A buried aluminum block-based PCB transverse heat conduction structure optimization method

Abstract

The application discloses a PCB transverse heat conduction structure optimization method based on embedded aluminum blocks, and belongs to the technical field of PCB heat dissipation design. The method comprises the following steps: obtaining initial thermal distribution data and power device layout information of a hybrid substrate, performing heat flow vector field reconstruction processing on the initial thermal distribution data to generate a heat flow density distribution graph; identifying a transverse heat conduction bottleneck area according to the heat flow density distribution graph; determining an aluminum block embedding position and aluminum block shape parameters based on heat flow vector divergence calculation of the transverse heat conduction bottleneck area; and generating a hybrid substrate transverse heat conduction structure optimization scheme according to the aluminum block embedding position and shape parameters. The method accurately locates the heat conduction bottleneck through heat flow vector field analysis, and adopts a parameterized shape optimization means to realize accurate matching of the aluminum block layout and the heat conduction path, thereby improving the transverse heat diffusion efficiency of the hybrid substrate.

Description

Technical Field

[0001] This invention relates to the field of PCB heat dissipation design technology, specifically to a method for optimizing the lateral heat conduction structure of a PCB based on an embedded aluminum block. Background Technology

[0002] In the thermal management of hybrid substrate PCBs, the heat generated by power devices is mainly conducted laterally to the heat dissipation boundary through the substrate insulation layer. However, the thermal conductivity of ordinary PCB insulation materials is usually low, resulting in insufficient lateral heat diffusion capacity. This leads to localized heat concentration below the power devices, forming high-temperature hotspots. Existing lateral heat conduction enhancement solutions involve embedding high thermal conductivity materials, such as copper or aluminum blocks, in the substrate insulation layer. The high thermal conductivity of the metal materials forms lateral heat conduction channels, reducing the hotspot temperature. However, existing embedding solutions have two drawbacks. First, the placement of the aluminum blocks usually relies on experience or a uniformly arranged array, failing to accurately correspond to areas of heat flow congestion under actual operating conditions. This arrangement ignores the non-uniform heat flow distribution caused by differences in the layout of different power devices and the intensity of heat sources. This results in some areas where the embedded aluminum blocks are not on effective heat conduction paths, leading to material redundancy and increased costs, while the actual heat flow concentration areas may not be effectively dissipated. Second, the geometry of the aluminum blocks is often a simple rectangle or circle, lacking specific adaptation to the characteristics of local heat flow density vector distribution. Simple shapes are difficult to match the complex and ever-changing heat flow path framework. In areas where the heat flow direction changes sharply or branches, the aluminum block does not fit well with the dense heat flow path, and cannot maximize the lateral heat diffusion efficiency.

[0003] The problem to be solved by this invention is how to accurately determine the placement position of the aluminum block in the substrate layer based on the actual heat flow distribution characteristics, and how to generate the aluminum block outline shape that matches the local heat flow path shape, so as to avoid the loss of thermal conductivity caused by empirical uniform layout and fixed shape design. Summary of the Invention

[0004] The purpose of this invention is to provide a method for optimizing the lateral thermal conductivity structure of a PCB based on embedded aluminum blocks. This method can determine the embedded point of the aluminum block for the actual location of the heat flow congestion by reconstructing the heat flow vector field and identifying the lateral thermal conductivity bottleneck area. At the same time, it can improve the lateral thermal conductivity performance of the hybrid substrate by parametric shape optimization so that the aluminum block outline fully overlaps with the heat flow path skeleton.

[0005] To achieve the above objectives, the present invention provides the following technical solution: To solve the problem of local heat accumulation caused by insufficient lateral thermal conductivity in hybrid substrates, the present invention provides a PCB lateral thermal conductivity structure optimization method based on embedded aluminum blocks. By analyzing the vector field of the heat distribution state, the bottleneck area that needs to be enhanced for thermal conductivity is accurately located, and an embedded aluminum block layout scheme with a high-efficiency heat dissipation path is automatically generated.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for optimizing the lateral thermal conductivity structure of a PCB based on embedded aluminum blocks, the steps of which include: acquiring initial thermal distribution data and power device layout information of a hybrid substrate. As a preferred acquisition method, the infrared thermal image sequence acquired from the surface of the hybrid substrate is subjected to time-averaged filtering to extract the temperature field matrix reflecting steady-state operating conditions. Simultaneously, the power consumption values ​​and coordinate information of the power devices are parsed from the design file, and the two are aligned to generate a power heat source distribution map, providing accurate boundary conditions for subsequent analysis.

[0007] Based on the obtained initial heat distribution data, a heat flux vector field reconstruction process is performed to generate a heat flux density distribution map. In the technical solution of this invention, based on the Fourier heat conduction equation, the spatial gradient of the steady-state temperature field matrix is ​​solved using the finite element method to obtain the temperature gradient vector of each discrete element. Then, the temperature gradient vector of each element is multiplied by the material thermal conductivity tensor corresponding to that location to quantitatively obtain the heat flux density vector, and arranged according to the grid index to intuitively construct a heat flux density distribution map, completing the transformation from a temperature scalar field to a heat flux vector field.

[0008] Based on the heat flux density distribution map, the system automatically identifies lateral heat conduction bottleneck regions, which are areas where the heat flux density gradient exceeds a preset threshold. The preferred identification process involves calculating the sum of the partial derivatives of the heat flux density vector with respect to the coordinates within each grid cell to obtain the divergence value. Grid cells with abnormally high divergence values ​​and a continuous distribution are marked as candidate bottleneck regions. Morphological dilation is then applied to these candidate regions to merge adjacent or fractured areas, ultimately generating a complete and connected lateral heat conduction bottleneck region, ensuring comprehensive capture of weak points in heat dissipation.

