Mechanism model based electrical distribution box thermal fault prediction system

Abstract

The present application relates to the technical field of computer-aided design, in particular to a distribution box thermal fault prediction system based on a mechanism model, which comprises a model voxelization processing module, used for calling a three-dimensional model of a distribution box containing busbars and circuit breaker components, performing spatial discretization on the three-dimensional model of the distribution box, establishing thermal conduction connection edges between adjacent voxel units, and generating a voxelized dual graph.In the present application, the number of non-empty grids is counted and the streamline fractal dimension is calculated by covering different scale space grids, so that the geometric distribution characteristics of the flow field are converted into quantifiable complexity indicators, and then the flow field dimension attenuation coefficient distribution is generated in combination with the reference dimension value under the standard operating condition, thereby realizing quantitative identification of the dust deposition state of the air inlet.The thermal fault prediction is extended from a single temperature criterion to the structure coupling and flow evolution level, the forward-looking of thermal anomaly identification is improved, and the calculation burden caused by the model size is reduced.

Description

Technical Field

[0001] This invention relates to the field of computer-aided design technology, and in particular to a thermal fault prediction system for distribution boxes based on mechanistic models. Background Technology

[0002] Existing technologies often rely on overall thermal balance or local temperature thresholds for judgment during actual operation, lacking a detailed representation of internal geometric connectivity. Heat conduction paths are typically treated as overall equivalent parameters. When the internal structure is complex or there are local obstructions, the temperature rise distribution becomes uneven, and the overall parameter model struggles to accurately reflect local heat accumulation areas, masking early potential problems. Furthermore, airflow status is often judged using average airflow or single-point velocity. Even with some dust accumulation at the air inlet while the fan maintains its rated speed, the overall airflow value may remain stable, but the internal flow structure has already changed. The reduced local cooling capacity cannot be identified by traditional indicators, leading to the continued development of heat accumulation. Therefore, improvements are needed. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a distribution box thermal fault prediction system based on a mechanism model.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: a distribution box thermal fault prediction system based on a mechanism model includes: The model voxelization module is used to retrieve the 3D model of the distribution box containing busbars and circuit breaker components, perform spatial discretization on the 3D model of the distribution box, establish thermal conduction connection edges between adjacent voxel units, generate a voxelized dual graph, construct the Laplacian matrix corresponding to the voxelized dual graph, perform eigenvalue decomposition on the Laplacian matrix, extract the second smallest eigenvalue of the matrix and the corresponding eigenvector, cluster the voxel nodes according to the eigenvector, and establish the thermal resistance network topology. The thermal path parameter mapping module is used to extract the thermal conductivity and specific heat capacity parameters of the busbar and the shell, calculate the aggregation weight of each node group in the thermal resistance network topology, generate a regional aggregation feature vector, map the massive voxel parameters into discretized node parameters based on the regional aggregation feature vector, calculate the equivalent thermal resistance and heat capacity between nodes, and generate a compact thermal path model. The flow field dimension calculation module is used to call the compact thermal circuit model, simulate the air flow state inside the distribution box, extract the three-dimensional flow field streamline diagram, cover the three-dimensional flow field streamline diagram with spatial grids of different scales, count the number of non-empty grids covering the streamline trajectory, and calculate the streamline fractal dimension. The heat dissipation failure determination module is used to obtain the baseline dimension value under the standard operating conditions of the distribution box, combine it with the streamline fractal dimension to generate the flow field dimension attenuation coefficient distribution, and determine whether there is dust accumulation and blockage at the air inlet position based on the flow field dimension attenuation coefficient distribution, and generate the heat dissipation structure failure determination result.

[0005] Preferably, the step of obtaining the thermal resistance network topology is as follows: Retrieve the 3D model of the distribution box containing busbars and circuit breaker components, analyze the geometric outer shell surface and internal solid boundary of the 3D model of the distribution box, set the voxel side length parameters for spatial discretization, perform equidistant grid segmentation along the three-axis boundary box of the 3D model of the distribution box, mark the grid units falling into the solid boundary as voxel units, and form a set of voxel units. Traverse the three-dimensional index coordinates of each voxel in the set of voxel units, retrieve adjacent voxel pairs with shared faces or shared edges, generate a heat conduction connection edge for each pair of adjacent voxel units, record the voxel unit number pair and connection direction identifier corresponding to the heat conduction connection edge, and generate a voxelized dual graph. The degree of each voxel unit in the voxelized dual graph is summarized and a degree diagonal matrix is ​​constructed. The thermal conduction connection edges in the voxelized dual graph are summarized and an adjacency matrix is ​​constructed. The difference between the degree diagonal matrix and the adjacency matrix is ​​used to construct a Laplacian matrix. Eigenvalue decomposition is performed on the Laplacian matrix and the eigenvector corresponding to the second smallest eigenvalue is extracted. Based on the numerical distribution of each voxel unit component in the eigenvector, voxel units are clustered and isothermal regions are identified. The cross-regional connection relationships induced by thermal conduction connection edges between isothermal regions are summarized and a connection topology is formed to obtain the thermal resistance network topology.

[0006] Preferably, the step of obtaining the region aggregation feature vector is as follows: Extract the thermal conductivity, specific heat capacity and density parameters of the busbar and the shell, read the node group division results of the thermal resistance network topology, calculate the total volume of voxels of each node group, and calculate the aggregation weight value of each node group based on the material properties. Based on the aggregate weight values, the aggregate weight values ​​corresponding to each node group are read sequentially according to the node group numbering order in the thermal resistance network topology. The aggregate weight values ​​of all node groups are arranged into a one-dimensional numerical sequence according to the numbering order to generate a regional aggregate feature vector.

[0007] Preferably, the steps for obtaining the compact thermal circuit model are as follows: Based on the region aggregation feature vector, the aggregation weight value is used as a weighting factor to calculate the weighted average of the material thermal conductivity and specific heat capacity parameters of all voxel units contained in each node group, so as to obtain the equivalent physical properties of each node group. Combined with the geometric parameters of the connecting edges in the thermal resistance network topology, the equivalent thermal resistance value between nodes and the total heat capacity value of the nodes are calculated using the equivalent physical properties. The heating power load of the circuit breaker node is used as the energy source term, and it is substituted together with the equivalent thermal resistance value and the total heat capacity value of the nodes into the energy conservation differential equation between nodes to generate a compact thermal circuit model.

