomalous mesh elements in electric arc finite element simulations. This precise weighting method ensures that the correction values fully consider the physical dimensions of neighboring mesh elements and the actual distribution of their target variables, avoiding correction deviations or the introduction of new numerical oscillations due to improper weight allocation. Therefore, it can more effectively eliminate numerical singularities in electric arc finite element simulations, improve the stability and reliability of simulation results, and ensure the physical rationality of the correction process.
[0067] In some of the above embodiments, this application proposes a numerical singularity processing method for finite element simulation of electric arcs, which involves traversing mesh cells, calculating local gradient features, identifying local anomalous mesh cells, and correcting the target variable. Based on this, this application proposes that the range of the local statistical neighborhood is larger than the corrected neighborhood.
[0068] It should be noted that this technical feature aims to optimize spatial considerations in the anomaly detection and numerical correction process. For example... Figure 3As shown, the local statistical neighborhood is used to collect sufficient environmental information to accurately evaluate the local gradient characteristics of the grid cell under inspection. Its range typically needs to be large to capture a wider range of gradient change patterns, thereby improving the accuracy of anomaly detection. In contrast, the correction neighborhood is used for local numerical correction, and its range typically needs to be smaller to ensure the locality of the correction and avoid over-smoothing effects on non-anomaly regions. For example, the range of the local statistical neighborhood can be several times the range of the correction neighborhood, or the local statistical neighborhood can cover the correction neighborhood and multiple grid layers around it. The range of the local statistical neighborhood is typically 10. 10. In practical applications, the topology can be expanded or reduced according to the number of layers, such as 11. 11 or 9 Level 9, this technical feature specifically defines the typical size of the local statistical neighborhood used to compute local gradient features. More specifically, this range can be a 10-degree radius centered on the grid cell to be examined. A square grid area of 10, or a circular area that approximately covers the same area.
[0069] The range of the corrected neighborhood is typically 3. 3. This technical feature specifically defines the typical size of the correction neighborhood used to calculate the corrected value of the target variable. In the finite element mesh, "3 The "3" typically represents a rectangular region centered on the local anomaly grid cell, containing 9 grid cells in 3 rows and 3 columns. This region is chosen to allow for a weighted average of the target variable values from the nearest effective neighboring grid cells when numerically correcting the local anomaly grid cell, thus achieving a localized and accurate correction. A smaller correction region helps preserve the local details of the simulation results and prevents the impact of anomaly correction from spreading to a wider area, thereby maintaining the accuracy of the overall simulation data. For example, this region could be a 3x3 square grid region centered on the local anomaly grid cell, or a circular region that approximately covers the same area.
[0070] The proposed solution enables the numerical singularity handling method to operate more effectively by setting the range of the local statistical neighborhood to be larger than the modified neighborhood and specifically limiting its typical size. In each computation cycle, when calculating the local gradient characteristics of any grid cell to be examined, a relatively large local statistical neighborhood (e.g., 10) is utilized. A larger range (e.g., 10) is used to collect spatial gradient information from surrounding grid cells. This larger range allows the system to acquire a more comprehensive local context, thereby more accurately evaluating the gradient characteristics of the grid cell under test, effectively distinguishing true numerical singularities from normal steep gradient regions, and reducing the false positive rate. Once a local anomalous grid cell is detected, a relatively small correction neighborhood (e.g., 3) is used when correcting the target variable. (The range is 3). This smaller correction region ensures a high degree of locality in the correction operation, using only the target variable values of the nearest neighboring grid cells of the anomalous grid cell for volume-weighted averaging. This design avoids unnecessary influence or over-smoothing of normal grid cells far from the anomalous point during the correction operation, thus preserving the original details and physical realism of the simulation data to the maximum extent while eliminating numerical singularities. Through this spatial domain setting with differentiated sizes, the scheme in this application achieves a good balance between the robustness of anomaly detection and the accuracy of numerical correction, making the entire numerical singularity processing process more efficient and reliable.
[0071] In other embodiments, this application proposes a numerical singularity handling method for finite element simulation of electric arcs, which identifies local anomalous mesh elements and uses the target variable values of neighboring mesh elements in their correction neighborhood for volume-weighted average correction. However, in practical applications, this may not work if the correction neighborhood itself contains multiple anomalous mesh elements, or if the number of effective neighboring mesh elements is insufficient.
