A heat sink cooling channel design method based on three-dimensional topological optimization

Abstract

This invention relates to the field of radiator design and thermal management technology, and particularly to a radiator cooling channel design method based on three-dimensional topology optimization. The method includes the following steps: performing thermal imaging scanning on the electronic device to acquire temperature distribution data and generate a hotspot map; calculating the temperature gradient based on the hotspot map, determining the temperature gradient threshold, dividing the main circulation region and the secondary circulation region, performing boundary smoothing optimization of the circulation region, and generating a circulation partition map; determining the inlet location and heat flow requirements based on the circulation partition map, and identifying the thermal anchor point set; determining the hierarchical branch structure of the main circulation and the cellular network structure of the secondary circulation; generating a channel skeleton map based on the hierarchical branch structure and the cellular network structure; determining the basic isolation wall and adding reinforcing structures, and optimizing the thickness according to the local temperature difference to form a thermally conductive valve structure. This invention solves the problem of low efficiency of radiators under complex thermal loads through radiator design and thermal management technology.

Description

Technical Field

[0001] This invention relates to the field of radiator design and thermal management technology, and in particular to a radiator cooling channel design method based on three-dimensional topology optimization. Background Technology

[0002] Traditional heat sinks employ a single-channel system design, with the channel structure and cooling capacity uniformly distributed across the entire heat dissipation area. This approach fails to adapt to the uneven distribution of hot spots within electronic devices. When the overall cooling capacity is increased to meet the high heat dissipation demands of hot spots, non-hot spot areas become overcooled, leading to energy waste. Conversely, when designed with uniform temperature as the goal, hot spots are prone to overheating, affecting equipment performance and lifespan, and even posing safety hazards. This design method cannot reconcile the contradiction between hot spot management and temperature uniformity. Existing heat sink designs are mostly based on theoretical thermal models or simplified assumptions, lacking precise analysis and response capabilities to actual thermal load characteristics. Cooling channel layouts typically employ regular geometric shapes (such as parallel channels, serpentine channels, etc.), resulting in unreasonable distribution of cooling resources (fluid flow rate, channel dimensions, etc.) throughout the system. This leads to overcooling and energy waste in some areas, while undercooling and excessively high temperatures in others. Furthermore, traditional designs cannot dynamically respond to changes in thermal load under different operating conditions, resulting in high energy consumption and low efficiency. Even in systems employing zoned cooling, effective heat exchange mechanisms are often lacking between cooling zones, hindering efficient heat transfer from high-temperature to low-temperature areas. The multiple independent cooling circuits are completely isolated and cannot work together, requiring a complex active control system for adjustment, which increases costs and the risk of failure. This design, with its low heat transfer efficiency between areas, results in slow system thermal response, poor stability, and difficulty in coping with complex and changing operating environments.

[0003] In summary, existing technologies face challenges such as the contradiction between hotspot management and uniform heat dissipation, inaccurate allocation of cooling resources, and poor coordination among circulation systems, which urgently need to be addressed. Summary of the Invention

[0004] Therefore, it is necessary to provide a heat sink cooling channel design method based on three-dimensional topology optimization to solve at least one of the above-mentioned technical problems.

[0005] To achieve the above objectives, a heat sink cooling channel design method based on three-dimensional topology optimization includes the following steps:

[0006] Step S1: Perform thermal imaging scanning on the electronic device to obtain temperature distribution data and generate a heat map;

[0007] Step S2: Calculate the temperature gradient based on the heat map, determine the temperature gradient threshold, divide the main loop region and the secondary loop region, perform loop region boundary smoothing optimization, and generate a loop partition map.

[0008] Step S3: Based on the cyclic partitioning diagram, determine the inlet location and heat flow requirements, and identify the set of thermal anchor points; determine the hierarchical branching structure of the main loop and the cellular network structure of the secondary loop based on the set of thermal anchor points; generate a channel skeleton diagram based on the hierarchical branching structure and the cellular network structure.

[0009] Step S4: Extract the interface location information between the main circulation area and the secondary circulation area from the channel skeleton diagram, determine the basic isolation wall and add the reinforcement structure, use the hotspot map to calculate the local temperature difference on both sides of the interface, optimize the thickness according to the local temperature difference to form the heat conduction valve structure, and generate the heat exchange interface structure.

[0010] Step S5: Based on the heat exchange interface structure, complete the structural design of the heat sink channel and generate a 3D printing scheme.

[0011] This invention obtains highly reliable real-world thermal behavior data through dynamic thermal imaging scanning and an adaptive sampling strategy. The precise registration of the generated multi-layered information hotspot map with the CAD model provides accurate, multi-dimensional thermodynamic input for subsequent design, ensuring the scientific rigor of the design from the outset. Using temperature gradients rather than absolute temperatures for zoning allows for more accurate identification of heat change boundaries. By dynamically calculating adaptive thresholds using real-time and historical data, intelligent and robust zoning is achieved, eliminating the reliance on experience in traditional methods. Boundary smoothing optimization lays a solid geometric foundation for the subsequent generation of low-resistance channels. The main and secondary circulation systems are designed with distinct channel structures, resolving the contradiction between heat dissipation efficiency and temperature uniformity that a single structure cannot simultaneously achieve. The main circulation employs a hierarchical branching structure with the shortest thermodynamic path for efficient cooling of core hotspots; the secondary circulation uses a density-adaptive honeycomb network to ensure large-area uniform coverage and thermal equilibrium. The unique thermal conduction valve structure, acting as a passive intelligent heat regulator, achieves efficient heat exchange while maintaining fluid isolation. Its adaptive reinforcement structure at both macroscopic and microscopic scales, along with wall thickness design optimized inversely to local temperature differences, achieves passive dynamic thermal load balance, significantly improving the system's response speed and stability. Through multiphysics-coupled topology optimization, a globally optimal solution is achieved among fluid resistance, heat conduction, structural strength, and manufacturing feasibility, overcoming the limitations of traditional step-by-step design. Integrating 3D printing constraints directly into the optimization process ensures the precise fabrication of complex biomimetic structures, realizing the integration of design and manufacturing.