[0008] Preferably, the steps for obtaining the three-dimensional flow field streamline diagram are as follows: The compact thermal circuit model is invoked, and the boundary flow conditions of the air inlet and the exhaust fan are set. Based on the laws of conservation of mass and momentum, the velocity vector field distribution and pressure distribution of the air flow inside the box are iteratively solved, and the air flow state inside the distribution box is output. Based on the airflow state inside the distribution box, the set of air velocity vector starting points passing through the air inlet area is selected, and the velocity vector trajectory point sequence is generated by segment integration along the air velocity vector direction, and a three-dimensional flow field streamline diagram is rendered.

[0009] Preferably, the step of obtaining the streamline fractal dimension is as follows: The three-dimensional flow field streamline diagram is covered by spatial grids of different scales, and the magnitude of the streamline velocity vector in each spatial grid is statistically analyzed to generate a weighted spatial grid coverage statistical sequence. The streamline fractal dimension is calculated based on the weighted spatial grid coverage statistical sequence.

[0010] Preferably, the step of obtaining the flow field dimension attenuation coefficient distribution is as follows: Obtain the current operating condition parameters of the distribution box and retrieve the standard operating condition reference dimension value that matches the current operating condition parameters from the preset database; The dimension attenuation ratio is obtained by dividing the fractal dimension of the streamline by the reference dimension value of the standard operating condition. At the same time, the starting coordinates of the velocity vector of the streamline are traced back to determine the inlet grid position corresponding to each streamline. The dimension attenuation ratio is mapped to the grid index of all streamline starting points, and the average attenuation degree in each grid area is calculated to generate the flow field dimension attenuation coefficient distribution.

[0011] Preferably, the steps for obtaining the failure determination result of the heat dissipation structure are as follows: Traverse each grid region in the flow field dimension attenuation coefficient distribution, compare the attenuation coefficient in the region with the preset blockage threshold, filter out abnormal grid sets with attenuation coefficients lower than the blockage threshold, locate the actual physical blockage area of ​​the air inlet according to the index coordinates of the abnormal grid sets, and generate a heat dissipation structure failure judgment result.

[0012] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, the three-dimensional model of a distribution box containing busbars and circuit breaker components is spatially discretized, and the thermal conduction connection relationship between voxels is established. This transforms the continuous thermal conduction path into a computable voxelized dual graph structure. Then, the second smallest eigenvalue and its eigenvector are extracted through Laplace matrix eigenvalue decomposition, achieving structural clustering of voxel nodes. This creates a distinguishable equivalent partition structure between the main thermal conduction channel and weakly coupled regions in mathematical space. Based on this, a thermal resistance network topology is constructed, transforming the originally dispersed geometric information into a physically meaningful thermally coupled network structure. Furthermore, the node group aggregation weight is calculated by combining the material's thermal conductivity and specific heat capacity parameters, forming a regional... The system aggregates feature vectors and maps massive voxel parameters to discretized node parameters, compressing three-dimensional field variables into a clearly structured one-dimensional thermal path parameter system. This generates a compact thermal path model while maintaining energy conservation, achieving a stable mapping from high-dimensional thermal fields to low-dimensional network models. Simultaneously, based on the compact thermal path model, the system derives airflow states and extracts three-dimensional streamline diagrams. By statistically analyzing the number of non-empty grids across different scales of spatial grids and calculating the streamline fractal dimension, the geometric distribution characteristics of the flow field are transformed into quantifiable complexity indicators. Combined with baseline dimension values ​​under standard operating conditions, the system generates a flow field dimension decay coefficient distribution, enabling quantitative identification of inlet ash accumulation and blockage. This expands thermal fault prediction from a single temperature criterion to the levels of structural coupling and flow evolution, improving the foresight of thermal anomaly identification while reducing the computational burden caused by model size. Attached Figure Description

[0013] Figure 1 This is a system flowchart of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0015] Please see Figure 1 The present invention provides a technical solution: a distribution box thermal fault prediction system based on a mechanism model includes: The model voxelization module is used to retrieve the 3D model of the distribution box containing busbars and circuit breaker components, perform spatial discretization on the 3D model of the distribution box, establish thermal conduction connection edges between adjacent voxel units, generate a voxelized dual graph, construct the Laplacian matrix corresponding to the voxelized dual graph, perform eigenvalue decomposition on the Laplacian matrix, extract the second smallest eigenvalue of the matrix and the corresponding eigenvector, cluster the voxel nodes according to the eigenvector, and establish the thermal resistance network topology. The thermal path parameter mapping module is used to extract the thermal conductivity and specific heat capacity parameters of the busbar and shell materials, calculate the aggregation weight of each node group in the thermal resistance network topology, generate regional aggregation feature vectors, map massive voxel parameters into discretized node parameters based on regional aggregation feature vectors, calculate the equivalent thermal resistance and heat capacity values ​​between nodes, and generate a compact thermal path model. The flow field dimension calculation module is used to call the compact thermal circuit model to simulate the air flow state inside the distribution box, extract the three-dimensional flow field streamline map, cover the three-dimensional flow field streamline map with spatial grids of different scales, count the number of non-empty grids covering the streamline trajectory, and calculate the streamline fractal dimension. The heat dissipation failure determination module is used to obtain the baseline dimension value under the standard operating conditions of the distribution box, combine it with the streamline fractal dimension to generate the flow field dimension attenuation coefficient distribution, and based on the flow field dimension attenuation coefficient distribution, determine whether there is dust accumulation and blockage at the air inlet position, and generate the heat dissipation structure failure determination result.

