A computer CPU heat dissipation optimization device and optimization method

Abstract

The application discloses a computer CPU heat dissipation optimization device and optimization method, which comprises the following steps: generating a CPU heat dissipation multi-source boundary condition set; obtaining an initial air duct topology flow field model; generating a CPU heat dissipation flow field bottleneck diagnosis set; forming a CPU heat dissipation collaborative optimization variable set; initializing a shark optimization algorithm population based on the CPU heat dissipation collaborative optimization variable set to obtain a shark optimization candidate topology population set; performing shark optimization iterative search based on the shark optimization candidate topology population set to obtain a shark optimization air duct candidate solution set; and performing adaptive air duct topology reconstruction based on the shark optimization air duct candidate solution set to generate an adaptive reconstructed air duct topology set. The application avoids the problem that local optimization is effective but the overall flow field is deteriorated in the traditional method, and can still maintain stable heat dissipation performance improvement under a complex heat source distribution scene.

Description

Technical Field

[0001] This invention relates to the technical field of computer CPU heat dissipation optimization device and optimization method, and more particularly to a computer CPU heat dissipation optimization device and optimization method. Background Technology

[0002] As CPU power density and transient load fluctuations continue to increase, cooling systems place higher demands on chassis airflow organization, heat exchange efficiency of heat sink fins, and directional airflow capability in hot spot areas. Existing CPU cooling designs typically rely on fixed airflow structures, empirical fan speed control, or single simulation verification, which fails to fully reflect the fluid-structure-thermal coupling relationship between the CPU package, thermal interface layer, heatsink base, heatsink fins, fan, and chassis air inlet and outlet. In actual operation, the compact internal structure of the chassis and complex airflow paths easily lead to problems such as insufficient penetration velocity of the heatsink fins, localized hot air stagnation, airflow recirculation, and continuous concentration of hot spots on the CPU core.

[0003] Traditional thermal optimization methods often rely on single temperature or fan power consumption metrics, lacking a comprehensive evaluation of the synergistic relationship between CPU core maximum temperature, average junction temperature, thermal resistance, airflow uniformity, and changes in airflow topology. For reconfigurable structures such as adjustable baffles, bypass ducts, guide angles, and fin gaps, existing technologies typically lack topological parameter representation and dynamic filtering mechanisms for addressing flow field bottlenecks. This leads to optimization results that are effective in localized cooling but result in increased overall pressure drop, higher fan power consumption, or exacerbated recirculation vortices. Summary of the Invention

[0004] One objective of this invention is to provide a computer CPU heat dissipation optimization device and method. This invention avoids the problem of traditional methods where local optimization is effective but the overall flow field deteriorates, maintaining stable heat dissipation performance even in complex heat source distribution scenarios.

[0005] A method for optimizing computer CPU heat dissipation according to an embodiment of the present invention includes: Collect structural boundary data, operational heat source data, and environmental status data of the CPU cooling system, perform preprocessing, and generate a set of multi-source boundary conditions for CPU cooling. Based on the set of multi-source boundary conditions for CPU heat dissipation, a fluid-structure-thermal coupling computational domain for CPU heat dissipation is constructed to obtain the initial airflow topology model. Based on the initial duct topology flow field model, the baseline flow field is solved to obtain the temperature field, velocity field and vorticity field corresponding to the CPU cooling system. The bottleneck features of the flow field are extracted according to the CPU core hot spot temperature gradient, the heat sink fin penetration wind speed and the duct return flow intensity, and a CPU cooling flow field bottleneck diagnosis set is generated. Based on the CPU heat dissipation flow field bottleneck diagnosis set, a reconfigurable parameter vector of air duct topology is constructed, and combined with air duct size constraints, fan power consumption constraints and CPU safe temperature constraints, a set of CPU heat dissipation collaborative optimization variables is formed. The population of the rare trevally optimization algorithm is initialized based on the set of CPU heat dissipation collaborative optimization variables. Each rare trevally individual is encoded as a candidate airflow topology reconstruction scheme. A multi-objective fitness function is constructed based on the highest CPU core temperature, average CPU junction temperature, heat dissipation thermal resistance and airflow uniformity to obtain the set of rare trevally candidate topology populations. Based on the candidate topology population set of the rare trevally, perform the rare trevally optimization iterative search, so that the leading rare trevally migrates towards the air duct topology region with low thermal resistance and low pressure loss, so that the following rare trevally gathers towards the high heat transfer region corresponding to the CPU core hot spot according to the hot spot neighborhood guidance weight, so that the disturbing rare trevally performs topology jump search according to the local eddy residence time and air duct return intensity, and obtain the candidate solution set of the rare trevally optimized air duct; Based on the candidate solution set of the optimized air duct, an adaptive air duct topology reconstruction is performed to generate an adaptive reconstructed air duct topology set.

[0006] Optionally, the preprocessing includes coordinate unification, time alignment, unit normalization, and outlier data removal.

[0007] Optionally, the construction of the CPU heat dissipation fluid-structure-thermal coupling computational domain based on the set of multi-source boundary conditions for CPU heat dissipation includes: Based on the set of multi-source boundary conditions for CPU heat dissipation, spatial coordinate registration and boundary contour closure checks are performed on each structural object to generate CPU heat dissipation structural boundary registration results. Based on the boundary registration results of the CPU heat dissipation structure, the CPU heat dissipation system is divided into regions according to the solid heat conduction region, the fluid flow region and the adjustable flow conduction region, and the CPU heat dissipation partition calculation domain is obtained. Based on the CPU heat dissipation partition computational domain, a set of CPU heat dissipation flow solid-thermal coupling boundary attributes is formed; Based on the set of boundary properties of CPU heat dissipation flow-structure-thermal coupling, the CPU package, thermal interface layer, heat sink base, heat sink fins, fan, chassis air inlet, chassis air outlet and reconfigurable air duct components are abstracted as flow field topology nodes, and each flow field topology node is kept in a one-to-one correspondence with its corresponding physical structure region, material properties, boundary properties and spatial position, so as to obtain the set of CPU heat dissipation flow field topology nodes. Based on the CPU heat dissipation flow field topology node set, airflow connection edges are established between the fan, chassis air inlet, heat sink fin gap, reconfigurable air duct assembly and chassis air outlet according to the adjacent airflow relationship. Thermal contact connection edges are established between the CPU package, thermal interface layer, heat sink base and heat sink fin according to the thermal contact relationship. Thermal constraint connection edges are established between the reconfigurable air duct assembly and adjacent fluid flow areas according to the flow constraint relationship, thus obtaining the CPU heat dissipation topology connection edge set. Based on the CPU heat dissipation topology connection edge set, each topology connection edge is labeled with its corresponding flow direction, passability, heat transfer path, airflow blockage status, baffle constraint status, and airflow influence range. Consistency screening is performed on conflicting topology connection edges to ensure that airflow connection edges, thermal contact connection edges, and airflow constraint connection edges in the same spatial area do not have overlapping physical meanings, thus obtaining the CPU heat dissipation topology constraint relationship set. Based on the set of CPU heat dissipation topology constraints, the CPU heat dissipation partition computational domain, the CPU heat dissipation flow field topology node set, and the CPU heat dissipation topology connection edge set are associated and encapsulated. The association and encapsulation results are used as input for subsequent baseline flow field solution to obtain the initial air duct topology flow field model.

[0008] Optionally, the baseline flow field solution based on the initial duct topology flow field model includes: The initial attitude is read based on the initial duct topology flow field model, and the initial attitude is limited to the duct structure state before the search of the Zhenyu optimization algorithm is performed, and the CPU heat dissipation baseline solution input set is generated. Based on the CPU heat dissipation baseline solution input set, thermal conductivity solution conditions are configured for the solid regions where the CPU package, thermal interface layer, heat sink base and heat sink fins are located, and flow solution conditions are configured for the fluid regions where the fan, chassis inlet, chassis outlet, heat sink fin gap and reconfigurable air duct components are located. Heat exchange relationship is set at the interface between the solid region and the fluid region to obtain the CPU heat dissipation baseline fluid-solid-thermal coupling solution conditions. Based on the baseline fluid-structure-thermal coupling solution conditions for CPU heat dissipation, the baseline flow field solution is performed to obtain the temperature field results of the CPU heat dissipation system under the initial air duct topology flow field model. Based on the temperature field results, the airflow state of the fan outlet neighborhood, heat sink inlet area, heat sink middle channel, heat sink tail area, chassis air outlet neighborhood, and reconfigurable air duct component neighborhood is further extracted to determine the airflow direction, airflow speed, airflow stagnation area, and airflow bypass area in each area, thus obtaining the velocity field results corresponding to the CPU cooling system. Based on the velocity field results, the vortex field results are obtained by identifying the vortex field in the fan outlet impact area, the heat sink fin gap shear area, the airflow guide component backwind area, the chassis wall near-wall area, and the air outlet confluence area. Based on the vortex field results, the local vortex location, vortex duration range, vortex retention tendency, and vortex interference degree of the main flow channel are determined, and the vortex field results corresponding to the CPU cooling system are obtained. Based on the temperature field results, the highest temperature location, hot spot distribution range, temperature change trend of hot spot diffusion to the heat dissipation interface layer, and temperature change trend of hot spot transmission to the heat sink base in the CPU core area are extracted. It is then determined whether the CPU core hot spot temperature gradient is concentrated in a local area, and the area where the temperature gradient is concentrated is marked as the hot spot heat conduction bottleneck area, thus obtaining the CPU core hot spot temperature gradient characteristics. Based on the velocity field results, the airflow penetration state from the inlet to the outlet of the heat sink fins is extracted to determine whether the fan airflow effectively enters the gap between the heat sink fins, and to determine whether there are airflow short circuits, local low speeds, and insufficient heat exchange at the tail between the heat sink fins. The fin channels with insufficient penetration airflow and corresponding high temperatures are marked as fin heat exchange bottleneck areas, and the characteristics of heat sink fin penetration airflow are obtained. Based on the vorticity field and velocity field results, the return flow location, return flow direction, return flow range, and degree of interference of the return flow on the main air intake channel are extracted. It is determined whether there is reverse airflow circulation on the leeward side of the reconfigurable duct component, the corner of the chassis, and the leading edge of the air outlet. The area with high return flow intensity and obstruction of hot air exhaust is marked as the duct return flow bottleneck area, and the duct return flow intensity characteristics are obtained. Based on the CPU core hot spot temperature gradient characteristics, heat sink fin penetration wind speed characteristics, and airflow return intensity characteristics, the flow field bottleneck characteristics are merged according to the heat source side bottleneck, heat exchange side bottleneck, and exhaust side bottleneck to generate a CPU heat dissipation flow field bottleneck diagnosis set.