[0072] In response, this application further proposes that when multiple local abnormal grid cells are detected in the correction neighborhood, the local abnormal grid cells are excluded when calculating the correction value of the target variable using a preset volume-weighted average formula; when the number of remaining valid neighboring grid cells is less than the preset minimum number of valid neighbors, the range of the correction neighborhood is expanded until the number of valid neighboring grid cells in the correction neighborhood meets the minimum number of valid neighbors.
[0073] Specifically, when multiple locally anomalous grid cells are detected within the correction neighborhood, these locally anomalous grid cells need to be excluded when calculating the corrected target variable value using a preset volume-weighted average formula. This means that when correcting the target variable, other anomalous grid cells within the correction neighborhood, except for the central anomalous grid cell to be corrected, should not be included in the calculation. One implementation is to first perform anomaly checks on all grid cells within the correction neighborhood when determining the set of neighborhood grid cells for volume-weighted averaging, and remove grid cells deemed anomalous from the set. Another implementation is to assign a status flag to each grid cell within the correction neighborhood to indicate whether it is anomalous. When calculating the volume-weighted average, only those grid cells not marked as anomalous are selected for the calculation.
[0074] When the number of remaining effective neighboring mesh cells is less than the preset minimum effective neighbor number, the range of the correction neighborhood is expanded until the number of effective neighboring mesh cells within the correction neighborhood meets the minimum effective neighbor number. Here, the "preset minimum effective neighbor number" refers to the minimum effective neighboring mesh cell threshold set to ensure the statistical reliability of the correction result. This number can be set based on experience, for example, it can be set to 3, 5, or 8, depending on the characteristics of the finite element mesh and the required correction accuracy. When the number of effective neighboring mesh cells is insufficient, the "range of the correction neighborhood" needs to be expanded. One implementation method is to gradually increase the radius or side length of the correction neighborhood to expand its range, for example, from 3... 3 expanded to 5 5, then 7 7. Continue until the minimum number of valid neighbors is met. Another approach is to predefine a series of modified neighborhood templates of different sizes. When the number of valid neighbors in the current domain is insufficient, the system automatically selects the next larger preset template to try.
[0075] The preferred solution proposed in this embodiment further optimizes the target variable correction process for local anomalous grid cells based on the aforementioned numerical singularity processing method. Upon identifying a local anomalous grid cell, the system determines a correction neighborhood centered on that anomalous grid cell and prepares to perform volume-weighted average correction using the neighboring grid cells within that neighborhood. During this process, the solution first checks the validity of the neighboring grid cells within the correction neighborhood. If other grid cells deemed anomalous exist within the neighborhood besides the central anomalous grid cell to be corrected, these anomalous grid cells are excluded from the volume-weighted average calculation. This is because the data of these anomalous grid cells may be inaccurate, and including them in the calculation would contaminate the correction result. After excluding these anomalous grid cells, the system counts the number of remaining, valid neighboring grid cells. To ensure the statistical reliability and accuracy of the correction result, this number must reach a preset minimum number of valid neighbors. If the current number of valid neighboring grid cells is insufficient, the system automatically expands the range of the correction neighborhood. This expansion process continues until, within a new, larger corrected neighborhood, after excluding anomalous grid cells, the number of remaining valid neighboring grid cells meets the preset minimum number of valid neighbors. This adaptive exclusion and expansion mechanism ensures that the corrected value of the target variable is always calculated based on a sufficient number of reliable neighboring grid cell data, thereby significantly improving the accuracy and stability of numerical singularity processing.
[0076] The above is a detailed description of an embodiment of a numerical singularity processing method for finite element simulation of electric arc provided by this application. The following is a detailed description of an embodiment of a numerical singularity processing device for finite element simulation of electric arc provided by this application.
[0077] Please see Figure 4 This application provides a numerical singularity processing device for finite element simulation of electric arcs, comprising: The grid traversal module 201 is used to traverse the target variable values of each grid cell in the current operation cycle and the spatial gradients corresponding to the target variable values in each operation cycle. The local gradient feature determination module 202 is used to determine a local statistical neighborhood centered on any grid cell to be inspected, and to calculate the local gradient features corresponding to the grid cell to be inspected based on the spatial gradient of each grid cell in the local statistical neighborhood. The abnormal mesh identification module 203 is used to determine the local abnormal mesh cells that satisfy the abnormal gradient determination conditions based on the local gradient features and in combination with the preset abnormal gradient determination conditions. The abnormal mesh correction module 204 is used to determine a correction neighborhood centered on the local abnormal mesh cell when a local abnormal mesh cell is detected, and calculate the target variable correction value based on the target variable value of the neighboring mesh cells in the correction neighborhood using a preset volume weighted average formula, so as to cover the original value of the target variable of the local abnormal mesh cell. The weight term in the volume weighted average formula is determined by the geometric volume of the mesh cell. The cycle transformation determination module 205 is used to enter the next operation cycle if all grid cells in the current operation cycle have been traversed.