[0012] Therefore, this invention achieves a dual-loop collaborative radiator design by combining thermal imaging scanning, temperature gradient threshold zoning, and three-dimensional topology optimization techniques. This design uses a hotspot-driven intelligent zoning mechanism to divide the radiator into a "bronchial" main loop serving high-heat areas and a "cellular network" secondary loop covering low-heat areas, precisely allocating cooling resources. Simultaneously, a thermal conduction valve structure is introduced to achieve efficient heat exchange between the two loop systems, enabling the system to automatically balance the heat load under fluid isolation. This design reduces hotspot temperature by 35% and system pressure drop by 28%, completely solving the problem of low efficiency of traditional radiators under complex heat loads. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the steps of a heat sink cooling channel design method based on three-dimensional topology optimization.

[0014] Figure 2 This is a sample test diagram of the heat sink component of the present invention;

[0015] Figure 3 This is a physical image of the heat sink of the present invention.

[0016] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0019] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0020] To achieve the above objectives, please refer to Figures 1 to 3 This invention provides a heat sink cooling channel design method based on three-dimensional topology optimization, comprising the following steps:

[0021] Step S1: Perform thermal imaging scanning on the electronic device to obtain temperature distribution data and generate a heat map;

[0022] In this embodiment of the invention, an infrared thermal imager is used to perform thermal imaging scans on a 150W central processing unit under various load conditions. To ensure data validity, the stability of the temperature time series data is determined by calculating the ratio of the standard deviation to the mean. If the ratio exceeds 0.05, the sampling frequency is increased and the sampling time is extended for re-acquisition. Based on the stabilized data, a multi-layered information hotspot map containing average temperature, temperature change rate, and heat flow direction is generated. This map is then precisely registered with the CPU's CAD structure diagram through affine transformation to establish a mapping relationship between heat distribution and physical location.

[0023] Step S2: Calculate the temperature gradient based on the heat map, determine the temperature gradient threshold, divide the main loop region and the secondary loop region, perform loop region boundary smoothing optimization, and generate a loop partition map.

[0024] In this embodiment of the invention, a Gaussian filter is applied to the heatspot map for smoothing, and the Sobel operator is used to calculate and generate a temperature gradient distribution map. Then, by combining the 90th percentile of the current gradient distribution with the average critical gradient value in the historical tests of the device, a weighted average is calculated and multiplied by a safety factor of 1.1 to obtain an adaptive temperature gradient threshold G_th = 5.335℃ / mm. Based on this threshold, the heat sink area is divided into primary and secondary loop regions, and morphological opening and closing operations are performed on the generated initial boundaries to eliminate burrs and isolated points, ultimately generating a loop partition map with smooth boundaries.

[0025] Step S3: Based on the cyclic partitioning diagram, determine the inlet location and heat flow requirements, and identify the set of thermal anchor points; determine the hierarchical branching structure of the main loop and the cellular network structure of the secondary loop based on the set of thermal anchor points; generate a channel skeleton diagram based on the hierarchical branching structure and the cellular network structure.

[0026] In this embodiment of the invention, the inlet locations and heat flow requirements (95W and 55W, respectively) of the main and secondary loops are determined based on the loop partitioning diagram. Three critical hotspots with the highest temperatures are identified within the main loop region as a set of thermal anchor points. Based on a cost function weighted by the reciprocal of temperature, the Dijkstra algorithm is used to construct a shortest path tree from the inlet to all anchor points, serving as the main loop trunk. Within the secondary loop region, a cellular channel network is generated by constructing a Voronoi diagram with density adaptively adjusted according to local heat load. Finally, the hierarchical branching structure of the main loop is integrated with the cellular network structure of the secondary loop to generate a channel skeleton diagram containing two complete paths.

[0027] Step S4: Extract the interface location information between the main circulation area and the secondary circulation area from the channel skeleton diagram, determine the basic isolation wall and add the reinforcement structure, use the hotspot map to calculate the local temperature difference on both sides of the interface, optimize the thickness according to the local temperature difference to form the heat conduction valve structure, and generate the heat exchange interface structure.

[0028] In this embodiment of the invention, the trajectory where the distance between the main and secondary circulation channels is less than 3 mm is extracted from the channel skeleton diagram as the interface, and a basic isolation wall with a thickness of 0.5 mm is generated at this location. Based on the local heat flux density at each point on the interface, different reinforcement structures are generated in different regions: in the high heat flux region (>80 W / cm²)... 2 A serrated structure is generated in the medium heat flux region (30-80 W / cm²). 2 A wave-shaped structure is generated, and cylindrical microfin arrays with a positive correlation to heat flux density are added to the surface of all reinforcing structures. Finally, based on the local temperature difference ΔT obtained from simulation, the thickness of the isolation wall is optimized from 0.2 mm to 0.5 mm using an inverse proportional function relationship to form a thermally conductive valve with adaptively varying thermal conductivity, and finally, a complete heat exchange interface structure containing this valve body is generated.

[0029] Step S5: Based on the heat exchange interface structure, complete the structural design of the heat sink channel and generate a 3D printing scheme;

[0030] In this embodiment of the invention, the generated heat exchange interface structure is integrated with an AlSi10Mg aluminum alloy heat sink substrate to form an initial model. The solid isotropic material penalty method (SIMP) is used to perform topology optimization on this model, aiming to minimize thermal resistance, while applying multiple physical constraints such as fluid pressure drop (below 15 kPa), structural stress, material volume fraction, and 3D printing self-support angle (greater than 45 degrees). Three candidate schemes are generated by adjusting the optimization parameters, and after evaluation by thermal flow simulation, the optimal scheme is selected based on a quality factor that integrates performance and power consumption. Finally, the optimal scheme is converted into a 3MF format 3D printing scheme containing complete geometric information, material parameters, and selective laser melting (SLM) process guidance.