[0016] The steps to obtain the thermal resistance network topology are as follows: Retrieve the 3D model of the distribution box containing busbars and circuit breaker components, analyze the geometric outer shell surface and internal solid boundary of the 3D model of the distribution box, set the voxel side length parameters for spatial discretization, perform equidistant grid segmentation along the three-axis boundary box of the 3D model of the distribution box, mark the grid units falling into the solid boundary as voxel units, and form a set of voxel units. Traverse the 3D index coordinates of each voxel in the voxel set, retrieve adjacent voxel pairs with shared faces or shared edges, generate a heat conduction connection edge for each pair of adjacent voxel pairs, record the voxel number pair and connection direction identifier corresponding to the heat conduction connection edge, and generate a voxelized dual graph. The degree of each voxel unit in the voxelized dual graph is summarized and a degree diagonal matrix is ​​constructed. The thermal conduction edges in the voxelized dual graph are summarized and an adjacency matrix is ​​constructed. The difference between the degree diagonal matrix and the adjacency matrix is ​​used to construct a Laplacian matrix. Eigenvalue decomposition is performed on the Laplacian matrix and the eigenvectors corresponding to the second smallest eigenvalue are extracted. Based on the numerical distribution of each voxel unit component in the eigenvector, voxel units are clustered and isothermal regions are identified. The cross-regional connection relationships induced by thermal conduction edges between isothermal regions are summarized and a connection topology is formed to obtain the thermal resistance network topology.

[0017] Specifically, the process involves retrieving a 3D model of the distribution box containing busbars and circuit breaker components. By reading geometric data files in STEP or IGES format, the geometric outer surface patches and internal solid boundaries of the 3D model are analyzed. The coordinate data of all geometric vertices is traversed to determine the extreme coordinates of the overall model in 3D space, and the length, width, and height of the bounding box are calculated. For example, the minimum X-axis coordinate is obtained. with the maximum coordinates minimum Y-axis coordinate with the maximum coordinates and the minimum coordinate of the Z-axis with the maximum coordinates Set the voxel side length parameter for spatial discretization, for example, set the voxel side length... The value is set to a fixed value between 0.005 meters and 0.01 meters, based on the minimum thickness of the busbar. This ensures that at least two voxel units are included in the thickness direction to guarantee the accuracy of heat conduction calculations. Equally spaced grid division is performed along the three-axis boundary frame of the distribution box 3D model, using a triple-loop structure from... to , to and to Iterate through all spatial coordinate points within the bounding box and calculate the center coordinates of each grid cell. The method of ray intersection is used to determine whether the center coordinates are inside the solid boundary. That is, a ray is emitted from the center point in any direction, and the number of intersections between the ray and the triangular facets on the model surface is counted. If the number of intersections is odd, it is determined to be inside. The grid cells that fall inside the solid boundary are marked as voxel cells, and all marked voxel cells are assigned a unique integer index number, for example, starting from 1 and incrementing. At the same time, the material property identifier of each voxel cell is recorded to form a voxel cell set.

[0018] Traverse the 3D index coordinates of each voxel in the voxel set, and establish a six-neighbor search rule determined by spatial positional relationships. That is, for a voxel with a spatial index of... Given the current voxel, retrieve the coordinates of its neighbors in the positive and negative directions of the X, Y, and Z axes, respectively. , and The process involves determining whether the coordinates of these neighbors exist in the valid index list of the voxel unit set, retrieving adjacent voxel unit pairs with shared faces or shared edges, and if neighboring voxels exist, it indicates that the two are in direct physical contact and have the conditions for heat conduction. For each pair of adjacent voxel unit pairs, a heat conduction connection edge is generated, and a unique edge ID is assigned to each connection edge. The voxel unit number pair corresponding to the heat conduction connection edge is recorded, for example, recorded as including the starting node. and termination node tuple The connection direction identifier is determined based on the relative position of the two, such as "positive X-axis" or "negative Z-axis". All connection edge information is stored in the edge list, which serves as the basic data structure for subsequent construction of sparse matrices, generating a voxelized dual graph.

[0019] The connection degree of each voxel unit in the voxelized dual graph is summarized and a degree diagonal matrix is ​​constructed. The degree is obtained by counting the number of connected edges associated with each voxel unit. ,in Number the voxels and construct the diagonal elements as A degree matrix in which all other elements are 0. Summarize the heat conduction connection edges in the voxelized dual graph and construct the adjacency matrix. If the voxels... With voxels If there are connecting edges, then set the adjacency matrix elements. Otherwise, it is 0. The Laplacian matrix is ​​constructed using the difference between the degree diagonal matrix and the adjacency matrix. The calculation formula is as follows: ,in Represents the Laplace matrix, Represents the degree matrix. Representing the adjacency matrix, perform eigenvalue decomposition on the Laplacian matrix and extract the eigenvector corresponding to the second smallest eigenvalue, then solve the characteristic equation. ,in It is an eigenvector. The eigenvalues ​​are selected as eigenvectors. The eigenvalues ​​whose numerical values ​​are second only to zero after sorting (i.e., Fiedler values) are chosen as the Fiedler vectors. Based on the numerical distribution of each voxel unit component in the eigenvector, voxel units are clustered and isothermal regions are identified. A clustering threshold is set. For example, take The one-dimensional K-Means clustering algorithm is used to cluster feature vector components with numerical differences less than 1. Voxels are grouped into the same category, and these voxels are considered to be tightly bound in terms of thermal conduction, thus forming a node group with a uniform internal temperature. The cross-regional connection relationships induced by thermal conduction connection edges between isothermal regions are summarized and a connection topology is formed. All original connection edges are traversed. If the two ends of the connection edge belong to different isothermal region clusters, a macroscopic thermal resistance connection is established between the two regions, resulting in a thermal resistance network topology.

[0020] The steps for obtaining the region aggregation feature vector are as follows: Extract the thermal conductivity, specific heat capacity, and density parameters of the busbar and shell materials. Read the node group division results of the thermal resistance network topology, calculate the total voxel volume of each node group, and calculate the aggregation weight value of each node group based on material properties. The calculation formula is as follows: ; in, For the first The aggregate weight value of each node group. For the first The volume-weighted average of the thermal conductivity of all voxel elements within a node group is used to characterize the equivalent thermal conductivity of that region. For the first The total volume of voxels in a group of nodes For the first The material density corresponding to each node group For the first The specific heat capacity of the material corresponding to each node group, product term Characterizing the first The total heat capacity of the node group This represents the number of nodes in the thermal resistance network topology. The summation index variable takes values ​​ranging from 1 to... ; Based on the aggregate weight values, the aggregate weight values ​​corresponding to each node group are read sequentially according to the node group numbering order in the thermal resistance network topology. The aggregate weight values ​​of all node groups are arranged into a one-dimensional numerical sequence according to the numbering order to generate a regional aggregate feature vector.