[0009] For the identified transverse heat conduction bottleneck regions, the embedding location and shape parameters of the aluminum block are determined by in-depth analysis of its heat flux vector divergence characteristics. Regarding the determination of the embedding location, a preferred approach is to extract the heat flux path skeleton within the transverse heat conduction bottleneck region. This skeleton is represented by a series of maximum gradient lines extending along the heat flux density vector direction. Then, the bifurcation nodes and extreme points with drastic curvature changes on the heat flux path skeleton are selected as candidate location nodes. Spatial clustering operations are performed on these nodes, and the cluster centers are used as the final embedding locations of the aluminum block. This ensures that the aluminum block can be accurately placed at the critical heat flux hubs where heat converges and changes direction.

[0010] Regarding the determination of the aluminum block shape parameters, this invention abstracts the aluminum block shape into a geometric figure controlled by a closed Bézier curve, and the area enclosed by it is calculated and constrained based on the preset aluminum block volume and the thickness of the substrate insulation layer. With the optimization objective of maximizing the overlap between the area enclosed by the closed Bézier curve and the area covered by the heat flow path framework, a simulated annealing algorithm is used to iteratively optimize the coordinate sequence of control points, ultimately generating aluminum block shape parameters that best conform to the actual heat flow expansion path, thereby achieving the highest lateral thermal conductivity within a limited volume.

[0011] As a preferred technical solution of the present invention, in order to eliminate redundant design and retain the core aluminum block that plays a decisive role in improving the overall thermal conductivity, the aforementioned determined aluminum block embedding positions and shape parameters are combined into an aluminum block layout matrix. By performing singular value decomposition on the thermal resistance network matrix corresponding to the layout matrix and extracting singular value vectors, a subset of aluminum blocks corresponding to singular vectors with singular values ​​greater than a set threshold is selected, and other aluminum blocks with low contribution are removed, ultimately forming a refined and efficient aluminum block embedding position and shape parameter.

[0012] When generating the optimized scheme for the lateral thermal conductivity structure of the hybrid substrate, which includes the above-mentioned optimized parameters, computer-aided manufacturing software is used to directly generate CNC machining path codes for processing the aluminum block embedding grooves based on the embedded position and shape parameters of the aluminum block. Simultaneously, the surface condition parameters of the aluminum block obtained after a series of surface treatments, including degreasing, alkaline washing, and electroless nickel plating, are associated with this machining path code and integrated into the scheme. This ensures high-fidelity transfer from the design scheme to the manufacturing process and improves the interface quality between the aluminum block and the substrate.

[0013] The method described in this invention incorporates an iterative verification and closed-loop adjustment mechanism after generating an optimized lateral thermal conductivity structure for a hybrid substrate. Specifically, the optimized scheme is input into finite element thermal simulation software to reproduce its thermal behavior, and the optimized temperature field distribution is calculated. The temperature deviation between the optimized temperature field and the initial thermal distribution data is calculated point by point to form a deviation matrix. If the maximum temperature gradient value in the deviation matrix still exceeds a preset gradient threshold, feedback adjustment is initiated: an adjustment window is defined centered on the point of maximum temperature gradient, and a position offset opposite to the original heat flux density vector is applied to the embedded position of the aluminum block within the window. Simultaneously, a scaling factor proportional to the magnitude of the temperature gradient is applied to the coordinates of the Bézier curve control points in the shape parameters of the aluminum block to adjust the coverage area of ​​the aluminum block. The steps of determining the position and shape of the aluminum block are re-executed using the updated parameters until the peak value of the temperature gradient is suppressed within the allowable threshold. This mechanism ensures that the final optimized structure achieves a uniform and sufficient lateral heat dissipation effect by iteratively eliminating residual heat concentration points.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows: By transforming initial thermal distribution data into a heat flux density distribution map and identifying continuous grid cell regions exceeding a preset threshold as lateral heat conduction bottlenecks based on divergence values, this technique explicitly locates weak points in heat conduction that are difficult to measure directly. Compared to the method of uniformly embedding aluminum blocks based on experience, the regions identified based on the divergence of the heat flux vector field precisely correspond to locations where heat flux density changes drastically and lateral conduction is obstructed. This allows subsequent aluminum block embedding operations to be performed directly on the problem origin, avoiding the waste of material in ineffective areas. A heat flux path skeleton is extracted from the lateral heat conduction bottleneck regions, and bifurcation nodes and curvature extrema on the skeleton are used as candidate location nodes. Spatial clustering is then used to determine the aluminum block embedding location. This method ensures that the embedding point falls precisely on key nodes where heat flux directions intersect or turn, leading to deterioration of lateral conduction conditions. Compared to regular array embedding, this method of targeted intervention based on heat flux topology features achieves precise alignment of the aluminum block with the required heat flux conduction location, achieving a more significant improvement in lateral heat diffusion with fewer aluminum blocks.

[0015] The aluminum block profile is described using parametric Bézier curves, and iterative optimization is performed with the goal of maximizing the overlap between the region enclosed by the curves and the area covered by the heat flow path framework. This shape determination method allows the aluminum block's shape to adaptively change with the curvature and bifurcation of the local heat flow path. Compared with using only rectangular or circular aluminum blocks, the optimized free-form curve profile extends along the dense heat flow path, fully filling the space covered by the heat flow framework, increasing the aluminum block's cross-section's area for receiving lateral heat flow and improving flow continuity. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a flowchart of a PCB lateral thermal conductivity structure optimization method based on embedded aluminum blocks; Figure 2 This is a flowchart of the method for reconstructing the heat flux density distribution of a hybrid substrate; Figure 3 This is a flowchart for identifying horizontal heat conduction bottleneck areas; Figure 4 This is a flowchart for determining the embedding location and shape parameters of the aluminum block based on the calculation of heat flow vector divergence in the transverse heat conduction bottleneck region; Figure 5 This is a flowchart of the method for optimizing the lateral thermal conductivity structure of a hybrid substrate. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] See Figure 1 This invention provides a method for optimizing the lateral thermal conductivity structure of a PCB based on embedded aluminum blocks, comprising: acquiring initial thermal distribution data and power device layout information of a hybrid substrate; performing heat flux vector field reconstruction processing on the initial thermal distribution data to generate a heat flux density distribution map; identifying lateral thermal conductivity bottleneck regions based on the heat flux density distribution map, wherein the lateral thermal conductivity bottleneck regions are regions where the heat flux density gradient exceeds a preset threshold; determining the embedded position and shape parameters of the aluminum block based on the heat flux vector divergence calculation of the lateral thermal conductivity bottleneck regions; and generating an optimization scheme for the lateral thermal conductivity structure of the hybrid substrate based on the embedded position and shape parameters of the aluminum block. Example 1:

[0020] In specific implementation, please refer to Figure 2 The method for obtaining the initial thermal distribution data and power device layout information of the hybrid substrate is as follows: Infrared thermal image sequences of the hybrid substrate surface are acquired. These sequences are continuously acquired by an infrared thermal imager at a fixed frame rate. Each frame records the temperature response value at each pixel location on the hybrid substrate surface. The infrared thermal image sequences are then subjected to time-averaged filtering. The implementation process of time-averaged filtering involves determining the number of frames for the time-averaged filter. , It is a preset positive integer. The value is determined based on the frame rate of the infrared thermal imager and the thermal stabilization time of the hybrid substrate. When the frame rate of the infrared thermal imager is... Frames per second, the time required for the hybrid substrate to reach thermal steady state after power is applied is seconds, Set as The floor value. The continuous values ​​in the infrared thermal image sequence. The temperature values ​​at the same pixel location in the frame are summed and divided by . The time-averaged temperature value of the pixel location is obtained, and the time-averaged temperature values ​​of all pixel locations constitute the steady-state temperature field matrix. Each row of the steady-state temperature field matrix corresponds to the row coordinate of a pixel on the surface of the hybrid substrate, and each column corresponds to the column coordinate of a pixel. The matrix elements are the steady-state temperature values ​​of that pixel location.

[0021] The power consumption value list and coordinate list of power devices are parsed from the design file of the hybrid substrate. The design file is circuit board design data containing component parameters and layout information, and can be in ODB++ or Gerber format. The nominal power consumption value of each power device is extracted from the design file to generate a power consumption value list, where each element represents the power consumption value of a single power device. The center coordinates of each power device on the hybrid substrate are extracted from the design file to generate a coordinate list, where each element represents a two-dimensional coordinate. The power consumption value list and the coordinate list are aligned, matching them one-to-one according to the component reference designator, to generate a power heat source distribution map. The power heat source distribution map is a two-dimensional data matrix with the same dimensions as the steady-state temperature field matrix. Non-zero elements in the matrix correspond to the center coordinates of the power devices, and the element value is equal to the power consumption value of that device; all other elements have a value of zero.

[0022] The method for reconstructing the heat flux vector field from the initial heat distribution data to generate a heat flux density distribution map is as follows: Based on the Fourier heat conduction equation, the spatial gradient of the steady-state temperature field matrix is ​​solved using the finite element method. The surface region of the hybrid substrate is discretized into a triangular finite element mesh, with each triangular finite element containing three nodes. The temperature values ​​of the nodes are obtained from the steady-state temperature field matrix through bilinear interpolation. Within each triangular finite element, a shape function is constructed, which is a linear function of the element node coordinates. The temperature field is interpolated using the shape function, and then the partial derivatives are calculated with respect to the spatial coordinates to obtain the temperature gradient vector of each triangular finite element. The temperature gradient vector is expressed as:

[0023] in, This represents the temperature gradient vector within a triangular finite element. For the node index of the triangular finite element, For nodes Temperature value, For nodes The gradient vector of the shape function with respect to spatial coordinates. Shape function The expression is , Let the area of ​​a finite element of a triangle be . , , The coefficients are related to the node coordinates. and These are the local spatial coordinates within a triangular finite element.

[0024] The temperature gradient vector of each finite element is multiplied by the corresponding material thermal conductivity tensor, which is pre-defined based on the anisotropic thermal conductivity of each layer of the hybrid substrate. For each triangular finite element, a pre-established material property mapping table is consulted according to its spatial position on the hybrid substrate to obtain the material thermal conductivity tensor corresponding to that triangular finite element. , This is a second-order tensor, represented as a 2×2 matrix in the two-dimensional case. The multiplication operation is a matrix multiplication, yielding the heat flux density vector of this triangular finite element. The calculation relationship is as follows The heat flux density vectors of all triangular finite elements are arranged according to the mesh index of the finite elements to generate a heat flux density distribution map. The heat flux density distribution map is a vector field, with each vector corresponding to a triangular finite element. The vector direction represents the direction of heat flow, and the vector magnitude represents the magnitude of heat flux density. Example 2:

[0025] In specific implementation, please refer to Figure 3 The method for identifying the transverse heat conduction bottleneck region based on the heat flux density distribution map is as follows: Calculate the divergence value of the heat flux density vector for each grid cell in the heat flux density distribution map. Each grid cell in the heat flux density distribution map corresponds one-to-one with a triangular finite element generated by the finite element method, and each grid cell contains a heat flux density vector. The divergence value is the sum of the partial derivatives of the heat flux density vector with respect to the coordinates within the grid cell. For the , Line number Column grid cells, divergence value The calculation process is as follows: using grid cells Centered on the grid, the heat flux density vector components of its adjacent grid cells are approximated using a central difference method, and the divergence value is... The expression is:

[0026] in, Indicates the first Line number The divergence value of the heat flux density vector of the column grid cell. For grid cells in Dimensions in the direction For grid cells in Dimensions in the direction Indicates the first Line number The heat flux density vector of the column grid cell is directional components, Indicates the first Line number The heat flux density vector of the column grid cell is directional components, Indicates the first Line number The heat flux density vector of the column grid cell is directional components, Indicates the first Line number The heat flux density vector of the column grid cell is The directional component. For grid cells located at the boundary of the heat flux density distribution map, the divergence value is calculated using the one-sided difference formula. The missing adjacent grid cell components in the one-sided difference formula are replaced by the boundary grid cell's own components.