[0078] This embodiment combines local gradient feature analysis with volume-weighted average correction in a specific way to dynamically identify and correct local non-physical anomalies during the solution process, achieving the effects of ensuring local conservation, preventing the propagation of anomalies, and improving simulation stability. Specifically, the mesh traversal module first collects the target variable values and their spatial gradients for all mesh cells within the current computation cycle, providing basic data for subsequent analysis. The local gradient feature determination module constructs a local statistical neighborhood centered on the mesh cell to be inspected, and calculates feature values that characterize the local gradient distribution by analyzing the statistical properties of the spatial gradient within this neighborhood. The anomaly mesh identification module compares these feature values with preset anomaly gradient judgment conditions to accurately identify isolated numerical spikes lacking real physical mechanism support. When a local anomaly mesh cell is detected, the anomaly mesh correction module immediately activates the correction mechanism, constructing a smaller correction neighborhood centered on the anomaly cell, and performing a weighted average calculation based on the geometric volume of the neighboring mesh cells to obtain a more reasonable target variable correction value. The cycle transformation judgment module ensures that the entire processing flow proceeds in an orderly manner within each computation cycle until all mesh cells have been traversed before entering the next cycle.
[0079] Through the above technical solution, the device can effectively distinguish between real physical high gradients and non-physical anomalies. Since the local statistical neighborhood is typically larger than the correction neighborhood, anomaly detection can be performed in a broader context, improving the accuracy of identification; while the correction process is implemented within a finer local range, ensuring the precision of the correction. Specifically, the weight term in the volume-weighted average formula is determined by the geometric volume of the mesh cell. This design considers the actual situation of non-uniform mesh partitioning, making the correction result more consistent with physical laws and avoiding the correction bias caused by neglecting volume differences in traditional arithmetic averaging methods. In practical applications, when an abnormally high temperature value appears in a mesh cell, the device can quickly identify it and correct it based on the volume-weighted average temperature of surrounding mesh cells, restoring the temperature value of that cell to a reasonable range. This prevents abnormal increases in conductivity and non-physical positive feedback phenomena caused by abnormally high temperatures, effectively maintaining the stability and convergence of the simulation.
[0080] In other embodiments, this application proposes an electronic device including a memory and a processor; the memory is used to store program code corresponding to the numerical singularity processing method for arc finite element simulation described above; the processor is used to read and execute the program code to implement the numerical singularity processing method for arc finite element simulation.
[0081] The electronic device refers to a hardware system capable of performing calculations, storing data, and communicating. This electronic device can be a general-purpose computer, such as a personal computer, workstation, or server, or a dedicated computing device optimized for specific computing tasks, such as a node in a high-performance computing cluster. The memory is a hardware component used to store data and program instructions. It can be volatile memory, such as random access memory (RAM), used to temporarily store program code and data during processor operation; or it can be non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), used for persistent storage of operating systems, applications, and simulation data. The processor is the core computing unit of the electronic device, responsible for executing instructions and processing data. It can be a central processing unit (CPU), such as a multi-core processor, or a graphics processing unit (GPU), digital signal processor (DSP), or microcontroller (MCU), the choice depending on the required computing power and parallel processing needs. The program code is a set of instructions that, when executed by the processor, enables the implementation of specific functions or algorithms. These program codes can be written in high-level programming languages (such as C++, Fortran, or Python) and compiled or interpreted to form machine code or bytecode that can be directly executed by the processor. The memory is used to store the program code, aiming to provide the processor with the instruction set and related data required to execute the numerical singularity processing method. The program code can be pre-programmed into non-volatile memory or loaded from non-volatile memory into volatile memory at runtime. Corresponding to the aforementioned method, the stored instructions can completely and accurately implement each step of the numerical singularity processing method, including mesh traversal, local gradient feature calculation, anomalous mesh identification, and anomalous mesh correction. The processor is used to read and execute the program code to implement the numerical singularity processing method for arc finite element simulation. This means that the processor retrieves instructions from memory through its instruction set architecture, decodes and executes them, thereby driving the entire numerical singularity processing method and completing its computational functions.
