Search

CN-121278252-B - Circuit board thermal management method based on multistage cooperation

CN121278252BCN 121278252 BCN121278252 BCN 121278252BCN-121278252-B

Abstract

The invention discloses a circuit board thermal management method based on multistage cooperation, which comprises the steps of constructing a thermal field real-time acquisition system based on a multi-mode sensor array, fusing spatial layout, thermal state data and functional attributes to form a structured thermal field sampling data set, carrying out normalization, filtering and feature dimension reduction processing on thermal field data to improve modeling accuracy and sensitivity, establishing a circuit board thermal field topology map based on the processed data, extracting structural features such as node centrality, clustering coefficients and connectivity to realize thermal distribution pattern recognition, dynamically generating a multi-objective optimized thermal regulation conflict resolution instruction according to a topology feature matching priority strategy library, and continuously optimizing strategy mapping through a feedback and online learning mechanism.

Inventors

  • ZHOU XUAN
  • ZHOU ZHINING
  • HOU DONGXIAO

Assignees

  • 梅州智科电路板有限公司

Dates

Publication Date
20260512
Application Date
20250922

Claims (10)

  1. 1. The circuit board thermal management method based on multistage cooperation is characterized by comprising the following steps of: S1, acquiring temperature, power density and heat flux density data of a plurality of heat source areas and a heat dissipation unit on a circuit board, and recording the spatial position and functional attribute of each sampling point by combining physical layout information of a thermal management system; S2, carrying out normalization and outlier filtering treatment on the acquired thermal field data to obtain preprocessed thermal field data; s3, constructing a circuit board thermal field topological graph based on the preprocessed thermal field data, wherein graph nodes represent heat sources or heat dissipation units, and graph edges represent heat flow coupling strength and direction; S4, extracting structural features of the circuit board thermal field topological graph, wherein the structural features comprise node centrality, clustering coefficients and heat flow path connectivity and are used for representing the type of the current heat distribution mode; S5, inputting the extracted topological features into a preset conflict resolution strategy matching module, and matching corresponding priority scheduling rules and feedback adjustment parameter sets in a strategy library based on feature similarity; S6, generating a preliminary conflict resolution instruction according to a matching result, wherein the preliminary conflict resolution instruction comprises multi-objective optimization weight distribution, a thermal regulation action sequence and execution priority sequencing; s7, continuously collecting circuit board thermal field state feedback data after executing the preliminary conflict resolution instruction, and calculating a deviation index between actual thermal field change and a prediction model; And S8, updating the mapping relation between the topological characteristic and the conflict resolution strategy through a lightweight online learning mechanism based on the deviation index.
  2. 2. The circuit board thermal management method based on multistage cooperation according to claim 1, wherein the step S1 specifically comprises: based on a physical layout design diagram of a thermal management system, topological position coordinates of each heat source area and a heat dissipation unit on a circuit board are obtained, and a space positioning information data set is formed; Disposing a multi-mode sensor array in each heat source area and each heat dissipation unit, and synchronously collecting temperature, power density and heat flux density to obtain a multi-dimensional thermal field state data stream; Performing zero drift compensation and nonlinear error correction on the acquired original thermal field data through a sensor calibration algorithm; Based on the function module division information of the heat source area, performing function attribute labeling on the acquisition points to generate a function-thermal state mapping table; And carrying out multi-source data fusion processing on the space positioning information data set, the multi-dimensional thermal field state data stream and the function-thermal state mapping table to generate a structured thermal field sampling data set.
  3. 3. The circuit board thermal management method based on multistage cooperation according to claim 2, wherein in the step S1, the multi-mode sensor array is arranged in a partitioned manner by adopting a density clustering and greedy coverage algorithm, the neighborhood radius is 1-10mm, the coverage radius is 1-5mm, the fault tolerance ratio is 1% -10%, and the redundancy per node is reduced by not less than 30%.
  4. 4. The circuit board thermal management method based on multistage cooperation according to claim 1, wherein the step S2 specifically comprises: Normalizing the acquired temperature, power density and heat flux density data of each region of the circuit board, and linearly transforming the multidimensional heterogeneous data to obtain a standardized thermal field data matrix; Performing time sequence filtering operation on the standardized thermal field data matrix based on a sliding window mechanism, and detecting and correcting abnormal values of a temperature and power density sequence by adopting an improved median filtering algorithm to obtain a filtered thermal field data set; performing spatial consistency check on each sampling point in the filtered thermal field data set, judging whether a local sensor drift phenomenon exists or not based on a gradient change threshold value of the heat flux density of an adjacent region, and dynamically compensating current sampling point data if drift is detected; Constructing a multidimensional feature vector space based on the corrected space consistency thermal field data, compressing the data dimension by adopting a principal component analysis method, and extracting main thermal field change characteristics; And performing dynamic window sliding average value calculation on the thermal field characteristic vector subjected to the principal component analysis and dimension reduction to generate a thermal field state sequence.
  5. 5. The method for thermal management of circuit boards based on multi-level collaboration according to claim 4, wherein the linear transformation in step S2 is specifically a linear transformation of multidimensional heterogeneous data based on a min-max normalization algorithm, and each type of physical quantity normalization interval is [0,1], and linear interval mapping of data under different physical dimensions is performed.
  6. 6. The circuit board thermal management method based on multistage cooperation according to claim 1, wherein the step S3 specifically comprises: Performing functional attribute classification on the heat source and the heat dissipation unit in the processed thermal field data, dividing a topological node set based on the spatial position information of the physical layout of the heat source and the heat dissipation unit, and establishing a graph structure basic representation of the thermal behavior of the circuit board; based on the Fourier heat conduction relation between the heat flux density and the temperature gradient, calculating the heat flux coupling strength between the adjacent heat source and the heat dissipation unit, and obtaining the quantitative representation of the heat flux transfer path; Directional modeling is carried out on the heat flow transmission path by adopting a directional diagram modeling method, and a heat flow transmission direction matrix is constructed; constructing a weighted directed graph model based on the heat flow coupling strength and the direction information between the nodes, wherein the nodes represent heat sources or heat dissipation units, and the edges represent heat flow coupling paths and the directivities thereof to form a primary structure of a thermal field topological graph; And (3) performing graph structure optimization processing on the constructed thermal field topological graph, and removing redundant heat flow paths by adopting a minimum spanning tree algorithm.
  7. 7. The method of claim 6, wherein the step S3 further comprises calculating the heat flow coupling strength between the nodes through fourier heat conduction relation and node space center distance, limiting the strength to [0,1] through normalization operation, determining the heat flow direction by combining the temperature difference, and performing node outflow unification by adopting weighted directional adjacency matrix and row normalization mode to finally form a weighted directional diagram.
  8. 8. The method for thermal management of circuit boards based on multi-stage collaboration according to claim 1, wherein the step S4 specifically includes: Carrying out node centrality calculation on graph nodes in the constructed circuit board thermal field topology map, and quantifying the relative importance of each heat source or heat dissipation unit in a heat flow network based on node degree centrality and a near centrality index to obtain a node influence characteristic vector; Based on the local adjacency relation in the map, calculating the clustering coefficient of each node, extracting the local aggregation characteristic of the thermal field, and forming a local thermal interaction strength characteristic index; Performing topology connectivity analysis on the heat flow path, and calculating the heat flow accessibility between any two nodes in the graph by adopting a shortest path algorithm to obtain a global heat flow path connectivity matrix; carrying out multidimensional feature fusion on the node centrality, the clustering coefficient and the heat flow path connectivity matrix, and constructing a thermal field topological feature space to form a unified topological feature description vector; and carrying out pattern recognition classification processing on the topological feature description vector based on a preset thermal distribution pattern classification rule, judging whether the current thermal field belongs to a uniform, hot spot aggregation type or gradient mutation type thermal distribution pattern, and outputting a pattern classification label.
  9. 9. The method for thermal management of circuit boards based on multi-stage collaboration according to claim 1, wherein the step S5 specifically comprises: feature coding is carried out on structural features of the thermal field topological graph, feature vectors are generated based on node centrality, clustering coefficients and heat flow path connectivity, and digital representation of a current heat distribution mode is constructed; executing a similarity measurement algorithm based on the feature vector, carrying out matching calculation on feature templates corresponding to various heat distribution modes in a pre-stored strategy library, and identifying the similarity degree of the current heat field topology and the existing modes in the strategy library; Screening a group of candidate strategy templates with the matching degree higher than a set threshold according to the similarity measurement result, and taking the candidate strategy templates as an input candidate set of a conflict resolution strategy matching module; based on the priority scheduling rule and the feedback regulation parameter set of the candidate strategy template, executing a multi-objective optimization weight distribution algorithm to generate a multi-level cooperative regulation strategy combination which is adaptive to the current heat distribution mode; And outputting the multi-stage cooperative regulation strategy combination to a conflict resolution instruction generation module.
  10. 10. The method of claim 9, wherein the step S5 further comprises weighting euclidean distance and cosine similarity double-index fusion by using feature vectors, introducing an empirical weight coefficient, and performing hash fingerprint de-duplication on repeated templates with a comprehensive similarity matching threshold of 0.8-1.0, wherein the candidate policy scale is controlled within 20% of the total template library.