[0027] Regions of consecutive grid cells with divergence values ​​exceeding a preset divergence threshold are marked as candidate bottleneck regions. The preset divergence threshold is determined as follows: The divergence values ​​of all grid cells in the heat flux density distribution map are statistically analyzed, generating a statistical histogram of the divergence values. The upper quartile of this histogram is used as the preset divergence threshold. The upper quartile is the value located at the 75th percentile after arranging all divergence values ​​in ascending order. This process is repeated for each grid cell in the heat flux density distribution map. When the first... Line number divergence value of column grid cells When the divergence exceeds a preset divergence threshold, the grid cell is marked as a bottleneck cell, and all bottleneck cells constitute the initial bottleneck cell set. A connected component labeling algorithm is then used to process the initial bottleneck cell set. This algorithm employs the eight-neighbor connectivity criterion, grouping spatially adjacent bottleneck cells into the same connected component, with each connected component representing a candidate bottleneck region.

[0028] Morphological dilation is applied to the candidate bottleneck region using a circular structuring element, with the radius of the circular structuring element being... Set to three times the maximum value of the grid cell size, radius The rationale for using a factor of three is to ensure that the interval between adjacent candidate bottleneck regions after expansion is less than [a certain value]. The expansion process allows for the connection and merging of regions while avoiding excessive expansion that could lead to the incorrect merging of previously separate heat flow bottleneck regions. The center of the circular structural element is placed sequentially at the center of each boundary grid cell in the candidate bottleneck region, incorporating all grid cells within the coverage area of ​​the circular structural element into the expanded region. After completing the expansion operation on all candidate bottleneck regions, overlapping or adjacent regions within the expanded region are merged into a single region, generating a transverse heat conduction bottleneck region. The criterion for merging adjacent regions is: if two expanded regions share a common boundary or common grid cells, they are merged into one transverse heat conduction bottleneck region. The generated transverse heat conduction bottleneck region is represented as a set of grid cells, with each region containing an index of a set of continuous grid cells. Example 3:

[0029] In specific implementation, please refer to Figure 4 Based on the calculation of heat flow vector divergence in the transverse heat conduction bottleneck region, the method for determining the embedding location of the aluminum block is as follows: A heat flux path skeleton is extracted for the transverse heat conduction bottleneck region. The heat flux path skeleton is the set of maximum gradient lines along the direction of the heat flux density vector. The heat flux density vector direction of each grid cell within the transverse heat conduction bottleneck region is used as the streamline tracing direction. A streamline tracing algorithm is used to generate a set of streamlines within the transverse heat conduction bottleneck region. The seed points of the streamline tracing algorithm are evenly distributed on the inflow boundary of the transverse heat conduction bottleneck region. The inflow boundary is the set of boundary grid cells in the transverse heat conduction bottleneck region where the heat flux density vector direction points inward into the region. On each streamline, the magnitude of the heat flux density vector is calculated, and the point where the magnitude reaches a local maximum along the streamline direction is marked as a gradient ridge. All gradient ridges on all streamlines are connected into lines. The connection method is to connect gradient ridges on adjacent streamlines whose spatial distance is less than a preset connection threshold with straight line segments. The preset connection threshold is set to twice the maximum value of the grid cell size. This setting allows adjacent gradient ridges on adjacent streamlines to be connected while avoiding misconnection across streamlines. The set of connected gradient ridges constitutes the heat flux path skeleton, which is represented as a set of interconnected or unconnected skeleton line segments.

[0030] Bifurcation nodes and curvature extrema on the heat flux path skeleton are used as candidate location nodes. The process for identifying bifurcation nodes is as follows: traverse each endpoint of a skeleton line segment in the heat flux path skeleton, count the number of skeleton line segments connected to that endpoint, and mark the endpoint as a bifurcation node when the number of connected skeleton line segments is greater than or equal to three. The process for identifying curvature extrema is as follows: perform curve parameterization on each skeleton line segment in the heat flux path skeleton, using arc length as the parameter. This indicates that the arc length parameter With equal intervals Sampling is performed to obtain a sequence of sampling points, with a step size of [missing information]. Set to half the minimum value of the grid cell size. Calculate the curvature value of the skeleton line segment at each sampling point. curvature value The calculation uses the three-point circular arc method, taking the sampling point and its three adjacent sampling points (a total of three points) to fit a circular arc. The curvature value is the reciprocal of the radius of the fitted circular arc. The curvature value is then... Arc length parameter The sampling points corresponding to local maxima in the curve are marked as curvature extrema. The criterion for determining local maxima is that the curvature value is greater than the curvature values ​​of the adjacent sampling points.

[0031] Candidate location nodes are spatially clustered, and the cluster centers are selected as the locations for embedding aluminum blocks. The spatial clustering process employs a density-based spatial clustering algorithm, which includes two preset parameters: neighborhood radius and density-based spatial clustering algorithm. and the minimum number of neighborhood points of the core point Neighborhood radius The value is set to five times the maximum value of the grid cell size. This value is chosen because at this scale, candidate location nodes belonging to different heat flow channels can be distinguished, while candidate location nodes belonging to the same heat flow channel can be grouped together. Minimum neighborhood number of core points. The value is set to three to ensure that a cluster contains at least three candidate location nodes to form effective cluster centers. When executing the density-based spatial clustering algorithm, the neighborhood radius of each candidate location node is first calculated. The number of adjacent candidate location nodes within the range, where the number of adjacent candidate location nodes is greater than or equal to Candidate location nodes are marked as core points. Starting from any core point, its neighborhood radius is calculated. All candidate location nodes within the range are grouped into the same cluster, and this cluster is continuously expanded through neighborhood relationships until no new candidate location nodes can be added, thus completing the construction of a cluster. This process is repeated until all core points have been visited. Candidate location nodes not included in any cluster are marked as noise points and removed. For each generated cluster, the arithmetic mean of the coordinates of all candidate location nodes in the cluster is calculated. The coordinates of the arithmetic mean are used as the cluster center, which is then used as the location for embedding the aluminum block.