[0021] Specifically, in the formula for calculating the aggregation weight, the importance of high thermal conductivity materials (such as copper busbars) in the model dimensionality reduction process is amplified at the numerical level by squaring the thermal conductivity. At the same time, the heat capacity term composed of volume, density and specific heat capacity is retained, ensuring that the final aggregate weight can reflect both the heat transfer efficiency of the region and the heat storage inertia of the region. For the first The volume-weighted average of the thermal conductivity of all voxel elements within a node group is used to characterize the equivalent thermal conductivity of that region. This parameter is obtained by iterating through the nth node group. Each node group index list contains all voxel unit numbers. Based on the material property ID marked by each voxel unit, the corresponding standard thermal conductivity is retrieved from a pre-set engineering material thermophysical property database. For example, the value is 400 W / (m·K) for copper, 50 W / (m·K) for steel, and 0.026 W / (m·K) for insulating air. If the node group contains multiple media, a weighted calculation is performed based on the proportion of each medium's voxel count to the total number of voxels in the node group. For example, if copper voxels account for 80% and air voxels account for 20% in the node group, the formula is used to calculate the standard thermal conductivity. The average thermal conductivity of this region was calculated. For the first The total volume of voxels in a group of nodes is obtained by calculating the voxel volume of the nth node group. The total number of valid voxel units contained in the node clustering results It also calls the voxel side length parameters set in the spatial discretization stage of the preceding steps. This side length parameter The value is usually set based on the dimensions of the smallest solid component inside the distribution box (such as the thickness of the busbar), for example, a value of 0.01 meters, using the cubic volume calculation formula. Calculate the geometric volume of a single voxel, and then calculate the total number of voxels. Multiply by the volume of a single voxel, i.e. This parameter quantifies the actual physical space occupied by a thermal node in three-dimensional space; For the first The material density corresponding to each node group is obtained by determining the density of the first node group based on the 3D design BOM or material property list of the distribution box. The material type of the physical component represented by each node group is determined by extracting the corresponding density value from the physical constant table, with the unit being kg / m³. 3 For example, for a busbar node group made of pure copper, the density of copper is extracted to be 8900 kg / m³. 3 For the shell node group made of carbon steel, the extracted steel density is 7850 kg / m³. 3 For a group of nodes containing mixed components, the equivalent density is calculated using the volume average method, which calculates the sum of the masses of each component and divides it by the total volume. This parameter determines the mass contained in a substance per unit volume. For the first The specific heat capacity of the material corresponding to each node group is obtained by consulting the standard material thermodynamic property table. The specific heat capacity of the material corresponding to each node group at the operating temperature is expressed in J / (kg·K). This parameter represents the amount of heat that a unit mass of material needs to absorb to raise its temperature by 1 Kelvin. For example, the specific heat capacity of copper is 390 J / (kg·K) and the specific heat capacity of air is 1005 J / (kg·K). By multiplying the specific heat capacity by the density and volume, the heat capacity of the node group can be accurately calculated. is the number of node groups in the thermal resistance network topology. This parameter comes from the total number of clusters generated after clustering the Laplacian matrix eigenvectors in the previous step, and is a positive integer. The summation index variable is used to traverse all node groups to calculate the normalized denominator; Calculations based on parameters: The area is divided into There are two node groups, where node group 1 is the core area of ​​the copper busbar and node group 2 is the mixed area of ​​surrounding air and insulation support.

[0022] Get basic parameters: Node group 1 (copper): number of voxels voxel side length thermal conductivity ,density Specific heat capacity ; Node group 2 (mixed): number of voxels voxel side length Weighted thermal conductivity Weighted density Weighted specific heat capacity ; Calculate intermediate variables: Calculate volume : ; ; Calculate the heat capacity product term : ; ; Calculate the numerator : ; ; Calculate the sum of the denominators : ; Calculate aggregate weights : ; ; The results indicate that although the volume of node group 2 is 5 times that of node group 1 and its total heat capacity is approximately twice that of node group 1, the thermal conductivity of node group 1 (copper busbar) is much higher than that of node group 2. Under the amplification effect of the square operation, It obtained a weight close to 1, which forcibly established the core position of the busbar in the thermal fault prediction model at the numerical level. This means that the compact model built later will mainly evolve around the thermal property changes of high thermal conductivity components, while ignoring the subtle disturbances in low thermal conductivity regions, thereby capturing the dynamic characteristics of the main heat source while ensuring computational efficiency.

[0023] Based on the aggregation weight value, create a group with a length equal to the total number of nodes. Using equal floating-point arrays as storage containers, and following the node group index order determined during the thermal resistance network topology generation stage, a loop control structure is used to traverse from index 1 to... Read the aggregate weights calculated for each node group one by one. During the reading process, a numerical validity check is performed, that is, a determination is made. Whether it is a non-negative and non-null value. If an abnormal null value is encountered, the weights of the neighboring node groups in the spatial topology of this node group are retrieved, and their average value is used to fill the null value. For example, if If missing, search for its direct neighbors. and ,calculate As an alternative value, to ensure the integrity of the data chain, the aggregated weight values ​​of all verified node groups are sequentially filled into an array according to their numbers, forming a one-dimensional numerical sequence. Then, normalization verification is performed on this sequence, and the values ​​of all elements in the sequence are accumulated. It is determined whether the sum is strictly equal to 1. If there is a deviation caused by floating-point operation error (e.g., the sum is 0.999999), each element in the sequence is divided by the current accumulated sum to complete the rescaling of the values ​​and generate a regional aggregated feature vector. This vector essentially constitutes a feature distribution map measured by the thermal importance of nodes. The value directly corresponds to the retention priority of each area in the distribution box in the thermal circuit simplification model, providing a quantitative weighting basis for parameter mapping in subsequent steps.