[0009] Optionally, the construction of the reconfigurable parameter vector for the airflow topology based on the CPU heat dissipation flow field bottleneck diagnosis set includes: Based on the CPU heat dissipation flow field bottleneck diagnosis set, the spatial location, bottleneck type, influence range and corresponding topological connection edge of the hot spot heat conduction bottleneck area, fin heat exchange bottleneck area and air duct return flow bottleneck area are read, and each type of bottleneck is associated with the flow field topology node and topological connection edge in the initial air duct topology flow field model to generate flow field bottleneck topology location results. Based on the topological location results of the flow field bottleneck, the set of regions affected by airflow duct reconstruction is obtained; Based on the set of the reconfigurable regions of the airway, a reconfigurable parameter vector of the airway topology is constructed, and the reconfigurable parameter results of the airway topology are obtained. Based on the reconfigurable parameters of the duct topology, duct size constraints are set to obtain the duct size constraint results. Fan power consumption constraints are set based on the reconfigurable parameters of the air duct topology, and the fan power consumption constraint results are obtained. Based on the reconfigurable parameters of the airflow topology, CPU safety temperature constraints are set to ensure that the maximum CPU core temperature, average CPU junction temperature, and hot spot duration are all within the safe operating range allowed by the CPU cooling system. Furthermore, any candidate airflow topology reconfiguration scheme must not cause the CPU core hot spot temperature to exceed the upper limit of the safety temperature at the expense of reducing local pressure loss during the optimization process, thus obtaining the CPU safety temperature constraint results. Based on the constraints of airflow size, fan power consumption, and CPU safe temperature, feasibility marking is performed on each parameter in the airflow topology reconfigurable parameter results. Parameters that satisfy all constraints are marked as searchable variables, parameters that violate any constraint are marked as prohibited variables, and parameters that are close to the constraint boundary but are still adjustable are marked as boundary variables, thus generating a set of CPU heat dissipation co-optimization variables.

[0010] Optionally, the initialization of the rare trevally optimization algorithm population based on the CPU heat dissipation collaborative optimization variable set includes: Based on the CPU heat dissipation collaborative optimization variable set, the reconfigurable parameters of the air duct topology marked as searchable variables and boundary variables are read, and the reconfigurable parameters of the air duct topology marked as prohibited variables are excluded. The flow guide angle parameter, diffusion angle parameter, baffle opening and closing parameter, bypass air duct width parameter, fin gap correction parameter, fan speed control parameter, and hot spot neighborhood flow guide weight parameter are sorted according to the corresponding flow field topology nodes and topology connection edges to generate the individual coding variable sequence of the rare trevally. Based on the individual coding variable sequence of the rare trevally, an individual expression method for candidate air duct topology reconstruction schemes is established, so that each rare trevally individual contains a complete set of guide angle adjustment state, diffusion angle adjustment state, baffle opening and closing state, bypass air duct width state, fin gap correction state, fan speed control state, and hot spot neighborhood guide weight state. Each rare trevally individual can uniquely correspond to a candidate air duct topology reconstruction scheme that can be verified by fluid-structure-thermal coupling, and the coding result of the rare trevally individual candidate scheme is obtained. Based on the coding results of the candidate schemes for the rare trevally individuals, population initialization is performed. Within the allowable range of air duct size constraints, fan power consumption constraints, and CPU safe temperature constraints, multiple rare trevally individuals are generated. These multiple rare trevally individuals are differentiated in terms of guide angle, bypass duct width, baffle opening and closing state, fan speed control amount, and hot spot neighborhood guide weight, so as to cover three candidate reconstruction directions: hot spot guide enhancement, fin penetration improvement, and backflow suppression, to obtain the initial rare trevally topological population results. Based on the initial rare trevally topological population results, the candidate wind tunnel topology reconstruction scheme corresponding to each rare trevally individual is mapped back to the initial wind tunnel topology flow field model. The corresponding topology connecting edges are marked with state tags for expansion, contraction, opening and closing, bypass, guide angle correction or fin gap correction, and the flow field topology nodes and topology connecting edges that have not been changed by the rare trevally individual are kept in their original state, thus obtaining a set of candidate wind tunnel topology flow field models. The candidate air duct topology reconstruction schemes are evaluated based on the candidate air duct topology flow field model set. The maximum CPU core temperature, average CPU junction temperature, heat dissipation thermal resistance and airflow uniformity under each candidate air duct topology reconstruction scheme are obtained, and a multi-objective evaluation index set of candidate schemes is obtained. A multi-objective fitness function was constructed based on a set of multi-objective evaluation indicators for candidate schemes to obtain the fitness evaluation results of individual rare trevally. Based on the fitness evaluation results of individual rare trevally, the rare trevally individuals in the initial rare trevally topological population results are sorted, labeled and archived. Rare trevally individuals that meet the constraints and have better fitness evaluation results are retained. At the same time, the candidate windway topology reconstruction scheme, variable value status, topology connection edge change status and multi-objective evaluation index corresponding to each rare trevally individual are recorded to obtain the rare trevally candidate topological population set.

[0011] Optionally, the step of performing rare trevally optimization iterative search based on the rare trevally candidate topological population set includes: Based on the candidate topology population set of rare trevally, the candidate air duct topology reconstruction scheme, multi-objective fitness function value, maximum CPU core temperature, average CPU junction temperature, thermal resistance, airflow uniformity, air duct topology connection edge status and constraint satisfaction status corresponding to each rare trevally individual are read, and rare trevally individuals that do not meet the air duct size constraint, fan power consumption constraint or CPU safe temperature constraint are removed from the current iteration valid search objects to generate the current valid rare trevally search population; Based on the current effective search population of rare trevally, the rare trevally individuals are sorted from best to worst according to their multi-objective fitness function values. Rare trevally individuals with fitness function values ​​in the advantageous range and simultaneously have lower maximum CPU core temperature, lower average CPU junction temperature, and lower thermal resistance are classified as leading rare trevally individuals. Rare trevally individuals with fitness function values ​​in the middle range and still have room for improvement are classified as following rare trevally individuals. Rare trevally individuals with fitness function values ​​in the disadvantageous range or with high corresponding airflow intensity are classified as disturbing rare trevally individuals. The results of the rare trevally individual role classification are obtained. Based on the role classification of the individual rare tuna, the state of the topology connection edge that contributes to the reduction of heat dissipation thermal resistance and the reduction of airflow pressure loss in its candidate airflow topology reconstruction scheme is read. The guide angle, diffusion angle, bypass airflow width and baffle opening and closing state are adjusted around the airflow topology regions corresponding to the heat dissipation fin inlet, heat dissipation fin tail, chassis air outlet neighborhood and low resistance exhaust path, so that the leading rare tuna individual migrates towards the airflow topology region with low thermal resistance and low pressure loss, and the migration result of the leading rare tuna is obtained. Based on the following trevally individuals in the individual role classification results, the heat spot neighborhood flow guidance weight and the high heat transfer area location corresponding to the CPU core heat spot are read. The candidate airflow topology reconstruction scheme corresponding to the following trevally individuals is moved closer to the flow guidance state related to the heat spot temperature drop in the leading trevally individuals. The flow guidance parameters close to the CPU core heat spot, above the heat sink base, the heat sink fin inlet, and the heat spot neighborhood channel are adjusted first, so that the following trevally individuals gather in the high heat transfer area corresponding to the CPU core heat spot according to the heat spot neighborhood flow guidance weight, and the following trevally gathering result is obtained. Based on the disturbed trevally individuals in the individual role classification results, the local vortex residence time and airflow recirculation intensity in the corresponding candidate airflow topology reconstruction scheme are read. The topology connection edges that cause hot air retention, reverse circulation of the main channel, or poor exhaust at the tail of the heat sink fins are identified. The state jump is performed on the topology connection edges, adjusting the bypass airflow duct that was originally in the closed state to the open candidate state, adjusting the baffle that originally blocked the exhaust to the weakening candidate state, and adjusting the guide angle that originally generated the recirculation to the correction candidate state, thus obtaining the disturbed trevally topology jump results. Based on the migration results of the leading trevally, the aggregation results of the following trevally, and the topology jump results of the disturbed trevally, a constraint consistency check is performed on the guide angle state, baffle opening and closing state, bypass ventilation duct width state, and fan speed control state that may conflict within the same trevally individual. This ensures that the optimized trevally individual still satisfies the duct size constraint, fan power consumption constraint, and CPU safe temperature constraint. Unrealizable or contradictory topology change states are deleted, resulting in the trevally iteratively updated population results. Based on the iterative update of the rare trevally population results, the multi-objective fitness function values ​​of each rare trevally individual are re-evaluated, and candidate airflow topology reconstruction schemes that can reduce the maximum CPU core temperature, reduce the average CPU junction temperature, reduce heat dissipation thermal resistance, or improve airflow uniformity are retained and archived as optimization candidate solutions for the current iteration, thus obtaining the rare trevally optimized airflow candidate solution set.

[0012] Optionally, the adaptive airway topology reconstruction based on the optimized airway candidate solution set includes: Based on the Zhenyu optimization of the candidate solution set for the air duct, the topology connection edge change status, CPU core maximum temperature change status, CPU average junction temperature change status, heat dissipation thermal resistance change status, airflow uniformity change status, air duct return flow intensity change status and constraint satisfaction status corresponding to each candidate air duct topology reconstruction scheme are read. The topology connection edges that appear repeatedly in each candidate air duct topology reconstruction scheme and have a stable contribution to heat dissipation improvement are taken as priority reconstruction objects, and a set of candidate edges for air duct topology reconstruction is generated. Based on the candidate edge set of airflow topology reconstruction, topology connection edges adjacent to the CPU core hot spot neighborhood, the top of the heat sink base, the heat sink fin inlet, and the local high-temperature fluid retention area are identified. It is determined whether the topology connection edge in the Zhenyu optimized airflow candidate solution set continuously corresponds to the decrease of the CPU core maximum temperature, the decrease of the CPU average junction temperature, or the enhancement of heat transfer in the hot spot area. When it meets the hot spot temperature decrease contribution threshold, the topology connection edge is marked as a hot spot cooling contribution edge, and the hot spot cooling contribution edge marking result is obtained. Based on the hot spot cooling contribution edge labeling results, the hot spot cooling contribution edge is subjected to air duct expansion or flow enhancement processing to obtain the hot spot flow enhancement reconstruction result. Based on the candidate edge set of the air duct topology reconstruction, identify topology connection edges with sufficient air volume but low contribution to the reduction of the maximum CPU core temperature, the reduction of the average CPU junction temperature, or the improvement of airflow uniformity. Then, determine whether the topology connection edge is in a state of excessive ventilation, bypass short circuit, ineffective exhaust, or additional burden on fan power consumption. When it meets the pressure loss redundancy threshold, mark the topology connection edge as a pressure loss redundant edge and obtain the pressure loss redundant edge marking result. Based on the marking results of the pressure loss redundant edge, the pressure loss redundant edge is subjected to air duct contraction or baffle weakening to obtain the pressure loss redundancy suppression and reconstruction results. Based on the candidate edge set of wind duct topology reconstruction, the topological connection edge of the backflow vortex region is identified and generated, and the topological connection edge is marked as the backflow vortex region edge to obtain the backflow vortex region edge marking result. Based on the backflow vortex region edge marking results, the backflow vortex region edge is processed by adding a bypass ventilation duct or correcting the guide angle to obtain the backflow vortex region suppression and reconstruction results. Based on the candidate edge set of duct topology reconstruction, identify topological connection edges with low heat transfer contribution and obtain the low heat transfer contribution edge labeling results; Based on the low heat transfer contribution edge marking results, the low heat transfer contribution edges are merged or deleted. The hot spot flow enhancement reconstruction results, pressure loss redundancy suppression reconstruction results, backflow vortex suppression reconstruction results, and low heat transfer contribution edge processing results are then uniformly collected to generate an adaptive reconstructed duct topology set.