[0032] In practical implementation, the method for determining the shape parameters of the aluminum block based on the heat flow vector divergence calculation of the transverse heat conduction bottleneck region is as follows: The shape of the aluminum block is parameterized as a coordinate sequence of control points for a closed Bézier curve. The closed Bézier curve is constructed by splicing cubic Bézier curves, each defined by four control points. The coordinate sequence of these control points includes... Group control points, The number of cubic Bézier curve segments that make up a closed Bézier curve. The value is set to six. The reason for choosing six is ​​to ensure that the closed Bézier curve can describe the shape of non-convex and irregular aluminum blocks while controlling the number of optimization variables. The area enclosed by the closed Bézier curve is equal to the preset aluminum block volume divided by the thickness of the hybrid substrate insulation layer. The preset aluminum block volume is pre-set based on heat dissipation enhancement requirements, and the hybrid substrate insulation layer thickness is obtained from the hybrid substrate design file. The area enclosed by the closed Bézier curve is calculated using Green's formula. In each iteration step, the coordinates of the control points are scaled to ensure that the area equals the preset value.

[0033] The goal is to maximize the overlap between the region enclosed by the closed Bézier curve and the area covered by the heat flow path skeleton. A simulated annealing algorithm is used to iteratively optimize the coordinate sequence of control points, generating the shape parameters of the aluminum block. (Overlap rate) The calculation expression is:

[0034] in, Indicates the overlap rate. This represents the region enclosed by a closed Bézier curve. This represents a buffer zone covering the area of ​​the heat flow path skeleton. The buffer zone is formed by extending a preset buffer width to both sides of each line segment in the heat flow path skeleton. Get the preset buffer width. Set to twice the maximum value of the grid cell size. This represents the area of ​​the intersection between the region enclosed by the closed Bézier curve and the buffer zone covered by the heat flow path framework. This represents the area of ​​the buffer zone covered by the heat flow path skeleton.

[0035] The process of iteratively optimizing the coordinate sequence of control points using the simulated annealing algorithm is as follows: Initialize temperature parameters , Set to 100, termination temperature Set to 0.01, temperature decay coefficient Set to 0.95, and perform at each temperature. iteration The value is set to be ten times the number of control points in the control point coordinate sequence. In each iteration, a control point is randomly selected, and its coordinates are subjected to a set of parameters that have a mean of zero and a standard deviation of [value missing]. Gaussian perturbation, standard deviation The initial value is set to the maximum value in the mesh cell size, and decreases linearly with decreasing temperature. The overlap rate of the closed Bézier curves after perturbation is calculated. The overlap rate after the perturbation is compared with the overlap rate before the perturbation. If the overlap rate increases, the perturbation is accepted; if the overlap rate decreases, the perturbation is rejected with probability. Accepting disturbances, among which This represents the amount by which the overlap rate decreases. The current temperature. When the temperature... Below the termination temperature When the simulated annealing algorithm terminates, the coordinate sequence of the control point with the highest overlap rate is output as the shape parameter of the aluminum block. Example 4:

[0036] In specific implementation, please refer to Figure 5 The method for combining the aluminum block embedding position and aluminum block shape parameters into an aluminum block layout matrix is ​​as follows: The location of the aluminum block is indicated by two-dimensional coordinates. The shape parameters of the aluminum block are represented by a coordinate sequence of control points of a closed Bézier curve, and the coordinate sequence includes... There are 1 control point, and the coordinates of each control point are two-dimensional coordinates. , The value is determined by the number of cubic Bézier curve segments set in the simulated annealing algorithm. The embedding position and shape parameters of each aluminum block are combined into an aluminum block parameter vector, where the first two components represent the embedding position. coordinates and Coordinates, with subsequent components being the coordinates of each control point in the control point coordinate sequence. coordinates and The coordinates are arranged alternately. The aluminum block parameter vectors of all aluminum blocks are arranged in rows to form an aluminum block layout matrix. The number of rows in the aluminum block layout matrix is ​​equal to the number of aluminum blocks, and the number of columns is equal to the dimension of the aluminum block parameter vector.

[0037] In practice, the singular value decomposition of the thermal resistance network matrix corresponding to the aluminum block layout matrix is ​​performed as follows: A thermal resistance network matrix is ​​constructed to reflect the resistance characteristics of heat conduction paths between different regions of the hybrid substrate. The surface of the hybrid substrate is divided into finite element meshes. Each region has a thermal resistance network matrix. The square matrix, the thermal resistance network matrix of the first Line number Column elements Indicates the first The region and the first The thermal resistance value between the regions. When the first region... The region and the first When regions are adjacent in a finite element mesh, the thermal resistance value The calculation method is to divide the center distance between the two regions by the product of the contact area between the two regions and the thermal conductivity of the material; when the first... The region and the first When the regions are not adjacent, the thermal resistance value Take infinity; when equal At that time, thermal resistance value Set to zero. Establish a correspondence between the thermal resistance network matrix and the aluminum block layout matrix. Each aluminum block in the aluminum block layout matrix corresponds to the row and column index of the region where the aluminum block is buried in the thermal resistance network matrix.

[0038] Singular value decomposition is performed on the thermal resistance network matrix. The expression for singular value decomposition is:

[0039] in, Represents the thermal resistance network matrix. for OK A real matrix of columns, It is a left singular vector matrix. for OK An orthogonal matrix of columns The Listed as number A left singular vector It is a singular value diagonal matrix. for OK A diagonal matrix of columns, Diagonal elements The values ​​are singular and arranged in descending order. It is a right singular vector matrix. for OK An orthogonal matrix of columns The Listed as number A right singular vector, superscript This indicates the matrix transpose.