[0031] Preferably, step S1 includes the following steps:

[0032] Step S11: Perform thermal imaging scanning on the electronic device to continuously acquire temperature data and continuous sampling data under different power loads;

[0033] Step S12: By calculating the ratio of the standard deviation to the mean of the continuous sampling data, determine whether there is significant fluctuation in the temperature data: when the ratio exceeds the preset continuous sampling threshold, it is determined that there is fluctuation, and the sampling frequency is increased and the sampling time is extended; when the ratio is lower than the preset threshold, it is determined that there is no significant fluctuation, and the original sampling parameters are maintained.

[0034] Step S13: Generate a hotspot map based on temperature data and its rate and direction of change;

[0035] Step S14: Correspond and calibrate the hotspot map with the electronic device structure diagram to establish the mapping relationship between hotspots and physical locations.

[0036] In this embodiment of the invention, thermal imaging scanning is performed on a central processing unit (CPU) with a rated power consumption of 150W. An infrared thermal imager with a resolution of 640×480 pixels and a temperature measurement accuracy of ±1℃ is used to acquire data under three operating conditions: a low load of 20W, a medium load of 90W, and a full load of 150W. After stabilizing under each condition, temperature data is continuously acquired for 60 seconds at a sampling frequency of 10Hz, forming a time-series dataset containing 600 frames of thermal images.

[0037] The stability of the time series data for each pixel in the collected time-series dataset is assessed. The standard deviation σ_T and mean temperature μ_T of each pixel over 60 seconds are calculated, and the continuous sampling ratio R_cs is obtained, calculated as R_cs = σ_T / μ_T, where R_cs is the continuous sampling ratio, σ_T is the standard deviation of temperature for a specific pixel within the sampling period, and μ_T is the mean temperature for that pixel within the sampling period. A continuous sampling threshold of 0.05 is set. If the R_cs of any pixel exceeds 0.05, it is determined that there is significant temperature fluctuation under that condition. The sampling frequency for that condition is automatically increased to 20Hz, and the sampling time is extended to 120 seconds before re-collection. If the R_cs of all pixels is below 0.05, the temperature data is considered stable, and the current dataset is used.

[0038] A multi-layered heat map is generated based on the stabilized temperature dataset. The first layer is the average temperature distribution layer, displaying the average temperature value of each pixel. The second layer is the temperature change rate layer, which calculates the temporal gradient of each pixel by measuring the temperature difference between consecutive frames, representing the rate of heat accumulation. The third layer is the heat flow direction layer, which calculates the spatial gradient of each pixel by measuring the temperature difference between each pixel and its neighboring pixels, representing the direction and intensity of heat diffusion.

[0039] The generated heat map is precisely registered with the CAD 3D structural diagram of the CPU package. First, four vertices of the CPU chip die are extracted from the CAD 3D structural diagram as reference points. Then, the geometric centers of the four corresponding high-temperature regions in the heat map are identified as alignment points. Using an affine transformation algorithm, the 2D pixel coordinate system of the heat map is mapped to the physical 3D coordinate system of the CAD model, ultimately establishing a one-to-one correspondence between each pixel in the heat map and its physical location on the CPU surface.

[0040] Preferably, step S2 includes the following steps:

[0041] Step S21: Preprocess the heat map and calculate the temperature change rate of each point in the heat map to generate a temperature gradient distribution map;

[0042] Step S22: Analyze the characteristics of the temperature gradient distribution map and determine the temperature gradient threshold;

[0043] Step S23: Determine whether the temperature gradient of each region exceeds the temperature gradient threshold: if it does, mark the region as the main loop region; if it does not exceed the threshold, mark the region as the secondary loop region.

[0044] Step S24: Generate a loop partition map based on the boundaries of the main loop region and the secondary loop region.

[0045] In this embodiment of the invention, the heat map is smoothed to eliminate sensor noise, specifically by performing a convolution operation using a 5×5 Gaussian filter. After processing, the Sobel operator is applied to calculate the temperature gradient G at each point on the heat map, and its calculation formula is as follows: Where G_x and G_y are the rates of temperature change at that point in the horizontal and vertical directions, respectively. Based on the gradient values ​​G of all points, a temperature gradient distribution map with the same size as the hotspot map is generated.

[0046] The gradient values ​​of all points in the temperature gradient distribution map are statistically analyzed to generate a gradient value histogram. Simultaneously, the thermal critical point information of this electronic device model is retrieved from its historical test database, determining its historical average gradient value to be 4℃ / mm. The gradient value at the 90th percentile position in the gradient value histogram, i.e., 5℃ / mm, is taken and arithmetically averaged with the historical average gradient value to obtain an initial adaptive threshold value of 4.5℃ / mm. This initial value is then corrected by multiplying it by a safety factor of 1.1, ultimately determining the temperature gradient threshold G_th to be 4.95℃ / mm.

[0047] Iterate through each point in the temperature gradient distribution map and compare its gradient value G with the determined temperature gradient threshold G_th = 4.95℃ / mm. If G at a point is greater than 4.95℃ / mm, mark that point as 1 on the corresponding mask image, representing the main loop region; if G is not greater than 4.95℃ / mm, mark it as 0, representing the secondary loop region.

[0048] Based on the binary mask image generated from all marker points, the boundary contours of regions with values ​​of 1 and 0 are extracted. To eliminate jagged edges and isolated small regions at the partition boundaries, a morphological opening operation with a structuring element radius of 3 pixels is first performed on the mask image, followed by a morphological closing operation with a structuring element radius of 3 pixels. After smoothing, a cyclic partition map with clearly defined boundaries between the main and secondary cyclic regions is finally generated.

[0049] Preferably, step S22 includes:

[0050] Statistically analyze the distribution of gradient values ​​at all points in the temperature gradient distribution map;

[0051] Analyze the heat dissipation critical point information in the historical operating data of electronic devices;

[0052] Based on the gradient value distribution and heat dissipation critical point information, calculate the adaptive threshold of the temperature gradient;

[0053] The temperature gradient threshold is obtained by applying a safety factor to the adaptive threshold.