[0013] A computer CPU heat dissipation optimization device includes a data acquisition component (100), a boundary processing component (200), a flow field modeling component (300), a bottleneck diagnosis component (400), an optimization search component (500), an airflow reconstruction component (600), and a control output component (700). The data acquisition component (100) is connected to the boundary processing component (200), the boundary processing component (200) is connected to the flow field modeling component (300), the flow field modeling component (300) is connected to the bottleneck diagnosis component (400), the bottleneck diagnosis component (400) is connected to the optimization search component (500), the optimization search component (500) is connected to the air duct reconstruction component (600), the air duct reconstruction component (600) is connected to the control output component (700), and the control output component (700) is used to output air duct adjustment parameters, fan control parameters, and flow guidance control parameters.

[0014] The beneficial effects of this invention are: This invention transforms traditional heat dissipation optimization methods based on geometric models or single simulation results into a structured expression based on topological nodes and topological connection edges by constructing a fluid-structure-thermal coupling topology graph model and a collaborative optimization mechanism for bottleneck diagnosis driving variable reconstruction. This allows the coupling relationships between the CPU package, thermal interface layer, heat sink base, heat sink fins, fan, and airflow components to be uniformly modeled and computably expressed. By simultaneously incorporating airflow relationships, thermal contact relationships, and flow guidance constraints into the topological connection edge system, and extracting multi-source bottleneck features such as hot spot temperature gradient, fin penetration velocity, and airflow return intensity, the optimization process is upgraded from structural parameter adjustment to directional repair of flow field bottlenecks, enhancing the analytical capability for multi-path coupled flow inside complex chassis.

[0015] This invention introduces the swarm optimization algorithm and combines it with differentiated topology search strategies for three types of individuals: leader, follower, and disturbance. This achieves a synergistic unity of global search capability and local fine-tuning capability during duct topology optimization. The leader swarm optimizer rapidly converges to low thermal resistance and low pressure loss regions based on multi-objective fitness. The follower swarm optimizer enhances directional airflow to the CPU core region based on the hotspot neighborhood guidance weight. The disturbance swarm optimizer performs topology jump search targeting recirculation vortex regions and airflow stagnation regions. This avoids the problem of traditional swarm intelligence algorithms easily getting trapped in local optima in complex flow field optimization. Furthermore, the invention introduces the physical meaning of the flow field into the search process, ensuring that topology changes directly correspond to hotspot cooling, recirculation suppression, and heat transfer enhancement objectives, thus improving search efficiency and the engineering feasibility of the solution.

[0016] This invention constructs an adaptive airflow topology reconstruction mechanism based on contribution thresholds, enabling dynamic screening and hierarchical adjustment of airflow structures. It can perform differentiated processing for different flow field problem types. For hot spot areas, airflow expansion and flow enhancement strategies are used to improve local heat transfer capacity. For pressure loss redundant areas, airflow contraction and baffle weakening strategies are used to reduce system resistance. For backflow vortex areas, bypass airflow channel addition and flow angle correction strategies are used to improve exhaust path. For low heat transfer contribution areas, topology merging or deletion is performed to reduce structural complexity. The hierarchical reconstruction method based on flow field contribution ensures that the optimization process not only focuses on temperature indicators but also takes into account fan power consumption, airflow organization, and structural rationality, achieving a coordinated improvement in the overall performance of the CPU cooling system under multiple constraints. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a computer CPU heat dissipation optimization method proposed in this invention; Figure 2 This is a structural diagram of a computer CPU heat dissipation optimization device proposed in this invention. Detailed Implementation

[0018] Example 1: Reference Figures 1-2 A method for co-optimizing CPU heat dissipation flow field based on the rare carp optimization algorithm and adaptive airflow topology reconstruction includes: Collect structural boundary data, operational heat source data, and environmental status data of the CPU cooling system, perform preprocessing, and generate a set of multi-source boundary conditions for CPU cooling. Based on the set of multi-source boundary conditions for CPU heat dissipation, a fluid-structure-thermal coupling computational domain for CPU heat dissipation is constructed to obtain the initial airflow topology model. Based on the initial duct topology flow field model, the baseline flow field is solved to obtain the temperature field, velocity field and vorticity field corresponding to the CPU cooling system. The bottleneck features of the flow field are extracted according to the CPU core hot spot temperature gradient, the heat sink fin penetration wind speed and the duct return flow intensity, and a CPU cooling flow field bottleneck diagnosis set is generated. Based on the CPU heat dissipation flow field bottleneck diagnosis set, a reconfigurable parameter vector of air duct topology is constructed, and combined with air duct size constraints, fan power consumption constraints and CPU safe temperature constraints, a set of CPU heat dissipation collaborative optimization variables is formed. The population of the rare trevally optimization algorithm is initialized based on the set of CPU heat dissipation collaborative optimization variables. Each rare trevally individual is encoded as a candidate airflow topology reconstruction scheme. A multi-objective fitness function is constructed based on the highest CPU core temperature, average CPU junction temperature, heat dissipation thermal resistance and airflow uniformity to obtain the set of rare trevally candidate topology populations. Based on the candidate topology population set of the rare trevally, the rare trevally optimization iterative search is performed. According to the multi-objective fitness function value of the candidate airway topology reconstruction scheme, the rare trevally individuals are divided into leading rare trevally individuals, following rare trevally individuals, and disturbing rare trevally individuals. The leading rare trevally individuals are made to migrate towards the airway topology region with low thermal resistance and low pressure loss. The following rare trevally individuals are made to gather towards the high heat transfer region corresponding to the CPU core hot spot according to the hot spot neighborhood guidance weight. The disturbing rare trevally individuals are made to perform topology jump search according to the local eddy residence time and airway backflow intensity, so as to obtain the candidate solution set of the rare trevally optimized airway. Based on the candidate solution set of the optimized air duct, an adaptive air duct topology reconstruction is performed. For topology connection edges that meet the hot spot temperature reduction contribution threshold, air duct expansion or flow enhancement processing is performed. For topology connection edges that meet the pressure loss redundancy threshold, air duct contraction or baffle weakening processing is performed. For topology connection edges that generate backflow vortex regions, bypass air duct addition or flow angle correction processing is performed. For topology connection edges with low heat transfer contribution, merging or deletion processing is performed to generate an adaptively reconstructed air duct topology set.

[0019] In this embodiment, preprocessing includes coordinate unification, time alignment, unit normalization, and outlier data removal.

[0020] In this embodiment, a CPU heat dissipation fluid-structure-thermal coupling computational domain is constructed based on a set of multi-source boundary conditions for CPU heat dissipation, including: Based on the set of multi-source boundary conditions for CPU heat dissipation, spatial coordinate registration and boundary contour closure checks are performed on each structural object to generate CPU heat dissipation structural boundary registration results. In Example 1, the CPU package size, CPU heat generation area location, thermal interface layer thickness, heat dissipation fin arrangement direction, chassis air inlet location, chassis air outlet location, and reconfigurable air duct component installation range are read based on the CPU heat dissipation multi-source boundary condition set.

[0021] Based on the boundary registration results of the CPU heat dissipation structure, the CPU heat dissipation system is divided into regions according to the solid heat conduction region, the fluid flow region and the adjustable flow conduction region, and the CPU heat dissipation partition calculation domain is obtained. In Example 1, the CPU package, thermal interface layer, heat sink base and heat sink fins are classified as solid thermal conduction area, the fan outlet area, chassis air inlet, chassis air outlet and heat sink fin gap are classified as fluid flow area, and the air guide shroud, openable baffle, bypass ventilation channel, air guide plate and movable partition are classified as adjustable airflow area.

[0022] Based on the CPU heat dissipation partition computational domain, a set of CPU heat dissipation flow solid-thermal coupling boundary attributes is formed; In Example 1, the thermally conductive material properties, contact thermal resistance properties, and heat source loading position are set for the solid thermally conductive region; the air inlet boundary, air outlet boundary, fan drive boundary, and wall adhesion boundary are set for the fluid flow region; and the variable channel boundary, variable flow angle boundary, and baffle opening and closing boundary are set for the adjustable flow guiding region.

[0023] Based on the set of boundary properties of CPU heat dissipation flow-structure-thermal coupling, the CPU package, thermal interface layer, heat sink base, heat sink fins, fan, chassis air inlet, chassis air outlet and reconfigurable air duct components are abstracted as flow field topology nodes, and each flow field topology node is kept in a one-to-one correspondence with its corresponding physical structure region, material properties, boundary properties and spatial position, so as to obtain the set of CPU heat dissipation flow field topology nodes. Based on the CPU heat dissipation flow field topology node set, airflow connection edges are established between the fan, chassis air inlet, heat sink fin gap, reconfigurable air duct assembly and chassis air outlet according to the adjacent airflow relationship. Thermal contact connection edges are established between the CPU package, thermal interface layer, heat sink base and heat sink fin according to the thermal contact relationship. Thermal constraint connection edges are established between the reconfigurable air duct assembly and adjacent fluid flow areas according to the flow constraint relationship, thus obtaining the CPU heat dissipation topology connection edge set. Based on the CPU heat dissipation topology connection edge set, each topology connection edge is labeled with its corresponding flow direction, passability, heat transfer path, airflow blockage status, baffle constraint status, and airflow influence range. Consistency screening is performed on conflicting topology connection edges to ensure that airflow connection edges, thermal contact connection edges, and airflow constraint connection edges in the same spatial area do not have overlapping physical meanings, thus obtaining the CPU heat dissipation topology constraint relationship set. Based on the set of CPU heat dissipation topology constraints, the CPU heat dissipation partition computational domain, the CPU heat dissipation flow field topology node set, and the CPU heat dissipation topology connection edge set are associated and encapsulated. The association and encapsulation results are used as input for subsequent baseline flow field solution to obtain the initial air duct topology flow field model.

[0024] In Example 1, the associated encapsulation enables each flow field topology node to invoke its corresponding boundary conditions, and enables each topology connection edge to characterize its corresponding airflow relationship, thermal contact relationship, or flow constraint relationship.

[0025] In this embodiment, baseline flow field solution is performed based on the initial duct topology flow field model, including: The initial attitude is read based on the initial duct topology flow field model, and the initial attitude is limited to the duct structure state before the search of the Zhenyu optimization algorithm is performed, and the CPU heat dissipation baseline solution input set is generated. In Example 1, the initial posture includes the CPU package heat source distribution, the contact state of the thermal interface layer, the heat conduction path of the heat sink base, the heat sink fin arrangement channel, the fan airflow direction, the chassis air inlet boundary, the chassis air outlet boundary, and the reconfigurable airflow duct assembly.

[0026] Based on the CPU heat dissipation baseline solution input set, thermal conductivity solution conditions are configured for the solid regions where the CPU package, thermal interface layer, heat sink base and heat sink fins are located, and flow solution conditions are configured for the fluid regions where the fan, chassis inlet, chassis outlet, heat sink fin gap and reconfigurable air duct components are located. Heat exchange relationship is set at the interface between the solid region and the fluid region to obtain the CPU heat dissipation baseline fluid-solid-thermal coupling solution conditions. Based on the baseline fluid-structure-thermal coupling solution conditions for CPU heat dissipation, the baseline flow field solution is performed to obtain the temperature field results of the CPU heat dissipation system under the initial air duct topology flow field model. In Example 1, the baseline flow field solution allows the heat generated by the CPU package to be transferred to the heat sink base and heat sink fins through the thermal interface layer, and the airflow driven by the fan to form a baseline heat dissipation flow path through the chassis air inlet, the gap between the heat sink fins, the reconfigurable air duct assembly and the chassis air outlet.