[0024] The steps to obtain the compact thermal circuit model are as follows: Based on the regional aggregation feature vector, the aggregation weight value is used as a weighting factor to calculate the weighted average of the material thermal conductivity and specific heat capacity parameters of all voxel units contained in each node group, so as to obtain the equivalent physical properties of each node group. Combined with the geometric parameters of the connecting edges in the thermal resistance network topology, the equivalent thermal resistance value between nodes and the total heat capacity value of the nodes are calculated using the equivalent physical properties. The heating power load of the circuit breaker node is used as the energy source term, and it is substituted together with the equivalent thermal resistance value and the total heat capacity value of the nodes into the energy conservation differential equation between nodes to generate a compact thermal circuit model.

[0025] Specifically, based on the region aggregation feature vector, the weight ratio corresponding to each element in the vector is extracted. The aggregation weight value is used as the core weighting factor to calculate the weighted average of the material thermal conductivity and specific heat capacity parameters of all voxel units contained in each node group. The specific process is as follows: for the th For a group of nodes, the equivalent physical properties of the node group are obtained by summing the products of the thermal conductivity of all voxels within the group and their corresponding weights, and then dividing by the sum of the weights. This includes the equivalent thermal conductivity. Equivalent specific heat capacity and equivalent density ,in This represents the average mass density of the node group at the macroscopic level. It is calculated by taking a volume-weighted average of the material densities of each voxel within the node group. , here The total volume of the node group Using the density of a single voxel and the geometric parameters of the connecting edges in the thermal resistance network topology, the equivalent thermal resistance between nodes and the total heat capacity of the nodes are calculated using equivalent physical properties. Based on the discrete form of Fourier's law of heat conduction, the node... With nodes The thermal resistance between them is calculated using the following formula: ,in Representative node group Geometric centroid and node group The Euclidean distance between the geometric centroids, in meters. This represents the common contact area between two node groups at the contact interface, expressed in square meters. This area is obtained by calculating the total area of ​​all shared voxel surfaces between the two node groups. The equivalent thermal conductivity of the interface between two nodal groups is typically calculated using the harmonic averaging method. This reflects the actual impedance of heat flow across the interface, while also calculating the nodes. The total heat capacity is calculated using the following formula: ,in For nodes The equivalent specific heat capacity, For nodes The equivalent density, For nodes The total volume is calculated by taking the heat load of the circuit breaker node as the energy source and collecting real-time current data from the circuit breaker. and preset contact resistance (For example, set to) Using Joule's law Real-time thermal power was calculated (For non-fever nodes) Substituting the equivalent thermal resistance and the total heat capacity of each node into the energy conservation differential equation between nodes, we can apply this equation to each node in the network. Constructing a form like The heat balance equations, in which For nodes instantaneous temperature, For time variables, Adjacent nodes instantaneous temperature, The term represents the sum of heat fluxes from all adjacent nodes, generating a compact thermal path model.

[0026] The steps for obtaining a 3D flow field streamline diagram are as follows: The compact thermal circuit model is invoked, and the boundary flow conditions of the air inlet and the exhaust fan are set. Based on the laws of mass conservation and momentum conservation, the velocity vector field distribution and pressure distribution of the air flow inside the box are solved iteratively, and the air flow state inside the distribution box is output. Based on the airflow state inside the distribution box, the set of air velocity vector starting points passing through the air inlet area is selected, and the velocity vector trajectory point sequence is generated by segment integration along the direction of the air velocity vector, and a three-dimensional flow field streamline diagram is rendered.

[0027] Specifically, the compact thermal circuit model is invoked, the fluid domain geometric boundary information defined in the model is read, and the inlet boundary flow conditions and exhaust fan boundary flow conditions are set. The specific values ​​of these flow conditions are derived from the cooling fan's product specifications or actual wind tunnel test data. For example, for a certain model of distribution box, the inlet volumetric flow rate is set. The flow rate is set to 0.05 cubic meters per second, corresponding to the fan's rated airflow under operating static pressure. The exhaust outlet is defined as a free outflow boundary, i.e., a relative pressure of 0 Pascals. The initial flow field velocity is set to 0 meters per second. Based on the laws of mass and momentum conservation, the finite volume method is used to discretize the governing equations. The continuity equation and the Navier-Stokes equations are solved simultaneously. During the solution process, the SIMPLE algorithm (semi-implicit pressure correlation equation algorithm) is used for coupled iteration of pressure and velocity, and a convergence residual threshold is set. for ,in This represents the maximum permissible deviation of physical quantities between two consecutive iterative calculations, used to determine whether the calculation process has reached a steady state. Iteratively solves for the velocity vector field distribution and pressure distribution of airflow within the box, recording the velocity components at each discretized grid node. and pressure value It outputs the airflow status inside the distribution box. This status data includes the fluid motion parameters of all three-dimensional spatial points from the air inlet to the air outlet, providing complete basic flow field data for subsequent streamline tracing and heat dissipation analysis.

[0028] The steps to obtain the streamline fractal dimension are as follows: Three-dimensional flow field streamline diagrams are covered by spatial grids of different scales, and the magnitude of streamline velocity vectors in each spatial grid is statistically analyzed to generate a weighted spatial grid coverage statistical sequence. Specifically, based on the airflow state inside the distribution box, the planar grid coordinates of the air inlet are read, and streamline seed points are evenly distributed in the air inlet area. For example, a 10x10 seed point array is set, resulting in 100 starting coordinates. The set of air velocity vector starting points passing through the air inlet area is selected, and a sequence of velocity vector trajectory points is generated by segment-by-segment integration along the direction of the air velocity vector. The trajectory is derived using the fourth-order Runge-Kutta integral method, and the formula is used... Calculate the spatial position at the next moment, where This represents the three-dimensional spatial coordinates of the streamline trajectory point obtained at the next time step, in meters. This represents the three-dimensional spatial coordinates of the streamline trajectory point at the current moment, in meters. Represents the air velocity vector at the current coordinate point, including The components in the three directions, in meters per second, are derived from the flow field calculation results. This represents the integration time step, in seconds, for example, set to 0.001 seconds. This step size determines the time resolution of trajectory tracking. Tracking continues until the trajectory touches the box wall or reaches the maximum integration step count (e.g., 5000 steps). All calculated discrete trajectory points are connected sequentially to render a 3D flow field streamline map. Spatial grids of different scales are used to cover the 3D flow field streamline map. First, the 3D bounding box size of the flow field region is determined, and the basic grid scale parameters are set, for example, the side length of the first-level grid is 0.005 meters, and the side length of the second-level grid is 0.01 meters. A 3D Cartesian grid system is constructed respectively. Each generated streamline trajectory is traversed, the grid cell index through which the streamline passes is identified, and the magnitude of the streamline velocity vector in each spatial grid is counted. For grids containing streamline trajectories, the average velocity magnitude of all trajectory points in the grid is calculated. If no streamline passes through the grid, it is marked as empty. The velocity magnitude values ​​of each non-empty grid are associated with the grid index and stored to generate a weighted spatial grid coverage statistical sequence. This sequence records in detail the momentum distribution characteristics of the flow field at different spatial resolutions.