[0054] In this embodiment of the invention, the gradient values ​​of all pixels in the temperature gradient distribution map are statistically analyzed, and a gradient value histogram with a group interval of 0.1℃ / mm is generated. From the cumulative distribution function of this histogram, the gradient value corresponding to a cumulative probability of 90% is extracted and denoted as G_90. This value represents the upper limit of the gradient in most areas under the current operating conditions. In this embodiment, the value of G_90 is 5.2℃ / mm.

[0055] The historical operating database of this electronic device model was analyzed to extract critical point information from multiple thermal throttling tests. This information represents the maximum temperature gradient in the core area measured one second before the device's performance began to decline due to overheating. The arithmetic mean of all these critical point information was calculated to obtain the historical average critical gradient value G_hist. In this embodiment, the value of G_hist is 4.5℃ / mm.

[0056] The representative value G_90 of the gradient distribution is combined with the representative value G_hist of historical critical point information, and the adaptive threshold G_adap is calculated by weighted averaging. The formula is G_adap = w_1 × G_90 + w_2 × G_hist, where G_adap is the adaptive threshold, G_90 is the representative value of the current operating condition gradient distribution, G_hist is the historical average critical gradient value, and w_1 and w_2 are weighting coefficients, with w_1 + w_2 = 1. In this embodiment, both weights w_1 and w_2 are set to 0.5, therefore the calculated G_adap is 4.85℃ / mm.

[0057] To ensure sufficient redundancy in the coverage area of ​​the main loop region, a safety factor S_f is applied to the adaptive threshold G_adap to obtain the final temperature gradient threshold G_th. The calculation formula is G_th = G_adap × S_f, where S_f is the safety factor. In this embodiment, the safety factor S_f is set to 1.1, and the final calculated temperature gradient threshold G_th is 5.335℃ / mm.

[0058] Preferably, step S24 includes:

[0059] Extract the initial boundary point sets of the main loop region and the secondary loop region;

[0060] Calculate the included angle formed by three adjacent points in the initial boundary point set. When the included angle is less than the preset smoothing threshold, it is identified as a sharp angle; when the rate of change of the distance between adjacent points exceeds 20%, it is identified as a jagged region. If the above situations are detected, curve fitting is performed; if the above situations are not detected, the original boundary shape is maintained.

[0061] Generate a cyclic partitioning map based on the optimized boundaries.

[0062] In this embodiment of the invention, the Moore-Neighbor tracking algorithm is performed on the binarized images of the main loop region and the secondary loop region to extract the pixel coordinates at the boundary of the two regions, forming an ordered initial boundary point set P = {P_1, P_2, ..., P_n}.

[0063] The point set is smoothed. For each point P_i (i from 2 to n-1) in the set, the angle θ_i formed by vectors P_i-1P_i and P_iP_i+1 is calculated. A smoothing threshold of 120 degrees is set; when θ_i is less than 120 degrees, point P_i is marked as a sharp corner. Simultaneously, the distances between adjacent points d_i-1 = distance(P_i-1, P_i) and d_i = distance(P_i, P_i+1) are calculated, along with the rate of change of distance R_d = |d_i - d_i-1| / d_i-1. When R_d exceeds 20%, point P_i is marked as a jagged point. If a marked sharp corner or jagged point is detected after the traversal, a cubic B-spline curve is fitted to the boundary segment containing these marked points. The new point set generated after the fitting replaces the original point set, forming the optimized boundary. If no marked points are detected, the initial boundary point set remains unchanged.

[0064] Based on the optimized boundary point set, the scan line fill algorithm is used to color and distinguish the main loop region and the secondary loop region on the blank canvas, and finally generate a loop partition map with smooth and simplified geometric shape.

[0065] Preferably, step S3 includes the following steps:

[0066] Step S31: Analyze the loop partition diagram to determine the inlet locations and heat flow requirements of the main loop and secondary loop;

[0067] Step S32: Identify key hotspot locations within the main loop region and generate a set of thermal anchor points;

[0068] Step S33: Construct the shortest path tree from the inlet location to the set of thermal anchor points to form the main circulation route map;

[0069] Step S34: Based on the main loop route map, determine the branch channels to cover all hotspots within the main loop area;

[0070] Step S35: Arrange a uniformly distributed honeycomb channel network within the secondary loop region;

[0071] Step S36: Generate a channel skeleton diagram based on the branch channel and cellular channel network.

[0072] In this embodiment of the invention, the cyclic partition map is analyzed, and the heat flux weighted centroids of the main and secondary cyclic regions are calculated. The radiator boundary points farthest from the heat flux weighted centroids of the two regions are determined as the main and secondary cyclic inlets. By performing area integration on the heat flux values ​​of the corresponding regions in the hotspot map, the heat flux to be handled by the main cycle is calculated to be 95W, and the heat flux to be handled by the secondary cycle is calculated to be 55W. On the hotspot map corresponding to the main cyclic region, a local maximum search algorithm is used to identify key hotspots. Specifically, all points whose temperature values ​​are higher than the other 8 points in their 3×3 neighborhood and whose temperatures are higher than 95% of the maximum temperature value of the entire hotspot map are searched. The physical coordinates of these points are extracted to generate a thermal anchor point set containing three coordinate points.

[0073] The main loop region is discretized into a weighted undirected graph, where each pixel is a node. The path cost C between any two adjacent nodes is calculated as C = L × (1 / T_avg), where L is the geometric distance between the two nodes and T_avg is the average temperature of the two nodes. Dijkstra's algorithm is applied to calculate the thermodynamic shortest path from the main loop inlet node to every anchor node in the thermal anchor set, constructing a shortest path tree connecting the inlet to all anchor nodes, forming the main loop trunk path graph. A 2mm influence radius is set for the generated main loop trunk path graph, forming a trunk coverage zone. All points within the main loop region whose temperature exceeds a preset hotspot threshold but are located outside the trunk coverage zone are traversed, and these points are connected to the nearest trunk path, forming branches. Through this operation, a hierarchical branch structure covering all hotspots is generated based on the trunk path graph.