[0027] Based on the temperature field results, the airflow state of the fan outlet neighborhood, heat sink inlet area, heat sink middle channel, heat sink tail area, chassis air outlet neighborhood, and reconfigurable air duct component neighborhood is further extracted to determine the airflow direction, airflow speed, airflow stagnation area, and airflow bypass area in each area, thus obtaining the velocity field results corresponding to the CPU cooling system. Based on the velocity field results, the vortex field results are obtained by identifying the vortex field in the fan outlet impact area, the heat sink fin gap shear area, the airflow guide component backwind area, the chassis wall near-wall area, and the air outlet confluence area. Based on the vortex field results, the local vortex location, vortex duration range, vortex retention tendency, and vortex interference degree of the main flow channel are determined, and the vortex field results corresponding to the CPU cooling system are obtained. Based on the temperature field results, the highest temperature location, hot spot distribution range, temperature change trend of hot spot diffusion to the heat dissipation interface layer, and temperature change trend of hot spot transmission to the heat sink base in the CPU core area are extracted. It is then determined whether the CPU core hot spot temperature gradient is concentrated in a local area, and the area where the temperature gradient is concentrated is marked as the hot spot heat conduction bottleneck area, thus obtaining the CPU core hot spot temperature gradient characteristics. Based on the velocity field results, the airflow penetration state from the inlet to the outlet of the heat sink fins is extracted to determine whether the fan airflow effectively enters the gap between the heat sink fins, and to determine whether there are airflow short circuits, local low speeds, and insufficient heat exchange at the tail between the heat sink fins. The fin channels with insufficient penetration airflow and corresponding high temperatures are marked as fin heat exchange bottleneck areas, and the characteristics of heat sink fin penetration airflow are obtained. Based on the vorticity field and velocity field results, the return flow location, return flow direction, return flow range, and degree of interference of the return flow on the main air intake channel are extracted. It is determined whether there is reverse airflow circulation on the leeward side of the reconfigurable duct component, the corner of the chassis, and the leading edge of the air outlet. The area with high return flow intensity and obstruction of hot air exhaust is marked as the duct return flow bottleneck area, and the duct return flow intensity characteristics are obtained. Based on the CPU core hot spot temperature gradient characteristics, heat sink fin penetration wind speed characteristics, and airflow return intensity characteristics, the flow field bottleneck characteristics are merged according to the heat source side bottleneck, heat exchange side bottleneck, and exhaust side bottleneck to generate a CPU heat dissipation flow field bottleneck diagnosis set.

[0028] In this embodiment, a reconfigurable parameter vector for the airflow topology is constructed based on the CPU heat dissipation flow field bottleneck diagnosis set, including: Based on the CPU heat dissipation flow field bottleneck diagnosis set, the spatial location, bottleneck type, influence range and corresponding topological connection edge of the hot spot heat conduction bottleneck area, fin heat exchange bottleneck area and air duct return flow bottleneck area are read, and each type of bottleneck is associated with the flow field topology node and topological connection edge in the initial air duct topology flow field model to generate flow field bottleneck topology location results. Based on the topological location results of the flow field bottleneck, the set of regions affected by airflow duct reconstruction is obtained; In Example 1, the heat sink base area, the heat sink fin inlet area, and the reconfigurable airflow component action area adjacent to the CPU core heat conduction bottleneck area are determined. The heat sink fin gap, local ventilation cross section, and airflow component attitude adjustment range are determined for the fin heat exchange bottleneck area. The bypass channel setting range, baffle weakening range, and airflow direction correction range are determined for the airflow return bottleneck area.

[0029] Based on the set of the reconfigurable regions of the airway, a reconfigurable parameter vector of the airway topology is constructed, and the reconfigurable parameter results of the airway topology are obtained. In Example 1, the following parameters are incorporated into the same parameter expression system: the guide angle parameter for changing the air intake direction, the diffusion angle parameter for changing the air outlet diffusion range, the baffle opening and closing parameter for changing the local airflow, the bypass duct width parameter for changing the bypass flow path, the fin gap correction parameter for changing the heat sink fin flow capacity, the fan speed control parameter for changing the airflow supply, and the hot spot neighborhood guide weight parameter for guiding the airflow toward the CPU core hot spot neighborhood.

[0030] Based on the reconfigurable parameters of the duct topology, duct size constraints are set to obtain the duct size constraint results. In Example 1, the airflow dimension constraint limits the airflow angle, diffusion angle, bypass airflow width, baffle opening and closing state, and fin gap correction amount to the installation space of the CPU cooling system, the internal clearance of the chassis, the external dimensions of the heat sink, the fan installation space, and the mechanical movement range of the reconfigurable airflow components. This avoids structural interference between the airflow topology reconfiguration scheme and the CPU package, heat sink base, heat sink fins, fan, or chassis wall.

[0031] Fan power consumption constraints are set based on the reconfigurable parameters of the air duct topology, and the fan power consumption constraint results are obtained. In Example 1, the fan power consumption constraint ensures that the fan speed control parameters, the air duct contraction state, the air duct expansion state, and the local baffle state together meet the sustainable operating power consumption range of the fan, and ensures that the increase in fan load caused by air duct blockage, air duct bend, or local backflow does not exceed the preset power consumption allowable range.

[0032] Based on the reconfigurable parameters of the airflow topology, CPU safety temperature constraints are set to ensure that the maximum CPU core temperature, average CPU junction temperature, and hot spot duration are all within the safe operating range allowed by the CPU cooling system. Furthermore, any candidate airflow topology reconfiguration scheme must not cause the CPU core hot spot temperature to exceed the upper limit of the safety temperature at the expense of reducing local pressure loss during the optimization process, thus obtaining the CPU safety temperature constraint results. Based on the constraints of airflow size, fan power consumption, and CPU safe temperature, feasibility marking is performed on each parameter in the airflow topology reconfigurable parameter results. Parameters that satisfy all constraints are marked as searchable variables, parameters that violate any constraint are marked as prohibited variables, and parameters that are close to the constraint boundary but are still adjustable are marked as boundary variables, thus generating a set of CPU heat dissipation co-optimization variables.

[0033] In this embodiment, the population of the rare trevally optimization algorithm is initialized based on the CPU heat dissipation collaborative optimization variable set, including: Based on the CPU heat dissipation collaborative optimization variable set, the reconfigurable parameters of the air duct topology marked as searchable variables and boundary variables are read, and the reconfigurable parameters of the air duct topology marked as prohibited variables are excluded. The flow guide angle parameter, diffusion angle parameter, baffle opening and closing parameter, bypass air duct width parameter, fin gap correction parameter, fan speed control parameter, and hot spot neighborhood flow guide weight parameter are sorted according to the corresponding flow field topology nodes and topology connection edges to generate the individual coding variable sequence of the rare trevally. Based on the individual coding variable sequence of the rare trevally, an individual expression method for candidate air duct topology reconstruction schemes is established, so that each rare trevally individual contains a complete set of guide angle adjustment state, diffusion angle adjustment state, baffle opening and closing state, bypass air duct width state, fin gap correction state, fan speed control state, and hot spot neighborhood guide weight state. Each rare trevally individual can uniquely correspond to a candidate air duct topology reconstruction scheme that can be verified by fluid-structure-thermal coupling, and the coding result of the rare trevally individual candidate scheme is obtained. Based on the coding results of the candidate schemes for the rare trevally individuals, population initialization is performed. Within the allowable range of air duct size constraints, fan power consumption constraints, and CPU safe temperature constraints, multiple rare trevally individuals are generated. These individuals are made to have differentiated distributions in terms of guide angle, bypass duct width, baffle opening and closing status, fan speed control amount, and hot spot neighborhood guide weight, so as to cover three candidate reconstruction directions: hot spot guide enhancement, fin penetration improvement, and backflow suppression, thus obtaining the initial rare trevally topological population results. Based on the initial rare trevally topological population results, the candidate wind tunnel topology reconstruction scheme corresponding to each rare trevally individual is mapped back to the initial wind tunnel topology flow field model. The corresponding topology connecting edges are marked with state tags for expansion, contraction, opening and closing, bypass, guide angle correction or fin gap correction, and the flow field topology nodes and topology connecting edges that have not been changed by the rare trevally individual are kept in their original state, thus obtaining a set of candidate wind tunnel topology flow field models. The candidate air duct topology reconstruction schemes are evaluated based on the candidate air duct topology flow field model set. The maximum CPU core temperature, average CPU junction temperature, heat dissipation thermal resistance and airflow uniformity under each candidate air duct topology reconstruction scheme are obtained, and a multi-objective evaluation index set of candidate schemes is obtained. In Example 1, the highest CPU core temperature is used to characterize the hot spot safety risk, the average CPU junction temperature is used to characterize the overall thermal load, the thermal resistance is used to characterize the degree of heat dissipation obstruction from the CPU package to the flowing air, and the airflow uniformity is used to characterize the degree of airflow balance between the heat sink fin gaps and the hot spot neighborhood.

[0034] A multi-objective fitness function was constructed based on a set of multi-objective evaluation indicators for candidate schemes to obtain the fitness evaluation results of rare trevally individuals. In Example 1, the multi-objective fitness function gives higher optimization priority to candidate airflow topology reconstruction schemes that have lower CPU core maximum temperature, lower CPU average junction temperature, smaller heat dissipation thermal resistance, and better airflow uniformity. It also weakens the fitness of candidate airflow topology reconstruction schemes that violate airflow size constraints, fan power consumption constraints, or CPU safe temperature constraints.

[0035] Based on the fitness evaluation results of individual rare trevally, the rare trevally individuals in the initial rare trevally topological population results are sorted, labeled and archived. Rare trevally individuals that meet the constraints and have better fitness evaluation results are retained. At the same time, the candidate windway topology reconstruction scheme, variable value status, topology connection edge change status and multi-objective evaluation index corresponding to each rare trevally individual are recorded to obtain the rare trevally candidate topological population set.