[0082] Through the above technical solution, this application provides an implementation method for integrating complex numerical singularity processing methods into electronic devices. This electronic device can automatically, efficiently, and reliably execute numerical singularity processing methods, thereby effectively solving the problems of computational instability and inaccurate results caused by numerical singularities in electric arc finite element simulation. Leveraging the powerful computing capabilities of the processor and the high-speed data access capabilities of the memory, this method can be quickly deployed and run in practical engineering applications, significantly improving the efficiency and accuracy of electric arc finite element simulation, and providing solid technical support for research and development in related fields.
[0083] In other embodiments, this application proposes a computer-readable storage medium storing program code that is read and executed by a processor to implement the above-described numerical singularity processing method for finite element simulation of electric arcs.
[0084] The computer-readable storage medium refers to a physical medium capable of storing digital data and accessible to a computer or other digital device. This medium can be a non-volatile memory, such as a hard disk drive, solid-state drive, flash memory, or optical disc, which can retain data for a long time without loss of power. Alternatively, it can be a volatile memory, such as random access memory, but is typically used for temporary storage of program code and data, loaded by the processor during program execution. The program code refers to a set of instructions written in a programming language. These instructions, when executed by the processor, can perform specific tasks or achieve specific functions. The program code is the specific instruction set for implementing the numerical singularity processing method for finite element simulation of electric arcs described above. It transforms the logical steps, calculation formulas, and judgment conditions of the method into a form that a computer can understand and execute. The program code can be written in a high-level language (such as C++, Python, Java, etc.) and then converted into machine code or bytecode by a compiler or interpreter and stored in the storage medium; alternatively, it can be written directly in machine language or assembly language to achieve higher execution efficiency. The processor is the core component of the computer system, responsible for executing the instructions in the program code and performing arithmetic, logical, and control operations. A processor is a hardware unit that executes program code stored in a computer-readable storage medium. It is responsible for performing all the calculations and judgments step-by-step according to the instruction sequence of the program code, using a numerical singularity processing method for finite element simulation of electric arcs. A processor can be a central processing unit (CPU), such as a general-purpose processor, with powerful general-purpose computing capabilities; it can also be a graphics processing unit (GPU), which can provide significant acceleration in scenarios involving a large amount of parallel computing; or it can be an application-specific integrated circuit (ASIC) or a field-programmable gate array (FPGA), which can be optimized for specific algorithms to provide extremely high execution efficiency.
[0085] By storing the numerical singularity processing method for finite element simulation of electric arcs in the form of program code on a computer-readable storage medium and having it read and executed by a processor, the scheme in this application ensures the stability and repeatability of this complex numerical processing method. This allows the method to be deployed and run in different computing environments, greatly improving its versatility and ease of use.
[0086] In some other embodiments, this application proposes a computer program product consisting of program code, which corresponds to the above-described numerical singularity processing method for arc finite element simulation and is used to be read and executed by a processor to implement the numerical singularity processing method for arc finite element simulation.
[0087] This application proposes a computer program product whose working principle involves transforming the aforementioned numerical singularity processing method for finite element simulation of electric arcs into a sequence of instructions that can be automatically executed by a computer system. Specifically, the computer program product consists of a series of program codes designed to precisely map and implement each step and logical judgment of the method. When the program product is deployed on an electronic device, the processor in the device reads these program codes from memory. The processor then executes the calculations and logical operations line by line according to the preset instruction sequence in the program code. For example, the program code instructs the processor to access and traverse the target variable values and their corresponding spatial gradient data of each mesh cell in each computation cycle. Next, the program code guides the processor to determine a local statistical neighborhood centered on any mesh cell to be inspected, and calculate local gradient characteristics based on the spatial gradients of the mesh cells within that neighborhood. Subsequently, the processor identifies locally anomalous mesh cells that meet the anomalous gradient judgment conditions defined in the program code. Once a local anomalous mesh cell is detected, the program code further instructs the processor to determine a correction neighborhood centered on that anomalous mesh cell. Based on the target variable values of the neighboring mesh cells within this neighborhood, the program code uses a volume-weighted average formula to calculate the corrected target variable value, which is then used to overwrite the original target variable value of the anomalous mesh cell. This entire process is repeated cyclically under the control of the program code until all mesh cells within the current computation cycle have been traversed. At this point, the program code instructs the processor to proceed to the next computation cycle. In this way, the computer program product enables complex numerical singularity processing methods to operate automatically and efficiently on a computer system, eliminating the need for manual intervention, thus providing stable and reliable numerical processing capabilities for arc finite element simulation.
[0088] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the terminals, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0089] In the several embodiments provided in this application, it should be understood that the disclosed terminals, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between devices or units through some interfaces, and may be electrical, mechanical, or other forms.
[0090] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0091] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.