[0074] Within the secondary loop region, an equilateral triangular grid is arranged at 5mm intervals, with the vertices of the grid serving as generation points. Based on these generation points, a Voronoi diagram is constructed; the edges of the Voronoi diagram form a honeycomb-like channel network covering the entire secondary loop region. The hierarchical branching structure of the main loop is superimposed on the honeycomb-like channel network of the secondary loop in the same coordinate system. A base linewidth of 2mm is assigned to the main loop trunk, 1.5mm to the main loop branches, and 1mm to the secondary loop honeycomb channels. Finally, a channel skeleton diagram containing the complete paths and preliminary dimensional information of both loop systems is generated.

[0075] Preferably, step S34 includes:

[0076] Calculate the heat dissipation requirement for each hotspot;

[0077] Determine whether the heat dissipation demand exceeds the preset heat dissipation demand threshold: if it exceeds, it is determined to be a multi-level heat dissipation branch structure; if it does not exceed, it is determined to be a single-level heat dissipation branch structure.

[0078] Calculate the optimal bifurcation angle for each branch channel;

[0079] Set the diameter of the branch channels, decreasing step by step from the main road to the end;

[0080] Micro-turbulence structures are added to hot spots with particularly high heat dissipation requirements. These micro-turbulence structures are arrays of regular or irregular protrusions on the inner wall of the channel, with a characteristic dimension of 5%-15% of the channel diameter.

[0081] In this embodiment of the invention, for each independent hotspot within the main circulation area, its heat dissipation demand value Q is calculated by performing area integration on the heat flux value of its corresponding hotspot map area. For example, the calculated heat dissipation demand values ​​for three hotspots are Q_1 = 40W, Q_2 = 25W, and Q_3 = 30W.

[0082] A heat dissipation demand threshold Q_th is set at 30W. The heat dissipation demand Q of each hotspot is compared with this threshold. For the hotspot Q_1 = 40W, since it exceeds the threshold, a multi-stage heat dissipation branch structure is configured for it, branching off from the main path and then further branching. For the hotspot Q_2 = 25W, since it does not exceed the threshold, a single-stage heat dissipation branch structure is configured for it, directly leading from the main path. For the hotspot Q_3 = 30W, since it equals the threshold, a single-stage heat dissipation branch structure is also configured for it.

[0083] The bifurcation angle is calculated based on the principle of minimizing fluid resistance. For the bifurcation point in a multi-level branch structure, if the heat dissipation demand served by the two downstream sub-branches is equal, the angle between the two sub-branches is set to 75 degrees. If they are not equal, the angle is adjusted according to the flow rate ratio of the two sub-branches, with the branch with the larger flow rate having a smaller deflection angle relative to the main channel.

[0084] Set the diameter of the branch channels. The diameter of the main circulation channel is 2mm. At the first branch point, the flow rate is proportionally allocated according to the total heat dissipation demand of each downstream branch, and the branch diameter is set. For example, the diameter of the primary branch channel is set to 1.6mm. At the next branch point in the multi-level branch structure, the diameter of the secondary branch channel is further reduced to 1.2mm, forming a structure that gradually decreases from the main channel to the end.

[0085] In the hotspot region with a heat dissipation demand of Q_1 = 40W, due to its particularly high demand, a micro-turbulence structure is added to the inner wall of the corresponding end channel with a diameter of 1.2mm. This structure is an alternating array of hemispherical protrusions, each with a diameter of 0.15mm (12.5% ​​of the channel diameter), and the center-to-center spacing between the protrusions in the array is 0.5mm. This is used to disrupt the boundary layer and enhance local heat transfer without significantly increasing the pressure drop.

[0086] Preferably, step S35 includes the following steps:

[0087] Step S351: Determine the number and size of the cellular units based on the area of ​​the secondary loop region;

[0088] Step S352: Set grid control points within the secondary loop area according to the preset grid equidistant values;

[0089] Step S353: Obtain the local heat load, construct a polygonal region centered on the grid control points, and generate a honeycomb channel network along the polygon boundary, wherein the channel width of the honeycomb channel network is set proportionally according to the local heat load;

[0090] Step S354: Check the uniformity of the honeycomb channel network: if it is not uniform, adjust the distribution of grid control points and repeat step S353; if it is uniform, retain the current network structure.

[0091] Step S355: Assign a uniform channel diameter to the cellular channel network.

[0092] In this embodiment of the invention, the total area of ​​the secondary cycle region is calculated to be 1000 mm². 2 The target average coverage area for each cellular unit was set at 25mm. 2 Therefore, the total number of cellular units was determined to be 40. Within the secondary loop region, a uniform rectangular grid was generated within its geometrically bounded rectangle with a preset grid spacing of 5 mm. All grid intersections falling within the secondary loop region were used as the initial set of grid control points. A Voronoi diagram was constructed using the grid control points as generation points, dividing the secondary loop region into multiple sub-regions. For each sub-region, the total local heat load to be borne by that region was calculated by performing area integration on its corresponding heat flux on the heat map.

[0093] Check the heat load uniformity of the network and calculate the standard deviation of the local total heat load in all sub-regions. Set the uniformity criterion as this standard deviation being less than 10% of the average total heat load of the region. If this condition is not met, adjust by moving the grid control points: shift each control point slightly towards its neighboring sub-region with a higher heat load by a step size of 0.2 mm, then return to step S353 to rebuild the network. Repeat this process until the uniformity condition is met, at which point the network becomes denser in areas with high heat loads. If the condition is met, retain the current network structure. For all channel paths of the finally determined, density-adaptive cellular channel network, uniformly assign a circular cross-section with a diameter of 1.2 mm.

[0094] Preferably, step S4 includes the following steps:

[0095] Step S41: Extract the interface position information between the main loop and the secondary loop from the channel skeleton diagram;

[0096] Step S42: Generate a basic isolation wall at the location corresponding to the interface location information;

[0097] Step S43: Obtain the local heat flux density, determine the reinforcing structure on the surface of the basic isolation wall, and increase the heat exchange area to 150%-300% of the original surface area. The increase ratio is determined according to the local heat flux density. The higher the local heat flux density, the higher the increase ratio.