[0036] In this embodiment, an iterative search for rare trevally optimization is performed based on the candidate topological population set of rare trevally, including: Based on the candidate topology population set of rare trevally, the candidate air duct topology reconstruction scheme, multi-objective fitness function value, maximum CPU core temperature, average CPU junction temperature, thermal resistance, airflow uniformity, air duct topology connection edge status and constraint satisfaction status corresponding to each rare trevally individual are read, and rare trevally individuals that do not meet the air duct size constraint, fan power consumption constraint or CPU safe temperature constraint are removed from the current iteration valid search objects to generate the current valid rare trevally search population; Based on the current effective search population of rare trevally, the rare trevally individuals are sorted from best to worst according to their multi-objective fitness function values. Rare trevally individuals with fitness function values ​​in the advantageous range and simultaneously have lower maximum CPU core temperature, lower average CPU junction temperature, and lower thermal resistance are classified as leading rare trevally individuals. Rare trevally individuals with fitness function values ​​in the middle range and still have room for improvement are classified as following rare trevally individuals. Rare trevally individuals with fitness function values ​​in the disadvantageous range or with high corresponding airflow intensity are classified as disturbing rare trevally individuals. The results of the rare trevally individual role classification are obtained. Based on the role classification of the individual rare tuna, the state of the topology connection edge that contributes to the reduction of heat dissipation thermal resistance and the reduction of airflow pressure loss in its candidate airflow topology reconstruction scheme is read. The guide angle, diffusion angle, bypass airflow width and baffle opening and closing state are adjusted around the airflow topology regions corresponding to the heat dissipation fin inlet, heat dissipation fin tail, chassis air outlet neighborhood and low resistance exhaust path, so that the leading rare tuna individual migrates towards the airflow topology region with low thermal resistance and low pressure loss, and the migration result of the leading rare tuna is obtained. Based on the following trevally individuals in the individual role classification results, the heat spot neighborhood flow guidance weight and the high heat transfer area location corresponding to the CPU core heat spot are read. The candidate airflow topology reconstruction scheme corresponding to the following trevally individuals is moved closer to the flow guidance state related to the heat spot temperature drop in the leading trevally individuals. The flow guidance parameters close to the CPU core heat spot, above the heat sink base, the heat sink fin inlet, and the heat spot neighborhood channel are adjusted first, so that the following trevally individuals gather in the high heat transfer area corresponding to the CPU core heat spot according to the heat spot neighborhood flow guidance weight, and the following trevally gathering result is obtained. Based on the disturbed trevally individuals in the individual role classification results, the local vortex residence time and airflow recirculation intensity in the corresponding candidate airflow topology reconstruction scheme are read. The topology connection edges that cause hot air retention, reverse circulation of the main channel, or poor exhaust at the tail of the heat sink fins are identified. The state jump is performed on the topology connection edges, adjusting the bypass airflow duct that was originally in the closed state to the open candidate state, adjusting the baffle that originally blocked the exhaust to the weakening candidate state, and adjusting the guide angle that originally generated the recirculation to the correction candidate state, thus obtaining the disturbed trevally topology jump results. Based on the migration results of the leading trevally, the aggregation results of the following trevally, and the topology jump results of the disturbed trevally, a constraint consistency check is performed on the guide angle state, baffle opening and closing state, bypass ventilation duct width state, and fan speed control state that may conflict within the same trevally individual. This ensures that the optimized trevally individual still satisfies the duct size constraint, fan power consumption constraint, and CPU safe temperature constraint. Unrealizable or contradictory topology change states are deleted, resulting in the trevally iteratively updated population results. Based on the iterative update of the rare trevally population results, the multi-objective fitness function values ​​of each rare trevally individual are re-evaluated, and candidate airflow topology reconstruction schemes that can reduce the maximum CPU core temperature, reduce the average CPU junction temperature, reduce heat dissipation thermal resistance, or improve airflow uniformity are retained and archived as optimization candidate solutions for the current iteration, thus obtaining the rare trevally optimized airflow candidate solution set.

[0037] In this embodiment, adaptive airway topology reconstruction is performed based on the candidate solution set of the optimized airway, including: Based on the Zhenyu optimization of the candidate solution set for the air duct, the topology connection edge change status, CPU core maximum temperature change status, CPU average junction temperature change status, heat dissipation thermal resistance change status, airflow uniformity change status, air duct return flow intensity change status and constraint satisfaction status corresponding to each candidate air duct topology reconstruction scheme are read. The topology connection edges that appear repeatedly in each candidate air duct topology reconstruction scheme and have a stable contribution to heat dissipation improvement are taken as priority reconstruction objects, and a set of candidate edges for air duct topology reconstruction is generated. Based on the candidate edge set of airflow topology reconstruction, topology connection edges adjacent to the CPU core hot spot neighborhood, the top of the heat sink base, the heat sink fin inlet, and the local high-temperature fluid retention area are identified. It is determined whether the topology connection edge in the Zhenyu optimized airflow candidate solution set continuously corresponds to the decrease of the CPU core maximum temperature, the decrease of the CPU average junction temperature, or the enhancement of heat transfer in the hot spot area. When it meets the hot spot temperature decrease contribution threshold, the topology connection edge is marked as a hot spot cooling contribution edge, and the hot spot cooling contribution edge marking result is obtained. Based on the hot spot cooling contribution edge marking results, airflow expansion or flow enhancement processing is performed on the hot spot cooling contribution edge. The airflow expansion processing is used to increase the effective flow space into the CPU core hot spot neighborhood or the heat sink fin inlet area. The flow enhancement processing is used to adjust the attitude of the reconfigurable airflow component and guide the airflow to preferentially pass through the high heat transfer area corresponding to the CPU core hot spot, so as to obtain the hot spot flow enhancement reconstruction result. Based on the candidate edge set of the air duct topology reconstruction, identify topology connection edges with sufficient air volume but low contribution to the reduction of the maximum CPU core temperature, the reduction of the average CPU junction temperature, or the improvement of airflow uniformity. Then, determine whether the topology connection edge is in a state of excessive ventilation, bypass short circuit, ineffective exhaust, or additional burden on fan power consumption. When it meets the pressure loss redundancy threshold, mark the topology connection edge as a pressure loss redundant edge and obtain the pressure loss redundant edge marking result. Based on the marking results of the pressure loss redundant edge, the pressure loss redundant edge is subjected to airflow contraction or baffle weakening. The airflow contraction is used to reduce the ineffective flow cross section and suppress airflow from bypassing the heat sink fins or hot spot neighborhood. The baffle weakening is used to reduce the obstruction effect of local baffles on the mainstream exhaust path and keep the fan power consumption within the allowable range, thus obtaining the pressure loss redundancy suppression reconstruction result. Based on the candidate edge set for topology reconstruction of the air duct, the topology connection edges that generate the backflow vortex region are identified. The topology connection edges that generate the backflow vortex region include those located on the leeward side of the reconfigurable air duct component, the tail of the heat sink fin, the corner of the chassis, and the leading edge of the chassis air outlet, and those that exhibit reverse airflow circulation, hot air retention, or continuous vortex enhancement. These topology connection edges are then marked as backflow vortex region edges, resulting in the backflow vortex region edge marking results. Based on the edge marking results of the backflow vortex region, the edge of the backflow vortex region is subjected to bypass ventilation duct addition or guide angle correction processing. The bypass ventilation duct addition processing is used to establish an exhaust path for the trapped hot air to bypass the backflow region, and the guide angle correction processing is used to change the local airflow injection direction or ejection direction and weaken the airflow reverse circulation, thus obtaining the backflow vortex region suppression and reconstruction results. Based on the candidate edge set of the air duct topology reconstruction, topological connection edges with low heat transfer contribution are identified. Topological connection edges with low heat transfer contribution include airflow connection edges or flow-guiding constraint connection edges that have failed to reduce the maximum CPU core temperature, the average CPU junction temperature, or the heat dissipation thermal resistance and have a weak effect on improving airflow uniformity. These are marked as low heat transfer contribution edges, and the low heat transfer contribution edge marking results are obtained. Based on the low heat transfer contribution edge marking results, the low heat transfer contribution edges are merged or deleted. The merging process is used to merge adjacent and overlapping duct connection relationships into the same valid duct connection relationship. The deletion process is used to remove invalid connection relationships that do not contribute enough to heat transfer and may increase structural complexity or fan load. The hot spot flow enhancement reconstruction results, pressure loss redundancy suppression reconstruction results, backflow vortex suppression reconstruction results and low heat transfer contribution edge processing results are uniformly collected to generate an adaptive reconstructed duct topology set.

[0038] A CPU heat dissipation flow field collaborative optimization device based on the rare carp optimization algorithm and adaptive air duct topology reconstruction includes a data acquisition component, a boundary processing component, a flow field modeling component, a bottleneck diagnosis component, an optimization search component, an air duct reconstruction component, and a control output component; The data acquisition unit is connected to the boundary processing unit to collect and output structural boundary data, operational heat source data, and environmental status data of the CPU cooling system. The boundary processing unit is connected to the flow field modeling unit to generate a set of multi-source boundary conditions for CPU cooling. The flow field modeling unit is connected to the bottleneck diagnosis unit to construct an initial duct topology flow field model and output the temperature field, velocity field, and vorticity field. The bottleneck diagnosis unit is connected to the optimization search unit to generate a set of bottleneck diagnoses for the CPU cooling flow field. The optimization search unit is connected to the duct reconstruction unit to perform iterative optimization search and output a set of candidate solutions for optimized ducts. The duct reconstruction unit is connected to the control output unit to generate an adaptive reconstructed duct topology set. The control output unit outputs duct adjustment parameters, fan control parameters, and airflow control parameters.

[0039] Example 2: During a workstation thermal overhaul test, the implementer discovered that a high-load computing host exhibited persistent hot spots above the CPU package when continuously running CPU-intensive compilation, rendering, and matrix operation tasks. Monitoring software showed that the CPU core's maximum temperature repeatedly approached the throttling protection threshold. Although the fan speed had been increased to a high level, hot air remained trapped at the tail of the heatsink fins. Further observation revealed sufficient airflow at the front of the chassis and no significant deficiency in fan exhaust speed. However, the airflow distribution after entering the heatsink fins was uneven. The airflow speed in the fin channel near the CPU core hot spot was low, while the airflow on the other side bypassed the hot spot area and was directly exhausted. Traditional methods, such as increasing the fan speed from 1450 RPM to 1850 RPM, only reduced the CPU core's maximum temperature from 96.4°C to 94.8°C, increased fan power consumption from 4.8 watts to 6.3 watts, and increased noise from 42.6 dB to 48.1 dB. This indicates that simply increasing airflow cannot effectively solve the problems of hot spot concentration and airflow recirculation.

[0040] The implementers conducted multi-source boundary data acquisition on the CPU cooling system. The data collected included CPU package dimensions, thermal interface layer thickness, heatsink base dimensions, fin spacing, fan position, chassis inlet dimensions, chassis exhaust dimensions, and the initial position of the adjustable baffle. The acquired data showed a CPU package area of ​​38 mm x 38 mm, an average thermal interface layer thickness of 0.12 mm, a heatsink base thickness of 6.5 mm, 48 heatsink fins, an average fin spacing of 1.85 mm, a fan diameter of 120 mm, an effective chassis inlet area of ​​9140 square millimeters, and an effective chassis exhaust area of ​​7820 square millimeters. During operation, the total thermal power of the CPU package remained stable between 214 watts and 226 watts under full load, with the maximum local heat flux density in the core area reaching 67.5 watts per square centimeter. In the environmental data, the intake air temperature ranged from 27.6°C to 28.4°C, the relative humidity ranged from 46% to 51%, and the airflow velocity outside the main unit was less than 0.15 meters per second. The implementers unified the data collected by different sensors to the same coordinate system and removed data points where the temperature probe's instantaneous change exceeded 3°C, with a removal rate of 1.7%.