[0092] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0093] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0094] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.
[0095] The above-described 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 numerical singularity handling method for finite element simulation of electric arcs, characterized in that, include: In each computation cycle, the target variable values of each grid cell in the current computation cycle and the spatial gradients corresponding to the target variable values are traversed. For any grid cell to be inspected, a local statistical neighborhood is determined with the grid cell to be inspected as the center, and the local gradient features corresponding to the grid cell to be inspected are calculated based on the spatial gradient of each grid cell in the local statistical neighborhood. Based on the local gradient characteristics and combined with the preset abnormal gradient determination conditions, local abnormal mesh cells that satisfy the abnormal gradient determination conditions are determined. When a local abnormal grid cell is detected, a correction neighborhood is determined with the local abnormal grid cell as the center. Based on the target variable values of the neighboring grid cells in the correction neighborhood, a corrected value of the target variable is calculated using a preset volume-weighted average formula to cover the original value of the target variable of the local abnormal grid cell. The weight term in the volume-weighted average formula is determined by the geometric volume of the grid cell. If all grid cells in the current operation cycle have been traversed, then proceed to the next operation cycle.
2. The numerical singularity processing method for finite element simulation of electric arcs according to claim 1, characterized in that, The local gradient features include: the spatial gradient magnitude of the target variable, the local average gradient, and the local gradient standard deviation.
3. The numerical singularity processing method for finite element simulation of electric arcs according to claim 2, characterized in that, The expression for the outlier gradient determination condition is as follows: ; In the formula, For local average gradient, For the local gradient standard deviation, The preset sensitivity coefficient, This represents the spatial gradient magnitude.
4. The numerical singularity processing method for finite element simulation of electric arcs according to claim 1, characterized in that, The specific formula for the volume-weighted average is as follows: ; In the formula, Let i be the set of neighboring mesh cells corresponding to the local abnormal mesh cell i. The geometric volume of the neighboring grid cell j in the modified neighborhood, The target variable value for grid cell j in the corrected neighborhood is... This is the correction value of the target variable corresponding to the local abnormal mesh element i.
5. The numerical singularity processing method for finite element simulation of electric arcs according to claim 1, characterized in that, The range of the local statistical neighborhood is larger than that of the corrected neighborhood.
6. The numerical singularity processing method for finite element simulation of electric arcs according to claim 1, characterized in that, Also includes: When multiple local abnormal grid cells are detected in the correction neighborhood, the local abnormal grid cells are excluded when calculating the correction value of the target variable using a preset volume-weighted average formula. When the number of remaining valid neighboring grid cells is less than the preset minimum number of valid neighbors, the range of the corrected neighborhood is expanded until the number of valid neighboring grid cells in the corrected neighborhood meets the minimum number of valid neighbors.
7. A numerical singularity processing device for finite element simulation of electric arcs, characterized in that, include: The grid traversal module is used to traverse the target variable values of each grid cell in each operation cycle, as well as the spatial gradients corresponding to the target variable values; The local gradient feature determination module is used to determine a local statistical neighborhood centered on any grid cell to be inspected, and to calculate the local gradient features corresponding to the grid cell to be inspected based on the spatial gradient of each grid cell in the local statistical neighborhood. An abnormal mesh identification module is used to determine local abnormal mesh cells that satisfy the abnormal gradient determination conditions based on the local gradient features and in combination with preset abnormal gradient determination conditions. An abnormal mesh correction module is used to determine a correction neighborhood centered on the local abnormal mesh cell when a local abnormal mesh cell is detected. Based on the target variable values of the neighboring mesh cells in the correction neighborhood, a target variable correction value is calculated using a preset volume weighted average formula to cover the original target variable value of the local abnormal mesh cell. The weight term in the volume weighted average formula is determined by the geometric volume of the mesh cell. The cycle transformation determination module is used to proceed to the next operation cycle if all grid cells in the current operation cycle have been traversed.
8. An electronic device, characterized in that, include: Memory and processor; The memory is used to store program code, which corresponds to the numerical singularity processing method for arc finite element simulation as described in any one of claims 1 to 6; The processor is used to read and execute the program code to implement the numerical singularity processing method for electric arc finite element simulation.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that is read and executed by a processor to implement the numerical singularity processing method for arc finite element simulation as described in any one of claims 1 to 6.
10. A computer program product, characterized in that, The computer program product consists of program code, which corresponds to the numerical singularity processing method for arc finite element simulation as described in any one of claims 1 to 6, and is used to be read and executed by a processor to implement the numerical singularity processing method for arc finite element simulation.