[0029] Based on the weighted spatial grid coverage statistical sequence, the streamline fractal dimension is calculated using the following formula: ; in, The streamline fractal dimension, and These are the grid scale values ​​for the first and second groups of spatial grids, respectively. , and They are respectively using grid scale and The momentum-weighted coverage at time t, the formula for calculating the momentum-weighted coverage is: , The total number of non-empty grid cells through which the streamline passes. For the first The average magnitude of streamline velocity vectors within a non-empty grid. This represents the maximum velocity threshold in the flow field.

[0030] Specifically, the formula for calculating the fractal dimension of streamlines utilizes the concept of fractal dimension to quantify the space-filling capacity and turbulence complexity of the cooling airflow inside the distribution box. The dimension is calculated by comparing the momentum coverage of the flow field at different observation scales (grid sizes). A higher fractal dimension can reflect whether the cooling airflow effectively fills the complex structural gaps inside the enclosure. A higher fractal dimension usually means that the airflow is more evenly distributed in space and covers fewer dead corners, thus having better global heat dissipation potential. A lower dimension, on the other hand, suggests that there is obvious airflow short circuit or a large area of ​​heat dissipation blind spot. and These are the grid scale values ​​for the first and second groups of spatial grids, respectively, in meters. The steps to obtain these two parameters are as follows: First, set the reference resolution based on the overall dimensions (length, width, and height) of the fluid domain of the distribution box. Typically, 1 / 2 to 1 / 5 of the smallest characteristic dimension of the fluid domain (such as the diameter of the heat dissipation holes) is selected as the small scale. ,Pick Two to four times as large scale For example, in actual setups, based on a heat dissipation channel width of 0.05 meters, the following settings are configured: The value is 0.005 meters, set as follows: The two scales, 0.01 meters and 0.01 meters respectively, must be fine enough to capture the details of the flow field, while maintaining a certain magnification difference to ensure the effectiveness of fractal feature calculations. and They are respectively using grid scale and The momentum-weighted coverage at time, a dimensionless parameter, is obtained by traversing all non-empty grids and reading the average velocity within each grid, based on the weighted spatial grid coverage statistical sequence generated in the previous step. Combined with the maximum velocity across the entire flow field Using the formula Perform calculations; The total number of non-empty grid cells through which streamlines pass is obtained by counting the number of grid cells that contain at least one streamline trajectory point at a specific grid scale, and is a positive integer. This is a summation index variable used to iterate through all non-empty grid cells, with values ​​ranging from 1 to... ; For the first The average magnitude of the streamline velocity vector within a non-empty grid, in meters per second, is obtained by identifying the area falling within the first non-empty grid. For all streamline trajectory points within a grid, extract the velocity vector magnitude of these points and calculate the average value; The maximum velocity threshold in the flow field is measured in meters per second. This parameter is obtained by scanning the velocity data in the entire flow field solution domain and selecting the maximum modulus value as the normalization benchmark, for example, 5 meters per second. Calculations based on parameters: Set the maximum velocity of the flow field .

[0031] For small-scale grids : The number of non-empty grid cells was obtained statistically. indivual; For example, 20 of the grids are high-speed zones (average speed) ), 30 grids are in the medium speed zone (average speed) ); Calculate coverage : Contribution of the high-speed zone: ; Mid-speed zone contribution: ; ; For large-scale grids : Due to grid merging, the number of non-empty grids decreased, and statistics showed... indivual; For example, the average velocity of these large grids, after smoothing, is approximately... ; Calculate coverage : ; Calculate fractal dimension : Calculate the logarithmic difference (numerator): ; ; molecular ; Calculate the logarithmic difference (denominator): ; ; denominator ; Calculate the final value: ; The results show that the calculated streamline fractal dimension is 2.61, which is close to 3. This indicates that the cooling airflow in the distribution box has a high space filling rate and a complex turbulent structure, which can cover various areas in three-dimensional space well. This suggests that the current heat dissipation structure design is highly efficient in using airflow for heat exchange. If the value is low (such as close to 1.5), it means that the airflow is mainly concentrated in a single channel, and there are obvious heat dissipation dead zones.

[0032] The steps for obtaining the distribution of the flow field dimension attenuation coefficient are as follows: Obtain the current operating condition parameters of the distribution box and retrieve the standard operating condition reference dimension value that matches the current operating condition parameters from the preset database; The dimension attenuation ratio is obtained by dividing the fractal dimension of the streamline by the reference dimension value under standard operating conditions. At the same time, the starting coordinates of the velocity vector of the streamline are traced back to determine the inlet grid position corresponding to each streamline. The dimension attenuation ratio is mapped to the grid index of all streamline starting points, and the average attenuation degree in each grid region is calculated to generate the flow field dimension attenuation coefficient distribution.