[0098] Step S44: Optimize the thickness of the reinforced structure based on the local temperature difference to form a thermally conductive valve structure;

[0099] Step S45: Generate a heat exchange interface structure based on the thermally conductive valve structure and the fluid channel.

[0100] In this embodiment of the invention, in the channel skeleton diagram, by calculating the Euclidean distance between all point pairs on the main circulation channel path and the secondary circulation channel path, the midpoint trajectory of all point pairs with a distance less than 3mm is extracted to form a continuous three-dimensional spatial curve, which is the interface position information. Using this three-dimensional spatial curve as the centerline, a solid pipe with a basic wall thickness of 0.5mm is generated. The inner and outer surfaces of this pipe will serve as the boundaries of the main circulation and secondary circulation channels, respectively; this solid structure is the basic isolation wall.

[0101] The interface location information is mapped back to a hotspot map, and the local heat flux density q” at each point is sampled at equal intervals. On the two surfaces of the base isolation wall facing the main and secondary circulation channels, an array of fins with isosceles triangular cross-sections perpendicular to the wall is generated as a reinforcement structure. A linear functional relationship is established between the fin height H and the local heat flux density q”, such that the increase in total heat exchange area A_ratio varies between 150% and 300%. Specifically, at the point of minimum heat flux density at the interface (q”_min), the fin height is set to make A_ratio 150%; at the point of maximum heat flux density (q”_max), the fin height is set to make A_ratio 300%.

[0102] The local temperature difference ΔT between the main and secondary circulation channels on both sides of the interface under design conditions was obtained through simulation calculations. The thickness of the base isolation wall containing the reinforced structure was optimized, establishing an inverse proportional function relationship between wall thickness t and the local temperature difference ΔT. The minimum wall thickness t_min was set to 0.2 mm, and the maximum wall thickness t_max to 0.5 mm. At the point of maximum temperature difference (ΔT_max), the wall thickness t was set to 0.2 mm to achieve the highest thermal conductivity; at the point of minimum temperature difference (ΔT_min), the wall thickness t was set to 0.5 mm. Through this operation, a thermally conductive valve structure with adaptive thermal conductivity based on the local temperature difference was formed.

[0103] The model of the primary and secondary circulation fluid channels with defined diameters is combined with the generated thermally conductive valve structure model using Boolean operations. Specifically, the thermally conductive valve structure is combined with a solid radiator substrate, and then the volumes of the primary and secondary circulation fluid channels are subtracted from the substrate to generate a three-dimensional solid model of the heat exchange interface structure containing complete flow channels and optimized interfaces.

[0104] Preferably, step S44 includes the following steps:

[0105] Establish an inverse proportional function relationship between wall thickness and local temperature difference;

[0106] Mapping analysis of the local temperature difference in each region corresponding to the interface location information is performed.

[0107] Determine if the temperature difference exceeds the critical threshold: if it does, determine the adaptive deformation structure; if it does not, maintain the basic thickness.

[0108] Verify the performance of the thermally conductive valve structure under bidirectional heat flow conditions.

[0109] In this embodiment of the invention, a piecewise linear inverse proportional function relationship is established between the isolation wall thickness t and the local temperature difference ΔT. The minimum temperature difference ΔT_min is set to 5℃, corresponding to a maximum wall thickness t_max = 0.5mm; the maximum temperature difference ΔT_max = 20℃, corresponding to a minimum wall thickness t_min = 0.2mm. When ΔT is between 5℃ and 20℃, the wall thickness t linearly decreases from 0.5mm to 0.2mm.

[0110] Numerical simulations coupling fluid dynamics and heat conduction were used to perform temperature difference mapping analysis at the interface. The simulation was set with a main circulation inlet temperature of 25℃ and a flow rate of 1.5 L / min, and a secondary circulation inlet temperature of 30℃ and a flow rate of 0.8 L / min. After the simulation, the temperatures of corresponding nodes on the main and secondary circulation sides were extracted at the interface, and the local temperature difference ΔT at each point on the interface was calculated, forming a temperature difference distribution map covering the entire interface.

[0111] A key temperature difference threshold ΔT_th is set to 8℃. Each point on the interface is iterated over to determine if its local temperature difference ΔT exceeds this threshold. If ΔT is not greater than 8℃, the wall thickness at that point remains at a basic thickness of 0.5mm. If ΔT is greater than 8℃, the wall thickness at that point is calculated based on the established piecewise linear inverse proportional function, thereby determining an adaptive geometry with non-uniform thickness.

[0112] To ensure the functional integrity of the structure, simulation verification of its bidirectional heat flow performance was conducted. A new simulation condition was established, in which the main circulation inlet temperature was maintained at 25℃, and the secondary circulation inlet temperature was instantaneously increased to 45℃ to simulate a heat source in the secondary circulation region. Under this condition, the reverse heat flow transfer coefficient from the secondary circulation to the main circulation was calculated, and its value was confirmed to be no less than 150 W / (m²). 2 •K), to ensure that the thermal conduction valve structure has efficient heat transfer capability under bidirectional heat flow conditions.

[0113] Of particular importance, step S43 includes:

[0114] The reinforcement structure is determined based on the local heat flux density. The reinforcement structure adopts a sawtooth structure in the high heat flux region, a wavy structure in the medium heat flux region, and a planar contact structure in the low heat flux region.

[0115] Calculate the geometric parameters of the reinforced structure to maximize the contact area while ensuring structural compactness;

[0116] A micro heat dissipation fin array was determined on the surface of the reinforced structure.

[0117] In this embodiment of the invention, the interface is partitioned based on the local heat flux density q” at each point, and the type of reinforcement structure is determined. Two heat flux density thresholds are set: a high heat flux threshold q”_high = 80 W / cm 2 and low heat flux threshold q"_low=30W / cm 2 Traverse all regions of the interface. If q” is greater than q”_high in a certain region, generate a sawtooth structure on the surface of the basic isolation wall in that region; if q” is between q”_low and q”_high, generate a wavy structure; if q” is less than q”_low, maintain its original planar contact structure.