[0040] In practice, the method for extracting singular value vectors and filtering subsets of aluminum blocks is as follows: From the singular value diagonal matrix Extracting the diagonal elements from the vector creates a singular value vector, which is: Calculate the cumulative energy ratio of the singular value vectors. The calculation method is the first The sum of squares of each singular value divided by the sum of squares of all singular values, where The number of singular values ​​selected. A preset singular value threshold is set to 0.95. This threshold is based on the following: in the matrix rank reduction approximation, when the cumulative energy ratio reaches 0.95, the retained singular vectors can cover 95% of the main heat conduction characteristics in the thermal resistance network matrix, while discarding less than 5% of the secondary characteristics results in a negligible contribution to the overall lateral heat conduction from the aluminum blocks. Each singular value in the singular value vector is correlated with the preset singular value threshold to find those that satisfy the cumulative energy ratio... The smallest value that is first greater than or equal to 0.95 The value will be sorted as the first. To the The left and right singular vectors corresponding to the singular values ​​of each bit are used as the selected singular vectors. The row indices of the selected singular vectors in the thermal resistance network matrix constitute the aluminum block subset index, and the aluminum block pointed to by the aluminum block subset index is the aluminum block subset. Elements other than the aluminum block subset are removed from the aluminum block embedding position and aluminum block shape parameters. The removal operation deletes the corresponding rows not included in the aluminum block subset index in the aluminum block layout matrix. The aluminum block embedding position and aluminum block shape parameters corresponding to the remaining rows are used as the generated aluminum block embedding position and aluminum block shape parameters.

[0041] In practical implementation, the optimized scheme for generating the lateral thermal conductivity structure of the hybrid substrate is implemented as follows: Based on the aluminum block's embedding location and shape parameters, computer-aided manufacturing (CAD) software is invoked to generate CNC machining path code for the aluminum block embedding groove. The CAD software receives the coordinate sequence of control points for the closed Bézier curve from the aluminum block's shape parameters, discretizes the closed Bézier curve into a contour trajectory approximated by straight line segments, with a discretization accuracy set to one-thousandth of the diagonal length of the control point bounding box. The contour trajectory is then translated coordinately based on the aluminum block's embedding location, ensuring the center of the contour trajectory coincides with the aluminum block's embedding location. The CAD software's contour milling strategy is then invoked, setting the tool diameter to 0.8 times the designed width of the aluminum block embedding groove and the feed rate to the recommended milling rate for the mixed substrate insulation layer material. This generates CNC machining path code for the aluminum block embedding groove. The CNC machining path code is in G-code format and includes G00 rapid positioning instructions, G01 linear interpolation instructions, G02 clockwise circular interpolation instructions, and G03 counterclockwise circular interpolation instructions.

[0042] The surface condition parameters of the aluminum block after sequential degreasing, alkaline washing, and electroless nickel plating were associated with the CNC machining path code. The degreasing treatment used an alkaline degreasing solution, a mixed aqueous solution of sodium hydroxide and sodium carbonate, at a temperature of 60°C for 10 minutes. The alkaline washing treatment used a sodium hydroxide solution with a mass concentration of 10%, at a temperature of 50°C for 5 minutes. The electroless nickel plating treatment used a plating solution with sodium hypophosphite as the reducing agent, maintaining the pH value between 4.5 and 5.0, at a temperature of 85°C, and for 30 minutes. After degreasing, alkaline washing, and electroless nickel plating, the surface condition parameters of the aluminum block were measured, including coating thickness, surface roughness, and coating adhesion. The coating thickness was measured using an X-ray fluorescence thickness gauge at five evenly distributed points on the aluminum block surface, and the arithmetic mean was taken as the coating thickness value. Surface roughness was measured using a contact roughness meter. The sampling length was set to 0.8 mm, and the evaluation length was set to 4 mm. The arithmetic mean deviation Ra of the profile was obtained as the surface roughness value. The coating adhesion was determined using the cross-cut test method. A cross grid with a spacing of 1 mm was drawn on the surface of the aluminum block. The grid was then adhered with tape and peeled off. The percentage of coating peeling area was counted, and the coating adhesion index was expressed as the percentage of peeling area.

[0043] Surface condition parameters are associated with CNC machining path codes. This association is achieved by embedding the surface condition parameters as comments in the header of the CNC machining path code file. The surface condition parameters are recorded as variables, with names of coating thickness, surface roughness, and coating adhesion, and values ​​representing the corresponding measured values. The CNC machining path code file with embedded surface condition parameters serves as an optimization scheme for the lateral thermal conductivity structure of the hybrid substrate. Example 5:

[0044] In practical implementation, after generating an optimized scheme for the lateral thermal conductivity structure of the hybrid substrate based on the embedded location and shape parameters of the aluminum block, the following steps are also included: The optimized lateral thermal conductivity structure of the hybrid substrate is input into the finite element thermal simulation software. This optimized structure includes complete definitions of the aluminum block embedding locations and shape parameters. After reading the optimized structure, the finite element thermal simulation software creates a 3D solid model of the hybrid substrate. The 3D solid model includes the structure of each layer of the hybrid substrate and the aluminum block entities embedded at their designated locations. The geometry of the aluminum block entities is generated by stretching the thickness of the hybrid substrate's insulating layer using closed Bezier curves from the aluminum block shape parameters. A tetrahedral finite element mesh is generated for the 3D solid model. The power consumption values ​​of each power device defined in the hybrid substrate design file are applied as heat source boundary conditions, and the convective heat transfer coefficient between the bottom surface of the hybrid substrate and the heat sink contact surface is applied as a heat dissipation boundary condition. The convective heat transfer coefficient is set based on empirical values ​​for natural convection or forced air cooling in the thermal design specifications. Steady-state thermal simulation calculations are performed to solve the Fourier heat conduction equation in the finite element model, obtaining the optimized temperature field distribution, which is the set of temperature values ​​for each node of the 3D solid model.