[0033] Specifically, the current operating parameters of the distribution box are obtained, and high-precision sensing devices are deployed at key electrical nodes and environmental monitoring points within the distribution box. For example, Hall effect current sensors are installed at the circuit breaker input to collect real-time load current values ​​(e.g., 0A to 100A range), and thermocouple temperature sensors are deployed at the top exhaust port and bottom air inlet of the box to obtain inlet and outlet temperature difference data. Simultaneously, the PWM control signal or speed feedback signal of the cooling fan is read. The above analog signals are converted into digital signals and encapsulated into an operating condition feature vector using a data acquisition card. ,in For load current, and The inlet and outlet air temperatures, To determine the fan speed, a baseline dimension value matching the current operating condition parameters is retrieved from a pre-defined database. This database is built based on large-scale offline fluid simulation experiments, pre-setting a series of standard operating conditions. For example, current steps of 10A cover the range from 0A to rated current, and ambient temperature steps of 5℃ cover the range from -10℃ to 50℃. Ideal heat dissipation simulations without dust accumulation or blockages are performed for each combination of operating conditions. The corresponding streamline fractal dimension is calculated and recorded, forming a key-value pair storage structure. During retrieval, an Euclidean distance minimization algorithm is used to find the corresponding value in the database. If no perfectly matching working condition is found among the nearest standard working points, the four closest sample points are selected, and the theoretical baseline dimension value under the current working condition is calculated using multilinear interpolation. This value represents the theoretical complexity of the flow field when the system is in a fully healthy state under the current load and heat dissipation conditions. For example, when the load is 45A and the ambient temperature is 30℃, the benchmark dimension obtained by interpolation may be 2.75, which can be used as a benchmark for subsequent evaluation of flow field degradation.

[0034] The dimension decay ratio is obtained by dividing the streamline fractal dimension by the standard operating condition reference dimension value, which is then read from the real-time calculated streamline fractal dimension. (e.g., 2.45), compared with the baseline dimension value obtained in the previous steps. (For example, 2.80) is compared to calculate the attenuation ratio. This ratio directly reflects the degree to which the overall quality of the current flow field is preserved relative to the ideal state. For example, if the calculated result is 0.875, it can be calculated by traversing back to the starting coordinates of the velocity vectors that generated the streamlines, iterating through all valid streamline objects generated in the current flow field, and extracting the three-dimensional spatial coordinates of each streamline at the start of the integration. Since the streamlines are all triggered from the inlet boundary, these coordinate points are mainly distributed within the two-dimensional plane region where the inlet is located. The inlet grid position corresponding to each streamline is determined, and the physical plane of the inlet is divided into... A regular grid matrix, for example, dividing a 0.2-meter multi-meter air inlet into... The grid cells are each 0.01 meters on each side, based on the starting coordinates. Determine the row and column indices of the grid it falls into. Establish a mapping list between grid indices and streamline origins, mapping the dimension decay ratio to the grid indices of all streamline origins. For each non-empty grid with streamline origins, assign the same dimension decay ratio value to all streamline origins contained within it. For blank grids where no streamlines are generated due to blockage, no numerical mapping is performed or the values ​​are marked as invalid. The average attenuation level within each grid region is calculated. All grid cells are traversed, and if a grid contains a streamline origin, the average value of all mapped values ​​within that grid is calculated (i.e., the mean value). If there is no streamline origin within the grid (i.e., the area does not participate in the effective heat dissipation cycle), the attenuation degree of the grid is forcibly assigned to 0. This distinguishes between normal flow areas and completely blocked areas, generating a flow field dimension attenuation coefficient distribution. This distribution is ultimately represented as a two-dimensional numerical matrix. Areas with higher values ​​(such as above 0.8) in the matrix represent airflow smoothly entering from that location and forming a complex heat dissipation flow field, while areas with values ​​of 0 or extremely low values ​​accurately depict the physical outline of the air inlet being covered by dust or foreign objects.

[0035] The steps for obtaining the failure determination result of the heat dissipation structure are as follows: Traverse each grid region in the flow field dimension attenuation coefficient distribution, compare the attenuation coefficient in the region with the preset blockage threshold, filter out abnormal grid sets with attenuation coefficients lower than the blockage threshold, locate the actual physical blockage area of ​​the air inlet based on the index coordinates of the abnormal grid sets, and generate the heat dissipation structure failure judgment result.

[0036] Specifically, it iterates through each grid region in the distribution of the flow field dimension attenuation coefficient, and uses a double loop structure to scan the value of each element in the two-dimensional matrix. The attenuation coefficient within the region is compared with a preset clogging threshold. The settings are based on critical heat dissipation effectiveness tests. The setting process is as follows: Different degrees of air inlet obstruction are simulated in a laboratory environment. When the obstruction area reaches 30%, it is found that the temperature rise of key components inside the distribution box exceeds the safety margin. At this time, the corresponding streamline fractal dimension attenuation ratio relative to the standard value usually drops to about 0.6. Considering the noise interference of on-site measurements, a safety threshold is set. This means that areas with an attenuation coefficient below 0.55 are considered to have lost their effective air intake capacity. An abnormal set of grids with an attenuation coefficient below the blockage threshold is selected, and during the traversal, all grids meeting the condition are... Grid index Add to the anomaly list, locate the actual physical blockage area of ​​the air inlet based on the index coordinates of the anomaly grid set, and use the spatial transformation matrix during grid division to index the grid rows and columns. The reverse projection is the physical coordinate range of the air inlet plane. For example, the grid (0,0) to (5,5) corresponds to the rectangular area of ​​0 to 5 cm in the lower left corner of the air inlet. By aggregating the physical coordinates of all abnormal grids, the shape and location of the actual blockage are delineated, generating the failure judgment result of the heat dissipation structure. The output includes a description of the blockage location (such as "severe blockage in the lower left area of ​​the air inlet") and the proportion of the blockage area, triggering the corresponding maintenance warning signal.

Claims

1. A distribution box thermal fault prediction system based on a mechanistic model, characterized in that, The system includes: The model voxelization module is used to retrieve the 3D model of the distribution box containing busbars and circuit breaker components, perform spatial discretization on the 3D model of the distribution box, establish thermal conduction connection edges between adjacent voxel units, generate a voxelized dual graph, construct the Laplacian matrix corresponding to the voxelized dual graph, perform eigenvalue decomposition on the Laplacian matrix, extract the second smallest eigenvalue of the matrix and the corresponding eigenvector, cluster the voxel nodes according to the eigenvector, and establish the thermal resistance network topology. The thermal path parameter mapping module is used to extract the thermal conductivity and specific heat capacity parameters of the busbar and the shell, calculate the aggregation weight of each node group in the thermal resistance network topology, generate a regional aggregation feature vector, map the massive voxel parameters into discretized node parameters based on the regional aggregation feature vector, calculate the equivalent thermal resistance and heat capacity between nodes, and generate a compact thermal path model. The flow field dimension calculation module is used to call the compact thermal circuit model, simulate the air flow state inside the distribution box, extract the three-dimensional flow field streamline diagram, cover the three-dimensional flow field streamline diagram with spatial grids of different scales, count the number of non-empty grids covering the streamline trajectory, and calculate the streamline fractal dimension. The heat dissipation failure determination module is used to obtain the baseline dimension value under the standard operating conditions of the distribution box, combine it with the streamline fractal dimension to generate the flow field dimension attenuation coefficient distribution, and determine whether there is dust accumulation and blockage at the air inlet position based on the flow field dimension attenuation coefficient distribution, and generate the heat dissipation structure failure determination result.