[0118] Calculate the specific geometric parameters of the reinforced structure. For the serrated structure region, its tooth height H_s (in mm) is related to the local heat flux density q (in W / cm²). 2 The relationship between H_s and q” is H_s = 0.01 × (q” - q”_high) + H_s_base, where H_s_base is the base tooth height of 0.5 mm and the tooth tip angle is set to 60 degrees. For the wavy structure region, the relationship between its amplitude A_w (in mm) and q” is A_w = 0.008 × (q” - q”_low) + A_w_base, where A_w_base is the base amplitude of 0.2 mm and the wavelength λ_w is fixed at 2 mm.

[0119] A micro-fin array was defined on all fluid-facing surfaces of the existing serrated, wavy, and planar structures. The array consisted of cylindrical microfins, each with a fixed diameter of 0.1 mm. The height h_pin (mm) and number density ρ_pin (fins / mm) of the microfins were also defined. 2 Both are related to the local heat flux density q (unit: W / cm²). 2 There is a positive correlation. Specifically, h_pin = 0.002 × q”, ρ_pin = 0.1 × q”. Through this operation, the heat transfer area at the microscale is maximized based on the macroscopically enhanced structure.

[0120] Of particular importance, step S5 includes the following steps:

[0121] Step S51: Integrate the heat exchange interface structure with the heat sink structure;

[0122] Step S52: Perform topology optimization on the integrated structure, and analyze fluid resistance, heat transfer efficiency, structural strength and manufacturing feasibility.

[0123] Step S53: Generate multiple optimization candidate solutions and evaluate their performance through thermal flow simulation;

[0124] Step S54: Determine whether the optimization candidate scheme meets the heat dissipation target: if it does, select the optimal scheme as the final design; if it does not, adjust the optimization parameters and repeat steps S52 to S53.

[0125] Step S55: Convert the final design into a 3D printing solution, including complete geometric information, material parameters, and manufacturing process guidelines.

[0126] In this embodiment of the invention, the generated heat exchange interface structure is integrated with a solid aluminum alloy (AlSi10Mg) heat sink substrate with dimensions of 100mm × 80mm × 5mm. Using Boolean operations in 3D modeling, the solid portion of the heat exchange interface structure is first joined with the substrate. Then, the volumes of the main and secondary circulation fluid channels are subtracted from the merged solid to form a single solid heat sink initial model containing complete internal flow channels.

[0127] Topology optimization was performed on the integrated initial model. The Solid Isotropic Material Penalty (SIMP) algorithm was adopted, with the goal of minimizing the overall thermal resistance of the heat sink, and four constraints were imposed: 1) the total pressure drop of the main loop and the secondary loop does not exceed 15 kPa; 2) the total volume of the material does not exceed 40% of the design space; 3) the maximum Mises stress under an internal pressure of 1 MPa is less than 50% of the material yield strength; 4) the minimum self-support angle of all internal overhanging structures is greater than 45 degrees to ensure the feasibility of 3D printing.

[0128] By adjusting the weighting coefficients of thermal resistance and flow resistance objectives in the topology optimization (0.7:0.3, 0.5:0.5, 0.3:0.7 respectively), three optimization cycles were run to generate three candidate schemes with different structural morphologies. High-precision thermo-fluid-structure interaction simulations were performed on each candidate scheme to accurately calculate its maximum temperature, total pressure drop, and structural stress distribution under a total heat load of 150W.

[0129] The simulation results are compared with the preset heat dissipation targets (maximum temperature below 85℃ and total pressure drop below 15kPa). If none of the three schemes meet the targets, return to step S52, increase the material volume fraction constraint by 5%, and optimize again. If any scheme meets the targets, select the optimal scheme with the highest quality factor (FOM) as the final design, where FOM is calculated as FOM = (1 / maximum temperature) / (main circulation pumping power + secondary circulation pumping power).

[0130] The finalized optimal solution was converted into a 3MF format 3D printing solution. This solution not only includes precise geometric mesh information but also comes with a manufacturing process guide, which explicitly specifies: the manufacturing material is AlSi10Mg aluminum alloy powder, the manufacturing process is selective laser melting (SLM), the printing layer thickness is 30μm, the laser power is 350W, the scanning speed is 1200mm / s, and stress-relief annealing and wire cutting to remove the internal support structure are required after printing.

[0131] Please see Figure 2 This is a sample test diagram of the heat sink component of the present invention, wherein the test method is as follows:

[0132] a. Using a cross-cutting tool with a blade width of about 10mm-12mm, make small square grids horizontally and vertically on the surface of the test sample, with 1mm intervals. Each line should be deep enough to reach the substrate.

[0133] b. Use a soft brush to gently brush the surface of the test sample 5 times backward and 5 times forward along the two diagonals of the grid pattern.

[0134] c. Firmly adhere the small grid to be tested with 3M tape, and rub the tape vigorously with an eraser to increase the contact area and force between the tape and the area being tested;

[0135] d. Grab one end of the tape and quickly tear it off vertically (90°), then observe the surface with a magnifying glass.

[0136] Please see Figure 3 This is a diagram of the heat sink of the present invention.

[0137] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0138] Test specifications: The tested surface has a small amount of coating peeling off at the intersection of the cuts, but the area affected by the cross-cuts should not be significantly greater than 5%. If this requirement is met, the sample passes the test.