[0041] The implementers constructed a CPU heat dissipation flow-structure-thermal coupling computational domain based on the aforementioned boundary conditions. The CPU package, thermal interface layer, heatsink base, heatsink fins, fan, chassis inlet, chassis outlet, and reconfigurable airflow components were mapped as flow field topology nodes. The model established eight main topology node types, which were further subdivided into 126 structural nodes and 312 topology connection edges, including 184 airflow connections, 62 thermal contact connections, and 66 flow-guiding constraint connections. The implementers found that in traditional geometric models, the area between the chassis inlet and the fan was simply considered a single inlet channel, while in the actual flow field, this region has an upper high-velocity channel and a lower low-velocity channel. Therefore, these were marked as two airflow branches in the topology model. The heatsink base and thermal interface layer were configured as thermal contact connections, the heatsink fin gaps and fan outlets were configured as airflow connections, and the adjustable baffle and the heatsink fin inlet edge were configured as flow-guiding constraint connections. After a topology consistency check, the implementers found that seven connecting edges spatially overlapped with the chassis wall, and three other airflow constraint edges were oriented in the opposite direction to the main air intake. After correction, an initial airflow topology model was obtained. This model can clearly represent the path of heat from the CPU package into the heatsink base, and also the path of airflow from the chassis intake into the heatsink fins and out.

[0042] During the baseline flow field solution process, the implementer set the total CPU heat output to 220 watts, the initial fan speed to 1600 RPM, and kept the adjustable baffle at its default opening angle. The results showed that the highest CPU core temperature was 96.4 degrees Celsius, the average CPU junction temperature was 88.9 degrees Celsius, the average surface temperature of the thermal interface layer was 82.3 degrees Celsius, the center temperature of the heatsink base was 78.6 degrees Celsius, the average air temperature at the inlet of the heatsink fins was 31.2 degrees Celsius, and the average air temperature at the outlet of the heatsink fins was 43.7 degrees Celsius. The velocity field results showed that the average airflow velocity at the fan outlet was 2.42 m / s, but the average penetration velocity after entering the effective channels of the heatsink fins was only 1.18 m / s. Specifically, the average airflow velocity in the 12 fin channels closest to the CPU hotspot was only 0.74 m / s, while the average airflow velocity in the 16 fin channels furthest from the hotspot reached 1.46 m / s, indicating a significantly uneven airflow distribution. Eddy field results show that a localized recirculation vortex region is formed between the heatsink tail and the chassis exhaust vent, accounting for 17.9% of the effective flow area volume. The average residence time of hot air in this region is 4.6 seconds. Hot spot analysis results show that the temperature in the upper right region of the CPU core is 13.8 degrees Celsius higher than that of the package edge region, and the hot spot temperature gradient is concentrated and lasts for more than 86% of the entire test cycle.

[0043] When constructing optimization variables, the implementers did not directly add new heat dissipation components. Instead, they extracted reconfigurable airflow parameters within the limits allowed by the existing structure. The allowable range for the adjustable airflow angle was limited to 10 to 38 degrees, the allowable range for the side ventilation duct width was 0 to 22 mm, the range for reducing the height of the local baffle was 0 to 9 mm, the range for effective airflow correction on the fin inlet side was 0 to 1.2 mm, and the allowable range for fan speed was 1350 to 1850 rpm. Due to the limited internal space of the chassis, when the width of the side ventilation duct exceeds 22 mm, it will interfere with the reserved space on the backplate of the graphics card; therefore, this range was marked as a prohibited variable. When the airflow angle exceeds 38 degrees, although it can enhance the airflow to the hot spot side, it will cause the low-speed zone on the other side of the heatsink to expand, so it was also excluded. After constraint screening, the implementers retained 53 searchable variables and 14 boundary variables, and eliminated 19 infeasible variables.

[0044] During the initialization phase of the Zhenyu optimization algorithm population, the implementers generated 40 candidate airflow topology reconstruction samples. Each sample corresponds to a candidate airflow topology scheme. For example, in one candidate sample, the guide angle is 24 degrees, the bypass duct width is 8 mm, the baffle weakening height is 3 mm, the fan speed is 1620 rpm, and the hotspot neighborhood guide weight is medium; in another candidate sample, the guide angle is 32 degrees, the bypass duct width is 18 mm, the baffle weakening height is 6 mm, the fan speed is 1580 rpm, and the hotspot neighborhood guide weight is high. The implementers calculated the CPU core maximum temperature, CPU average junction temperature, thermal resistance, and airflow uniformity corresponding to each candidate sample. Among the initial 40 samples, the worst-performing sample caused the CPU core temperature to reach a maximum of 97.1 degrees Celsius because the bypass ventilation channel was too wide, causing some airflow to bypass the heatsink fins. The better-performing sample reduced the CPU core temperature to a maximum of 90.6 degrees Celsius, but the fan power consumption still reached 6.0 watts. The initial sample with the best overall score achieved a CPU core temperature of 91.2 degrees Celsius, an average junction temperature of 83.5 degrees Celsius, a thermal resistance of 0.382 degrees Celsius per watt, and improved the fin channel airflow uniformity index from 0.61 in the traditional method to 0.76, thus obtaining the candidate topological population set of *Triplophysa pulcherrima*.

[0045] During the iterative search, samples with superior fitness and simultaneously low thermal resistance and low pressure drop were classified as "leader trevally" individuals. The implementers observed that a scheme corresponding to a leader individual, adjusting the guide angle to 30 degrees, the bypass duct width to 15 mm, and maintaining the fan speed at 1570 rpm, increased the inlet air velocity of the hotspot side fins from 0.74 m / s to 1.21 m / s without significantly increasing the fan load. Samples with moderate fitness were classified as "follower trevally" individuals. These samples primarily moved closer to the leader individual based on the hotspot neighborhood guide angle weight. For example, after a follower individual adjusted the guide angle from 22 degrees to 29 degrees, the hotspot temperature on the upper right side of the CPU core decreased from 94.3 degrees Celsius to 89.7 degrees Celsius. Samples with poor fitness or significant backflow were classified as "perturbed individuals." These samples underwent topological changes based on local vortex residence time and airflow backflow intensity. For example, after adding a 16mm bypass channel at the rear of the heatsink, the proportion of the backflow vortex volume in a perturbed individual decreased from 16.8% to 7.1%, and the average residence time of hot air decreased from 4.2 seconds to 2.1 seconds. After multiple iterations, the number of candidate samples remained at 40, but the top 12 superior samples were retained in each round, while approximately 8 long-term non-converging samples were replaced. Midway through the iterations, the group average of the highest CPU core temperature decreased from 94.9 degrees Celsius to 89.8 degrees Celsius, the group average of the average junction temperature decreased from 87.4 degrees Celsius to 82.6 degrees Celsius, and the group average of the fan power consumption decreased from 5.9 watts to 5.5 watts. In the final set of candidate solutions for the optimized airflow channel, the optimal candidate solution has a maximum CPU core temperature of 86.9 degrees Celsius, an average CPU junction temperature of 80.7 degrees Celsius, a thermal resistance of 0.331 degrees Celsius per watt, an airflow uniformity index of 0.84, and a recirculation vortex volume ratio of 4.8%.

[0046] During the adaptive airflow topology reconstruction phase, the implementers performed topology edge contribution analysis on the candidate solutions obtained from the optimization. Five topology connection edges adjacent to the CPU core hotspot neighborhood all showed significant cooling contributions in multiple preferred candidate solutions. One connection edge, after enhanced airflow, reduced the average temperature of the hotspot area by 3.4 degrees Celsius. Another connection edge, after airflow expansion, increased the inlet air velocity of the hotspot side fins by 0.31 m / s. Therefore, these edges were deemed to meet the hotspot temperature reduction contribution threshold, and airflow expansion and enhancement were implemented. A bypass connection edge on the right side of the radiator, although having a high flow velocity, resulted in a hotspot temperature reduction of less than 0.4 degrees Celsius and significantly increased fan pressure loss. Therefore, it was deemed to meet the pressure loss redundancy threshold, and airflow contraction was implemented. Two connection edges at the radiator's rear repeatedly exhibited reverse airflow circulation, corresponding to high backflow intensity. The implementers added a bypass duct to one connection edge and performed airflow angle correction on the other, changing the local airflow from inward entrainment to exhaust towards the outlet. Three other low-heat-transfer-contribution connection edges failed to significantly reduce the CPU core's maximum temperature or improve the average junction temperature among all preferred candidate solutions, and were merged into a simplified airflow connection. After reconstruction, in the final airflow topology, the guide angle is fixed at 31 degrees, the bypass duct width is fixed at 17 mm, the baffle weakening height is 5 mm, the target fan speed is 1560 rpm, and the hotspot neighborhood guide weight remains high.

[0047] To demonstrate the effectiveness of the method of this invention compared to traditional methods, the implementers conducted a comparison of multiple sets of candidate samples under the same load, chassis structure, and ambient temperature range. Traditional methods employ a fixed airflow structure, adjusting only the fan speed and the angle of the manual air guide vanes; the method of this invention utilizes the aforementioned Zhenyu optimization algorithm and adaptive airflow topology reconstruction method for optimization. The highest CPU core temperatures achieved by the traditional method in the 10 sets of samples were 96.4°C, 95.8°C, 97.0°C, 96.1°C, 95.6°C, 96.7°C, 95.9°C, 96.3°C, 96.8°C, and 95.7°C, with a corresponding average of 96.23°C. The highest CPU core temperatures achieved by the method of this invention in the same 10 sample groups were 86.9°C, 87.2°C, 86.5°C, 87.0°C, 86.3°C, 87.4°C, 86.8°C, 86.6°C, 87.1°C, and 86.4°C, respectively, with a corresponding average of 86.82°C, representing an average temperature reduction of 9.41°C. The average CPU junction temperature achieved by the traditional method was 88.6°C, while the average CPU junction temperature achieved by the method of this invention was 80.8°C, a reduction of 7.8°C. The average thermal resistance of the traditional method was 0.419°C per watt, while the average thermal resistance of the method of this invention was 0.333°C per watt, a reduction of approximately 20.5%. The average volume ratio of the recirculation vortex region in the conventional method is 17.2%, while the average volume ratio of the recirculation vortex region in the method of this invention is 4.9%, a reduction of approximately 71.5%. The average fan power consumption in the conventional method is 6.2 watts, while the average fan power consumption in the method of this invention is 5.4 watts, a reduction of approximately 12.9%. The average airflow uniformity in the finned channels in the conventional method is 0.62, while the average airflow uniformity in the finned channels in the method of this invention is 0.84, an improvement of approximately 35.5%.

[0048] The implementers also recorded several specific candidate topology samples to verify the interpretability of the algorithm's search process. Candidate sample A, using a small guide angle and no bypass duct, achieved a maximum CPU temperature of 94.6 degrees Celsius and a recirculation vortex volume ratio of 15.8%, indicating that simply slightly changing the guide direction cannot solve the tail retention problem. Candidate sample B, using a larger bypass duct but with an insufficient guide angle, achieved a maximum CPU temperature of 91.8 degrees Celsius and a fan power consumption of 5.9 watts, indicating that the bypass duct can reduce recirculation, but the hotspot side airflow is still insufficient. Candidate sample C, using a 31-degree guide angle, a 17 mm bypass duct, and a 5 mm baffle to reduce the height, achieved a maximum CPU temperature of 86.9 degrees Celsius, an average junction temperature of 80.7 degrees Celsius, a recirculation vortex volume ratio of 4.8%, and a fan power consumption of 5.4 watts, indicating that hotspot guidance, recirculation suppression, and pressure drop control are all optimized simultaneously. Candidate sample D further enlarged the bypass duct to 21 mm, but the CPU's highest temperature actually rose to 88.6 degrees Celsius. This was because some airflow bypassed the fin channel, resulting in insufficient heat exchange in the middle of the fins. These samples demonstrate that the method of this invention does not simply pursue maximum airflow or maximum channel size, but rather selects a solution with superior overall performance through multi-objective evaluation and topology reconstruction.