[0045] The optimized temperature field distribution is compared with the initial thermal distribution data point-by-point to calculate the temperature deviation, generating a deviation matrix. The initial thermal distribution data is the steady-state temperature field matrix extracted in the initial thermal distribution data acquisition step of the hybrid substrate, which records the temperature value at each pixel location on the hybrid substrate surface. The optimized temperature field distribution on the hybrid substrate surface is mapped to the same pixel coordinate grid as the steady-state temperature field matrix using an interpolation method employing shape function interpolation within triangular elements. A subtraction operation is performed between the optimized temperature value and the initial temperature value at the same pixel coordinate location, and the result of the subtraction operation is used as the temperature deviation value at that pixel coordinate location. The temperature deviation values ​​at all pixel coordinate locations are arranged according to the row and column structure of the steady-state temperature field matrix to generate the deviation matrix. Line number Column elements The calculation method is to subtract the initial temperature value from the optimized temperature value.

[0046] If the maximum temperature gradient in the deviation matrix exceeds the preset gradient threshold, the aluminum block embedding position and shape parameters are adjusted, and the steps of determining the aluminum block embedding position and shape parameters and generating the optimized scheme for the lateral thermal conductivity structure of the hybrid substrate are repeated until the maximum temperature gradient in the deviation matrix does not exceed the preset gradient threshold. The temperature gradient in the deviation matrix is ​​calculated as follows: for each element in the deviation matrix, its approximate partial derivative with central difference in the row and column directions is calculated. Line number Row-direction partial derivatives of column elements For the first Line number Column elements and the first Line number The partial derivative of the column direction is the difference between the column elements divided by twice the column spacing. For the first Line number Column elements and the first Line number The difference between column elements divided by twice the cell row spacing. Temperature gradient value. partial derivatives in the row direction partial derivatives in the column direction Take the square root of the sum of squares. Iterate through all elements in the deviation matrix and take the temperature gradient values. The maximum value is taken as the maximum temperature gradient. The preset gradient threshold is determined based on the thermal stress tolerance limit of the hybrid substrate material. The preset gradient threshold is set to five degrees Celsius per millimeter. The basis for this setting is that when the temperature gradient exceeds five degrees Celsius per millimeter, the thermal stress generated by the difference in thermal expansion between the insulating layer material and the aluminum block material in the hybrid substrate exceeds 50% of the allowable stress of the insulating layer material.

[0047] In practice, the methods for adjusting the embedding position and shape parameters of the aluminum block are as follows: The adjustment window is defined centered on the grid cell with the largest temperature gradient in the deviation matrix. The temperature gradient values ​​of all grid cells in the deviation matrix are then iterated through. Record the row and column indices of the grid cells corresponding to the maximum temperature gradient, and expand outward from that grid cell. Each grid cell defines an adjustment window, which is a square region with a side length of [value missing]. Each grid cell. The value is set to six, based on the principle that the adjustment window can cover the heat-affected zone around the point of maximum temperature gradient, while the size of the adjustment window does not exceed 80% of the boundary of the hybrid substrate.

[0048] Apply a position offset opposite in direction to the heat flux density vector at the aluminum block embedding location within the adjustment window. Obtain the heat flux density vectors of all grid cells within the adjustment window. The heat flux density vectors are extracted from the heat flux density distribution map. Take the average value of all heat flux density vectors within the adjustment window to obtain the average heat flux density vector. The average heat flux density vector The direction is reversed, which is used as the direction of the position offset, and the magnitude of the position offset is... The expression is:

[0049] in, This indicates the magnitude of the position offset. This is the offset scaling factor. This represents the maximum temperature gradient in the deviation matrix. This is the maximum value among the mesh cell sizes. Offset scaling factor. The value is set to 0.3, based on the principle that the offset of the aluminum block's embedding position in a single iteration does not exceed one-tenth of the side length of the adjustment window, ensuring stable convergence during the iteration process. The embedding position of the aluminum block within the adjustment window is then aligned with the average heat flux density vector. Distance of movement in the opposite direction This generates the updated aluminum block embedding location.

[0050] A scaling factor proportional to the temperature gradient is applied to the coordinates of the Bézier curve control points in the aluminum block shape parameters to generate updated aluminum block embedding positions and shape parameters. Scaling factor The calculation method is based on the maximum temperature gradient. The ratio to the preset gradient threshold. For each control point coordinate of the closed Bézier curve in the aluminum block shape parameters, a scaling factor is applied radially along the line connecting the control point coordinates and the aluminum block embedding position, centered on the aluminum block embedding position. The scaling operation moves the control point coordinates radially. After the movement, the distance between the control point and the embedded position of the aluminum block is equal to the original distance multiplied by the scaling factor. When the scaling factor When the scaling factor is greater than one, the control points expand outwards; when the scaling factor is greater than one, the control points expand outwards. When the value is less than one, the control point shrinks inward. The offset aluminum block embedding position is combined with the scaled aluminum block shape parameters to generate updated aluminum block embedding positions and shape parameters.

[0051] The steps of determining the aluminum block's embedding location and shape parameters based on the heat flow vector divergence calculation of the lateral thermal bottleneck region are repeated using the updated aluminum block embedding location and shape parameters. The steps of generating an optimized lateral thermal conduction structure scheme for the hybrid substrate based on the aluminum block embedding location and shape parameters are then repeated. The newly generated optimized lateral thermal conduction structure scheme is then input into the finite element thermal simulation software to calculate the temperature field distribution after the new round of optimization. Point-by-point temperature deviation calculations are performed between this new scheme and the initial thermal distribution data to determine whether the maximum temperature gradient value in the deviation matrix exceeds a preset gradient threshold. This iterative process is repeated until the maximum temperature gradient value in the deviation matrix does not exceed the preset gradient threshold, at which point the iteration terminates, and the final optimized lateral thermal conduction structure scheme for the hybrid substrate is output.

[0052] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for optimizing the lateral thermal conductivity structure of a PCB based on embedded aluminum blocks, characterized in that, The method includes: Obtain initial thermal distribution data and power device layout information of the hybrid substrate; The initial heat distribution data is reconstructed using a heat flux vector field to generate a heat flux density distribution map; The transverse heat conduction bottleneck region is identified based on the heat flux density distribution map. The transverse heat conduction bottleneck region is the region where the heat flux density gradient exceeds a preset threshold. Based on the heat flow vector divergence calculation of the transverse heat conduction bottleneck region, the embedding location and shape parameters of the aluminum block are determined. An optimized scheme for the lateral thermal conductivity structure of the hybrid substrate is generated based on the embedded position and shape parameters of the aluminum block.