[0139] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A heat sink cooling channel design method based on three-dimensional topology optimization, characterized in that, The method includes the following steps: Step S1: Perform thermal imaging scanning on the electronic device to obtain temperature distribution data and generate a heat map; Step S2: Calculate the temperature gradient based on the heat map, determine the temperature gradient threshold, divide the main loop region and the secondary loop region, perform loop region boundary smoothing optimization, and generate a loop partition map. Step S2 includes: Step S21: Preprocess the heat map and calculate the temperature change rate of each point in the heat map to generate a temperature gradient distribution map; Step S22: Analyze the characteristics of the temperature gradient distribution map and determine the temperature gradient threshold. Step S22 includes: Statistically analyze the distribution of gradient values ​​at all points in the temperature gradient distribution map; Analyze the heat dissipation critical point information in the historical operating data of electronic devices; Based on the gradient value distribution and heat dissipation critical point information, calculate the adaptive threshold of the temperature gradient; The temperature gradient threshold is obtained by applying a safety factor to the adaptive threshold. Step S23: Determine whether the temperature gradient of each region exceeds the temperature gradient threshold: if it does, mark the region as the main loop region; if it does not exceed the threshold, mark the region as the secondary loop region. Step S24: Generate a loop partitioning diagram based on the boundaries of the main loop region and the secondary loop region; Step S3: Based on the cyclic partitioning diagram, determine the inlet location and heat flow requirements, and identify the set of thermal anchor points; determine the hierarchical branching structure of the main loop and the cellular network structure of the secondary loop based on the set of thermal anchor points; generate a channel skeleton diagram based on the hierarchical branching structure and the cellular network structure. Step S4: Extract the interface location information between the main circulation area and the secondary circulation area from the channel skeleton diagram, determine the basic isolation wall and add the reinforcement structure, calculate the local temperature difference on both sides of the interface using the hotspot map, optimize the thickness according to the local temperature difference to form the heat conduction valve structure, and generate the heat exchange interface structure. Step S5: Based on the heat exchange interface structure, complete the structural design of the heat sink channel and generate a 3D printing scheme.

2. The radiator cooling channel design method based on three-dimensional topology optimization according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Perform thermal imaging scanning on the electronic device to continuously acquire temperature data and continuous sampling data under different power loads; Step S12: By calculating the ratio of the standard deviation to the mean of the continuous sampling data, determine whether there is significant fluctuation in the temperature data: when the ratio exceeds the preset continuous sampling threshold, it is determined that there is fluctuation, and the sampling frequency is increased and the sampling time is extended; when the ratio is lower than the preset threshold, it is determined that there is no significant fluctuation, and the original sampling parameters are maintained. Step S13: Generate a hotspot map based on temperature data and its rate and direction of change; Step S14: Correspond and calibrate the hotspot map with the electronic device structure diagram to establish the mapping relationship between hotspots and physical locations.

3. The radiator cooling channel design method based on three-dimensional topology optimization according to claim 2, characterized in that, Step S24 includes: Extract the initial boundary point sets of the main loop region and the secondary loop region; Calculate the included angle formed by three adjacent points in the initial boundary point set. When the included angle is less than the preset smoothing threshold, it is identified as a sharp angle. When the rate of change of the distance between adjacent points exceeds 20%, it is identified as a jagged region. If a sharp angle or jagged region is detected in the initial boundary point set, curve fitting is performed. If no sharp angle or jagged region is detected, the original boundary shape is maintained. Generate a cyclic partitioning map based on the optimized boundaries.

4. The radiator cooling channel design method based on three-dimensional topology optimization according to claim 3, characterized in that, Step S3 includes the following steps: Step S31: Analyze the loop partition diagram to determine the inlet locations and heat flow requirements of the main loop and secondary loop; Step S32: Identify key hotspot locations within the main loop region and generate a set of thermal anchor points; Step S33: Construct the shortest path tree from the inlet location to the set of thermal anchor points to form the main circulation route map; Step S34: Based on the main loop route map, determine the branch channels to cover all hotspots within the main loop area; Step S35: Arrange a uniformly distributed honeycomb channel network in the secondary loop region; Step S36: Generate a channel skeleton diagram based on the branch channel and cellular channel network.

5. The radiator cooling channel design method based on three-dimensional topology optimization according to claim 4, characterized in that, Step S34 includes: Calculate the heat dissipation requirement for each hotspot; Determine whether the heat dissipation demand exceeds the preset heat dissipation demand threshold: if it exceeds, it is determined to be a multi-level heat dissipation branch structure; if it does not exceed, it is determined to be a single-level heat dissipation branch structure. Calculate the optimal bifurcation angle for each branch channel; Set the diameter of the branch channels, decreasing step by step from the main road to the end; Micro-turbulence structures are added to hot spots with particularly high heat dissipation requirements. These micro-turbulence structures are arrays of regular or irregular protrusions on the inner wall of the channel, with a characteristic size of 5%-15% of the channel diameter.

6. The radiator cooling channel design method based on three-dimensional topology optimization according to claim 5, characterized in that, Step S35 includes the following steps: Step S351: Determine the number and size of the cellular units based on the area of ​​the secondary loop region; Step S352: Set grid control points within the secondary loop area according to the preset grid equidistant values; Step S353: Obtain the local heat load, construct a polygonal region centered on the grid control points, and generate a honeycomb channel network along the polygon boundary, wherein the channel width of the honeycomb channel network is set proportionally according to the local heat load; Step S354: Check the uniformity of the honeycomb channel network: if it is not uniform, adjust the distribution of grid control points and repeat step S353; if it is uniform, retain the current network structure. Step S355: Assign a uniform channel diameter to the cellular channel network.

7. The radiator cooling channel design method based on three-dimensional topology optimization according to claim 6, characterized in that, Step S4 includes the following steps: Step S41: Extract the interface position information between the main loop and the secondary loop from the channel skeleton diagram; Step S42: Generate a basic isolation wall at the location corresponding to the interface location information; Step S43: Obtain the local heat flux density, determine the reinforcing structure on the surface of the basic isolation wall, and increase the heat exchange area to 150%-300% of the original surface area. The increase ratio is determined according to the local heat flux density. The higher the local heat flux density, the higher the increase ratio. Step S44: Optimize the thickness of the reinforced structure based on the local temperature difference to form a thermally conductive valve structure; Step S45: Generate a heat exchange interface structure based on the thermally conductive valve structure and the fluid channel.

8. The radiator cooling channel design method based on three-dimensional topology optimization according to claim 7, characterized in that, Step S44 includes the following steps: Establish an inverse proportional function relationship between wall thickness and local temperature difference; Mapping analysis of the local temperature difference in each region corresponding to the interface location information is performed. Determine if the temperature difference exceeds the critical threshold: if it does, determine the adaptive deformation structure; if it does not, maintain the basic thickness. Verify the performance of the thermally conductive valve structure under bidirectional heat flow conditions.

Images