[0049] After completing the final optimization, the implementers applied the duct topology parameters obtained by the method of this invention to the actual prototype structure and conducted continuous high-load operation verification again: Under traditional methods, after the device has been running for a period of time, the highest temperature of the CPU core will fluctuate repeatedly between 94 degrees Celsius and 97 degrees Celsius, and there will be short-term frequency reduction records. Under the method of the present invention, the highest temperature of the CPU core is stable between 86 degrees Celsius and 88 degrees Celsius, and there is no frequency reduction record. The temperature fluctuation range of the hot spot area is reduced from 5.6 degrees Celsius in the traditional method to 2.1 degrees Celsius.

[0050] In the traditional method, infrared thermal imaging of the tail of the heat sink fins shows a continuous high-temperature band. In the method of the present invention, the high-temperature band is dispersed into multiple low-intensity temperature rise areas, indicating that the hot air exhaust path is smoother.

[0051] In traditional methods, the fan speed needs to be maintained at more than 1800 revolutions per minute to suppress the temperature rise. In the method of this invention, the fan speed can be maintained at about 1560 revolutions per minute to achieve a lower temperature.

[0052] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for optimizing CPU heat dissipation in computers, characterized in that, include: Collect structural boundary data, operational heat source data, and environmental status data of the CPU cooling system, perform preprocessing, and generate a set of multi-source boundary conditions for CPU cooling. Based on the set of multi-source boundary conditions for CPU heat dissipation, a fluid-structure-thermal coupling computational domain for CPU heat dissipation is constructed to obtain the initial airflow topology model. Based on the initial duct topology flow field model, the baseline flow field is solved to obtain the temperature field, velocity field and vorticity field corresponding to the CPU cooling system. The bottleneck features of the flow field are extracted according to the CPU core hot spot temperature gradient, the heat sink fin penetration wind speed and the duct return flow intensity, and a CPU cooling flow field bottleneck diagnosis set is generated. Based on the CPU heat dissipation flow field bottleneck diagnosis set, a reconfigurable parameter vector of air duct topology is constructed, and combined with air duct size constraints, fan power consumption constraints and CPU safe temperature constraints, a set of CPU heat dissipation collaborative optimization variables is formed. The population of the rare trevally optimization algorithm is initialized based on the set of CPU heat dissipation collaborative optimization variables. Each rare trevally individual is encoded as a candidate airflow topology reconstruction scheme. A multi-objective fitness function is constructed based on the highest CPU core temperature, average CPU junction temperature, heat dissipation thermal resistance and airflow uniformity to obtain the set of rare trevally candidate topology populations. Based on the candidate topology population set of the rare trevally, perform the rare trevally optimization iterative search, so that the leading rare trevally migrates towards the air duct topology region with low thermal resistance and low pressure loss, so that the following rare trevally gathers towards the high heat transfer region corresponding to the CPU core hot spot according to the hot spot neighborhood guidance weight, so that the disturbing rare trevally performs topology jump search according to the local eddy residence time and air duct return intensity, and obtain the candidate solution set of the rare trevally optimized air duct; Based on the candidate solution set of the optimized air duct, an adaptive air duct topology reconstruction is performed to generate an adaptive reconstructed air duct topology set.

2. The method for optimizing computer CPU heat dissipation according to claim 1, characterized in that, The preprocessing includes coordinate unification, time alignment, unit normalization, and outlier data removal.

3. The method for optimizing computer CPU heat dissipation according to claim 1, characterized in that, The construction of the CPU heat dissipation fluid-structure-thermal coupling computational domain based on the set of multi-source boundary conditions for CPU heat dissipation includes: Based on the set of multi-source boundary conditions for CPU heat dissipation, spatial coordinate registration and boundary contour closure checks are performed on each structural object to generate CPU heat dissipation structural boundary registration results. Based on the boundary registration results of the CPU heat dissipation structure, the CPU heat dissipation system is divided into regions according to the solid heat conduction region, the fluid flow region and the adjustable flow conduction region, and the CPU heat dissipation partition calculation domain is obtained. Based on the CPU heat dissipation partition computational domain, a set of CPU heat dissipation flow solid-thermal coupling boundary attributes is formed; Based on the set of boundary properties of CPU heat dissipation flow-structure-thermal coupling, the CPU package, thermal interface layer, heat sink base, heat sink fins, fan, chassis air inlet, chassis air outlet and reconfigurable air duct components are abstracted as flow field topology nodes, and each flow field topology node is kept in a one-to-one correspondence with its corresponding physical structure region, material properties, boundary properties and spatial position, so as to obtain the set of CPU heat dissipation flow field topology nodes. Based on the CPU heat dissipation flow field topology node set, airflow connection edges are established between the fan, chassis air inlet, heat sink fin gap, reconfigurable air duct assembly and chassis air outlet according to the adjacent airflow relationship. Thermal contact connection edges are established between the CPU package, thermal interface layer, heat sink base and heat sink fin according to the thermal contact relationship. Thermal constraint connection edges are established between the reconfigurable air duct assembly and adjacent fluid flow areas according to the flow constraint relationship, thus obtaining the CPU heat dissipation topology connection edge set. Based on the CPU heat dissipation topology connection edge set, each topology connection edge is labeled with its corresponding flow direction, passability, heat transfer path, airflow blockage status, baffle constraint status, and airflow influence range. Consistency screening is performed on conflicting topology connection edges to ensure that airflow connection edges, thermal contact connection edges, and airflow constraint connection edges in the same spatial area do not have overlapping physical meanings, thus obtaining the CPU heat dissipation topology constraint relationship set. Based on the set of CPU heat dissipation topology constraints, the CPU heat dissipation partition computational domain, the CPU heat dissipation flow field topology node set, and the CPU heat dissipation topology connection edge set are associated and encapsulated. The association and encapsulation results are used as input for subsequent baseline flow field solution to obtain the initial air duct topology flow field model.

4. The method for optimizing computer CPU heat dissipation according to claim 1, characterized in that, The baseline flow field solution based on the initial duct topology flow field model includes: The initial attitude is read based on the initial duct topology flow field model, and the initial attitude is limited to the duct structure state before the search of the Zhenyu optimization algorithm is performed, and the CPU heat dissipation baseline solution input set is generated. Based on the CPU heat dissipation baseline solution input set, thermal conductivity solution conditions are configured for the solid regions where the CPU package, thermal interface layer, heat sink base and heat sink fins are located, and flow solution conditions are configured for the fluid regions where the fan, chassis inlet, chassis outlet, heat sink fin gap and reconfigurable air duct components are located. Heat exchange relationship is set at the interface between the solid region and the fluid region to obtain the CPU heat dissipation baseline fluid-solid-thermal coupling solution conditions. Based on the baseline fluid-structure-thermal coupling solution conditions for CPU heat dissipation, the baseline flow field solution is performed to obtain the temperature field results of the CPU heat dissipation system under the initial air duct topology flow field model. Based on the temperature field results, the airflow state of the fan outlet neighborhood, heat sink inlet area, heat sink middle channel, heat sink tail area, chassis air outlet neighborhood, and reconfigurable air duct component neighborhood is further extracted to determine the airflow direction, airflow speed, airflow stagnation area, and airflow bypass area in each area, thus obtaining the velocity field results corresponding to the CPU cooling system. Based on the velocity field results, the vortex field results are obtained by identifying the vortex field in the fan outlet impact area, the heat sink fin gap shear area, the airflow guide component backwind area, the chassis wall near-wall area, and the air outlet confluence area. Based on the vortex field results, the local vortex location, vortex duration range, vortex retention tendency, and vortex interference degree of the main flow channel are determined, and the vortex field results corresponding to the CPU cooling system are obtained. Based on the temperature field results, the highest temperature location, hot spot distribution range, temperature change trend of hot spot diffusion to the heat dissipation interface layer, and temperature change trend of hot spot transmission to the heat sink base in the CPU core area are extracted. It is then determined whether the CPU core hot spot temperature gradient is concentrated in a local area, and the area where the temperature gradient is concentrated is marked as the hot spot heat conduction bottleneck area, thus obtaining the CPU core hot spot temperature gradient characteristics. Based on the velocity field results, the airflow penetration state from the inlet to the outlet of the heat sink fins is extracted to determine whether the fan airflow effectively enters the gap between the heat sink fins, and to determine whether there are airflow short circuits, local low speeds, and insufficient heat exchange at the tail between the heat sink fins. The fin channels with insufficient penetration airflow and corresponding high temperatures are marked as fin heat exchange bottleneck areas, and the characteristics of heat sink fin penetration airflow are obtained. Based on the vorticity field and velocity field results, the return flow location, return flow direction, return flow range, and degree of interference of the return flow on the main air intake channel are extracted. It is determined whether there is reverse airflow circulation on the leeward side of the reconfigurable duct component, the corner of the chassis, and the leading edge of the air outlet. The area with high return flow intensity and obstruction of hot air exhaust is marked as the duct return flow bottleneck area, and the duct return flow intensity characteristics are obtained. Based on the CPU core hot spot temperature gradient characteristics, heat sink fin penetration wind speed characteristics, and airflow return intensity characteristics, the flow field bottleneck characteristics are merged according to the heat source side bottleneck, heat exchange side bottleneck, and exhaust side bottleneck to generate a CPU heat dissipation flow field bottleneck diagnosis set.

5. The method for optimizing computer CPU heat dissipation according to claim 1, characterized in that, The construction of a reconfigurable parameter vector for the airflow topology based on the CPU heat dissipation flow field bottleneck diagnosis set includes: Based on the CPU heat dissipation flow field bottleneck diagnosis set, the spatial location, bottleneck type, influence range and corresponding topological connection edge of the hot spot heat conduction bottleneck area, fin heat exchange bottleneck area and air duct return flow bottleneck area are read, and each type of bottleneck is associated with the flow field topology node and topological connection edge in the initial air duct topology flow field model to generate flow field bottleneck topology location results. Based on the topological location results of the flow field bottleneck, the set of regions affected by airflow duct reconstruction is obtained; Based on the set of regions where the wind duct can be reconstructed, a wind duct topology reconstructable parameter vector is constructed to obtain the wind duct topology reconstructable parameter results. Based on the reconfigurable parameters of the duct topology, duct size constraints are set to obtain the duct size constraint results. Fan power consumption constraints are set based on the reconfigurable parameters of the air duct topology, and the fan power consumption constraint results are obtained. Based on the reconfigurable parameters of the airflow topology, CPU safety temperature constraints are set to ensure that the maximum CPU core temperature, average CPU junction temperature, and hot spot duration are all within the safe operating range allowed by the CPU cooling system. Furthermore, any candidate airflow topology reconfiguration scheme must not cause the CPU core hot spot temperature to exceed the upper limit of the safety temperature at the expense of reducing local pressure loss during the optimization process, thus obtaining the CPU safety temperature constraint results. Based on the constraints of airflow size, fan power consumption, and CPU safe temperature, feasibility marking is performed on each parameter in the airflow topology reconfigurable parameter results. Parameters that satisfy all constraints are marked as searchable variables, parameters that violate any constraint are marked as prohibited variables, and parameters that are close to the constraint boundary but are still adjustable are marked as boundary variables, thus generating a set of CPU heat dissipation co-optimization variables.