2. The PCB lateral thermal conductivity structure optimization method based on embedded aluminum blocks according to claim 1, characterized in that, The acquisition of initial thermal distribution data and power device layout information of the hybrid substrate includes: Infrared thermal image sequence of the hybrid substrate surface is acquired, and time-averaged filtering is performed on the infrared thermal image sequence to extract the steady-state temperature field matrix; The power consumption value list and coordinate list of the power devices are parsed from the design file of the hybrid substrate, and the power consumption value list and the coordinate list are aligned to generate a power heat source distribution map.

3. The PCB lateral thermal conductivity structure optimization method based on embedded aluminum blocks according to claim 1, characterized in that, The step of reconstructing the heat flux vector field from the initial heat distribution data to generate a heat flux density distribution map includes: Based on the Fourier heat conduction equation, the spatial gradient of the steady-state temperature field matrix is ​​solved using the finite element method to obtain the temperature gradient vector of each finite element. The heat flux density vector is obtained by multiplying the temperature gradient vector of each finite element with the material thermal conductivity tensor at the corresponding location. The heat flux density vectors are then arranged according to the mesh index of the finite elements to generate the heat flux density distribution map.

4. The PCB lateral thermal conductivity structure optimization method based on embedded aluminum blocks according to claim 1, characterized in that, The step of identifying the transverse heat conduction bottleneck region based on the heat flux density distribution map includes: Calculate the divergence value of the heat flux density vector in each grid cell of the heat flux density distribution map. The divergence value is the sum of the partial derivatives of the heat flux density vector with respect to the coordinates within the grid cell. The continuous grid cell region whose divergence value exceeds the preset divergence threshold is marked as a candidate bottleneck region; The candidate bottleneck region is subjected to morphological expansion processing, and adjacent regions are merged to generate the transverse thermal conduction bottleneck region.

5. The PCB lateral thermal conductivity structure optimization method based on embedded aluminum blocks according to claim 1, characterized in that, The calculation of the heat flow vector divergence based on the transverse thermal bottleneck region to determine the location of the aluminum block embedding includes: A heat flow path skeleton is extracted from the transverse heat conduction bottleneck region. The heat flow path skeleton is a set of maximum gradient lines along the direction of the heat flow density vector. The bifurcation nodes and curvature extrema points on the heat flow path skeleton are used as candidate location nodes. Spatial clustering is performed on the candidate location nodes, and the cluster center is selected as the location where the aluminum block is buried.

6. The PCB lateral thermal conductivity structure optimization method based on embedded aluminum blocks according to claim 1, characterized in that, The determination of the aluminum block shape parameters based on the heat flow vector divergence calculation of the transverse thermal bottleneck region includes: The shape of the aluminum block is parameterized as a coordinate sequence of control points of a closed Bézier curve, and the area enclosed by the closed Bézier curve is equal to the preset volume of the aluminum block divided by the thickness of the insulating layer of the hybrid substrate. With the goal of maximizing the overlap between the region enclosed by the closed Bézier curve and the region covered by the heat flow path skeleton, the coordinate sequence of the control points is iteratively optimized using a simulated annealing algorithm to generate the shape parameters of the aluminum block.

7. The method for optimizing the PCB lateral thermal conductivity structure based on embedded aluminum blocks according to claim 5 or 6, characterized in that, The determination of the aluminum block's embedding location and shape parameters also includes: The aluminum block embedding location and the aluminum block shape parameters are combined to form an aluminum block layout matrix; Singular value decomposition is performed on the thermal resistance network matrix corresponding to the aluminum block layout matrix to extract singular value vectors, and a subset of aluminum blocks corresponding to singular vectors whose singular values ​​are greater than a preset singular value threshold is selected. The aluminum block embedding location and aluminum block shape parameters are generated by removing elements other than the aluminum block subset from the aluminum block embedding location and aluminum block shape parameters.

8. The PCB lateral thermal conductivity structure optimization method based on embedded aluminum blocks according to claim 7, characterized in that, The optimized scheme for generating the lateral thermal conductivity structure of the hybrid substrate includes: Based on the aluminum block's embedding location and shape parameters, computer-aided manufacturing software is invoked to generate CNC machining path code for the aluminum block embedding groove; The surface state parameters of the aluminum block after sequential degreasing, alkaline washing, and electroless nickel plating are associated with the CNC machining path code to generate an optimized scheme for the lateral thermal conductivity structure of the hybrid substrate.

9. The method for optimizing the PCB lateral thermal conductivity structure based on embedded aluminum blocks according to claim 1, characterized in that, After generating the optimized scheme for the lateral thermal conductivity structure of the hybrid substrate based on the embedded position and shape parameters of the aluminum block, the method further includes: The optimized scheme of the lateral thermal conductivity structure of the hybrid substrate was input into the finite element thermal simulation software to calculate the optimized temperature field distribution. The optimized temperature field distribution is compared with the initial thermal distribution data point by point to calculate the temperature deviation, and a deviation matrix is ​​generated. If the maximum temperature gradient in the deviation matrix exceeds the preset gradient threshold, then adjust the embedding position and shape parameters of the aluminum block and repeat steps 4 and 5 until the maximum temperature gradient in the deviation matrix does not exceed the preset gradient threshold.

10. The method for optimizing the PCB lateral thermal conductivity structure based on embedded aluminum blocks according to claim 9, characterized in that, The adjustment of the embedded position and shape parameters of the aluminum block includes: The adjustment window is defined with the grid cell with the largest temperature gradient in the deviation matrix as the center; Within the adjustment window, a positional offset is applied to the embedded position of the aluminum block, with the direction opposite to the heat flux density vector; A scaling factor proportional to the temperature gradient is applied to the coordinates of the Bézier curve control points in the aluminum block shape parameters to generate updated aluminum block embedding positions and aluminum block shape parameters.

Images