2. The distribution box thermal fault prediction system based on a mechanistic model according to claim 1, characterized in that, The steps for obtaining the thermal resistance network topology are as follows: Retrieve the 3D model of the distribution box containing busbars and circuit breaker components, analyze the geometric outer shell surface and internal solid boundary of the 3D model of the distribution box, set the voxel side length parameters for spatial discretization, perform equidistant grid segmentation along the three-axis boundary box of the 3D model of the distribution box, mark the grid units falling into the solid boundary as voxel units, and form a set of voxel units. Traverse the three-dimensional index coordinates of each voxel in the set of voxel units, retrieve adjacent voxel pairs with shared faces or shared edges, generate a heat conduction connection edge for each pair of adjacent voxel units, record the voxel unit number pair and connection direction identifier corresponding to the heat conduction connection edge, and generate a voxelized dual graph. The degree of each voxel unit in the voxelized dual graph is summarized and a degree diagonal matrix is ​​constructed. The thermal conduction connection edges in the voxelized dual graph are summarized and an adjacency matrix is ​​constructed. The difference between the degree diagonal matrix and the adjacency matrix is ​​used to construct a Laplacian matrix. Eigenvalue decomposition is performed on the Laplacian matrix and the eigenvector corresponding to the second smallest eigenvalue is extracted. Based on the numerical distribution of each voxel unit component in the eigenvector, voxel units are clustered and isothermal regions are identified. The cross-regional connection relationships induced by thermal conduction connection edges between isothermal regions are summarized and a connection topology is formed to obtain the thermal resistance network topology.

3. The distribution box thermal fault prediction system based on a mechanistic model according to claim 1, characterized in that, The steps for obtaining the region aggregation feature vector are as follows: Extract the thermal conductivity, specific heat capacity and density parameters of the busbar and the shell, read the node group division results of the thermal resistance network topology, calculate the total volume of voxels of each node group, and calculate the aggregation weight value of each node group based on the material properties. Based on the aggregate weight values, the aggregate weight values ​​corresponding to each node group are read sequentially according to the node group numbering order in the thermal resistance network topology. The aggregate weight values ​​of all node groups are arranged into a one-dimensional numerical sequence according to the numbering order to generate a regional aggregate feature vector.

4. The distribution box thermal fault prediction system based on a mechanistic model according to claim 1, characterized in that, The steps for obtaining the compact thermal circuit model are as follows: Based on the region aggregation feature vector, the aggregation weight value is used as a weighting factor to calculate the weighted average of the material thermal conductivity and specific heat capacity parameters of all voxel units contained in each node group, so as to obtain the equivalent physical properties of each node group. Combined with the geometric parameters of the connecting edges in the thermal resistance network topology, the equivalent thermal resistance value between nodes and the total heat capacity value of the nodes are calculated using the equivalent physical properties. The heating power load of the circuit breaker node is used as the energy source term, and it is substituted together with the equivalent thermal resistance value and the total heat capacity value of the nodes into the energy conservation differential equation between nodes to generate a compact thermal circuit model.

5. The distribution box thermal fault prediction system based on a mechanistic model according to claim 1, characterized in that, The steps for obtaining the three-dimensional flow field streamline diagram are as follows: The compact thermal circuit model is invoked, and the boundary flow conditions of the air inlet and the exhaust fan are set. Based on the laws of conservation of mass and momentum, the velocity vector field distribution and pressure distribution of the air flow inside the box are iteratively solved, and the air flow state inside the distribution box is output. Based on the airflow state inside the distribution box, the set of air velocity vector starting points passing through the air inlet area is selected, and the velocity vector trajectory point sequence is generated by segment integration along the air velocity vector direction, and a three-dimensional flow field streamline diagram is rendered.

6. The distribution box thermal fault prediction system based on a mechanistic model according to claim 1, characterized in that, The steps for obtaining the streamline fractal dimension are as follows: The three-dimensional flow field streamline diagram is covered by spatial grids of different scales, and the magnitude of the streamline velocity vector in each spatial grid is statistically analyzed to generate a weighted spatial grid coverage statistical sequence. The streamline fractal dimension is calculated based on the weighted spatial grid coverage statistical sequence.

7. The distribution box thermal fault prediction system based on a mechanistic model according to claim 1, characterized in that, The steps for obtaining the distribution of the flow field dimension attenuation coefficient are as follows: Obtain the current operating condition parameters of the distribution box and retrieve the standard operating condition reference dimension value that matches the current operating condition parameters from the preset database; The dimension attenuation ratio is obtained by dividing the fractal dimension of the streamline by the reference dimension value of the standard operating condition. At the same time, the starting coordinates of the velocity vector of the streamline are traced back to determine the inlet grid position corresponding to each streamline. The dimension attenuation ratio is mapped to the grid index of all streamline starting points, and the average attenuation degree in each grid area is calculated to generate the flow field dimension attenuation coefficient distribution.

8. The distribution box thermal fault prediction system based on a mechanistic model according to claim 1, characterized in that, The steps for obtaining the failure determination result of the heat dissipation structure are as follows: Traverse each grid region in the flow field dimension attenuation coefficient distribution, compare the attenuation coefficient in the region with the preset blockage threshold, filter out abnormal grid sets with attenuation coefficients lower than the blockage threshold, locate the actual physical blockage area of ​​the air inlet according to the index coordinates of the abnormal grid sets, and generate a heat dissipation structure failure judgment result.

Images