6. The method for optimizing computer CPU heat dissipation according to claim 1, characterized in that, The initialization of the rare trevally optimization algorithm population based on the CPU heat dissipation collaborative optimization variable set includes: Based on the CPU heat dissipation collaborative optimization variable set, the reconfigurable parameters of the air duct topology marked as searchable variables and boundary variables are read, and the reconfigurable parameters of the air duct topology marked as prohibited variables are excluded. The flow guide angle parameter, diffusion angle parameter, baffle opening and closing parameter, bypass air duct width parameter, fin gap correction parameter, fan speed control parameter, and hot spot neighborhood flow guide weight parameter are sorted according to the corresponding flow field topology nodes and topology connection edges to generate the individual coding variable sequence of the rare trevally. Based on the individual coding variable sequence of the rare trevally, an individual expression method for candidate air duct topology reconstruction schemes is established, so that each rare trevally individual contains a complete set of guide angle adjustment state, diffusion angle adjustment state, baffle opening and closing state, bypass air duct width state, fin gap correction state, fan speed control state, and hot spot neighborhood guide weight state. Each rare trevally individual can uniquely correspond to a candidate air duct topology reconstruction scheme that can be verified by fluid-structure-thermal coupling, and the coding result of the rare trevally individual candidate scheme is obtained. Based on the coding results of the candidate schemes for the rare trevally individuals, population initialization is performed. Within the allowable range of air duct size constraints, fan power consumption constraints, and CPU safe temperature constraints, multiple rare trevally individuals are generated. These multiple rare trevally individuals are differentiated in terms of guide angle, bypass duct width, baffle opening and closing state, fan speed control amount, and hot spot neighborhood guide weight, so as to cover three candidate reconstruction directions: hot spot guide enhancement, fin penetration improvement, and backflow suppression, to obtain the initial rare trevally topological population results. Based on the initial rare trevally topological population results, the candidate wind tunnel topology reconstruction scheme corresponding to each rare trevally individual is mapped back to the initial wind tunnel topology flow field model. The corresponding topology connecting edges are marked with state tags for expansion, contraction, opening and closing, bypass, guide angle correction or fin gap correction, and the flow field topology nodes and topology connecting edges that have not been changed by the rare trevally individual are kept in their original state, thus obtaining a set of candidate wind tunnel topology flow field models. The candidate air duct topology reconstruction schemes are evaluated based on the candidate air duct topology flow field model set. The maximum CPU core temperature, average CPU junction temperature, heat dissipation thermal resistance and airflow uniformity under each candidate air duct topology reconstruction scheme are obtained, and a multi-objective evaluation index set of candidate schemes is obtained. A multi-objective fitness function was constructed based on a set of multi-objective evaluation indicators for candidate schemes to obtain the fitness evaluation results of rare trevally individuals. Based on the fitness evaluation results of individual rare trevally, the rare trevally individuals in the initial rare trevally topological population results are sorted, labeled and archived. Rare trevally individuals that meet the constraints and have better fitness evaluation results are retained. At the same time, the candidate windway topology reconstruction scheme, variable value status, topology connection edge change status and multi-objective evaluation index corresponding to each rare trevally individual are recorded to obtain the rare trevally candidate topological population set.

7. The method for optimizing computer CPU heat dissipation according to claim 1, characterized in that, The step of performing rare trevally optimization iterative search based on the candidate topological population set of rare trevally includes: Based on the candidate topology population set of rare trevally, the candidate air duct topology reconstruction scheme, multi-objective fitness function value, maximum CPU core temperature, average CPU junction temperature, thermal resistance, airflow uniformity, air duct topology connection edge status and constraint satisfaction status corresponding to each rare trevally individual are read, and rare trevally individuals that do not meet the air duct size constraint, fan power consumption constraint or CPU safe temperature constraint are removed from the current iteration valid search objects to generate the current valid rare trevally search population; Based on the current effective search population of rare trevally, the rare trevally individuals are sorted from best to worst according to their multi-objective fitness function values. Rare trevally individuals with fitness function values ​​in the advantageous range and simultaneously have lower maximum CPU core temperature, lower average CPU junction temperature, and lower thermal resistance are classified as leading rare trevally individuals. Rare trevally individuals with fitness function values ​​in the middle range and still have room for improvement are classified as following rare trevally individuals. Rare trevally individuals with fitness function values ​​in the disadvantageous range or with high corresponding airflow intensity are classified as disturbing rare trevally individuals. The results of the rare trevally individual role classification are obtained. Based on the role classification of the individual rare tuna, the state of the topology connection edge that contributes to the reduction of heat dissipation thermal resistance and the reduction of airflow pressure loss in its candidate airflow topology reconstruction scheme is read. The guide angle, diffusion angle, bypass airflow width and baffle opening and closing state are adjusted around the airflow topology regions corresponding to the heat dissipation fin inlet, heat dissipation fin tail, chassis air outlet neighborhood and low resistance exhaust path, so that the leading rare tuna individual migrates towards the airflow topology region with low thermal resistance and low pressure loss, and the migration result of the leading rare tuna is obtained. Based on the following trevally individuals in the individual role classification results, the heat spot neighborhood flow guidance weight and the high heat transfer area location corresponding to the CPU core heat spot are read. The candidate airflow topology reconstruction scheme corresponding to the following trevally individuals is moved closer to the flow guidance state related to the heat spot temperature drop in the leading trevally individuals. The flow guidance parameters close to the CPU core heat spot, above the heat sink base, the heat sink fin inlet, and the heat spot neighborhood channel are adjusted first, so that the following trevally individuals gather in the high heat transfer area corresponding to the CPU core heat spot according to the heat spot neighborhood flow guidance weight, and the following trevally gathering result is obtained. Based on the disturbed trevally individuals in the individual role classification results, the local vortex residence time and airflow recirculation intensity in the corresponding candidate airflow topology reconstruction scheme are read. The topology connection edges that cause hot air retention, reverse circulation of the main channel, or poor exhaust at the tail of the heat sink fins are identified. The state jump is performed on the topology connection edges, adjusting the bypass airflow duct that was originally in the closed state to the open candidate state, adjusting the baffle that originally blocked the exhaust to the weakening candidate state, and adjusting the guide angle that originally generated the recirculation to the correction candidate state, thus obtaining the disturbed trevally topology jump results. Based on the migration results of the leading trevally, the aggregation results of the following trevally, and the topology jump results of the disturbed trevally, a constraint consistency check is performed on the guide angle state, baffle opening and closing state, bypass ventilation duct width state, and fan speed control state that may conflict within the same trevally individual. This ensures that the optimized trevally individual still satisfies the duct size constraint, fan power consumption constraint, and CPU safe temperature constraint. Unrealizable or contradictory topology change states are deleted, resulting in the trevally iteratively updated population results. Based on the iterative update of the rare trevally population results, the multi-objective fitness function values ​​of each rare trevally individual are re-evaluated, and candidate airflow topology reconstruction schemes that can reduce the maximum CPU core temperature, reduce the average CPU junction temperature, reduce heat dissipation thermal resistance, or improve airflow uniformity are retained and archived as optimization candidate solutions for the current iteration, thus obtaining the rare trevally optimized airflow candidate solution set.

8. The method for optimizing computer CPU heat dissipation according to claim 1, characterized in that, The adaptive airway topology reconstruction based on the candidate solution set of the rare carp optimization includes: Based on the Zhenyu optimization of the candidate solution set for the air duct, the topology connection edge change status, CPU core maximum temperature change status, CPU average junction temperature change status, heat dissipation thermal resistance change status, airflow uniformity change status, air duct return flow intensity change status and constraint satisfaction status corresponding to each candidate air duct topology reconstruction scheme are read. The topology connection edges that appear repeatedly in each candidate air duct topology reconstruction scheme and have a stable contribution to heat dissipation improvement are taken as priority reconstruction objects, and a set of candidate edges for air duct topology reconstruction is generated. Based on the candidate edge set of airflow topology reconstruction, topology connection edges adjacent to the CPU core hot spot neighborhood, the top of the heat sink base, the heat sink fin inlet, and the local high-temperature fluid retention area are identified. It is determined whether the topology connection edge in the Zhenyu optimized airflow candidate solution set continuously corresponds to the decrease of the CPU core maximum temperature, the decrease of the CPU average junction temperature, or the enhancement of heat transfer in the hot spot area. When it meets the hot spot temperature decrease contribution threshold, the topology connection edge is marked as a hot spot cooling contribution edge, and the hot spot cooling contribution edge marking result is obtained. Based on the hot spot cooling contribution edge labeling results, the hot spot cooling contribution edge is subjected to air duct expansion or flow enhancement processing to obtain the hot spot flow enhancement reconstruction result. Based on the candidate edge set of the air duct topology reconstruction, identify topology connection edges with sufficient air volume but low contribution to the reduction of the maximum CPU core temperature, the reduction of the average CPU junction temperature, or the improvement of airflow uniformity. Then, determine whether the topology connection edge is in a state of excessive ventilation, bypass short circuit, ineffective exhaust, or additional burden on fan power consumption. When it meets the pressure loss redundancy threshold, mark the topology connection edge as a pressure loss redundant edge and obtain the pressure loss redundant edge marking result. Based on the marking results of the pressure loss redundant edge, the pressure loss redundant edge is subjected to air duct contraction or baffle weakening to obtain the pressure loss redundancy suppression and reconstruction results. Based on the candidate edge set of wind duct topology reconstruction, the topological connection edge of the backflow vortex region is identified and generated, and the topological connection edge is marked as the backflow vortex region edge to obtain the backflow vortex region edge marking result. Based on the backflow vortex region edge marking results, the backflow vortex region edge is processed by adding a bypass ventilation duct or correcting the guide angle to obtain the backflow vortex region suppression and reconstruction results. Based on the candidate edge set of duct topology reconstruction, identify topological connection edges with low heat transfer contribution and obtain the low heat transfer contribution edge labeling results; Based on the low heat transfer contribution edge marking results, the low heat transfer contribution edges are merged or deleted. The hot spot flow enhancement reconstruction results, pressure loss redundancy suppression reconstruction results, backflow vortex suppression reconstruction results, and low heat transfer contribution edge processing results are then uniformly collected to generate an adaptive reconstructed duct topology set.

9. A computer CPU heat dissipation optimization device, characterized in that, It includes a data acquisition component (100), a boundary processing component (200), a flow field modeling component (300), a bottleneck diagnosis component (400), an optimization search component (500), a duct reconstruction component (600), and a control output component (700). The data acquisition component (100) is connected to the boundary processing component (200), the boundary processing component (200) is connected to the flow field modeling component (300), the flow field modeling component (300) is connected to the bottleneck diagnosis component (400), the bottleneck diagnosis component (400) is connected to the optimization search component (500), the optimization search component (500) is connected to the air duct reconstruction component (600), the air duct reconstruction component (600) is connected to the control output component (700), and the control output component (700) is used to output air duct adjustment parameters, fan control parameters, and flow guidance control parameters.

Images