Description

Circuit board thermal management method based on multistage cooperation Technical Field The invention relates to the technical field of circuit board thermal management, in particular to a circuit board thermal management method based on multistage cooperation. Background With the rapid development of electronic products, the integration level and power consumption of circuit boards are continuously improved, and thermal management is becoming a key factor affecting the performance and reliability of the system. The main stream heat management system generally adopts a multi-target optimization method, and realizes circuit board temperature control, energy consumption management and reliability guarantee by adjusting strategies such as fan rotation speed, heat flow distribution, power consumption distribution and the like in real time. Meanwhile, multistage cooperative thermal regulation has become an important trend of high-end circuit board thermal management, namely, fine heat distribution management and resource sharing are realized through intelligent scheduling decomposition among board level, module level and even system level; The existing public schemes are mostly based on rule scheduling based on threshold values, mixed integer programming algorithm or static strategy library mapping, and partial high-end schemes are combined with finite element thermal simulation or thermal prediction models based on machine learning so as to achieve higher precision and better responsiveness. In recent years, the industry focuses on improving the processing capacity of a thermal management system on complex working conditions (such as transient high power consumption, non-uniform hot spot distribution, heat dissipation unbalance and the like) through stronger data fusion, self-adaptive scheduling and feedback mechanisms; Despite the advances made by the existing multi-stage thermal management and conflict resolution mechanisms for circuit boards, the following prominent technical drawbacks and unmet needs remain: (1) The strategy is stiff, the traditional rule driving or threshold value recognition conflict resolution method is difficult to adjust autonomously according to different thermal field distribution modes, the regulation and control logic of the method depends on expert experience or a solidification threshold value, and response is slow or misjudgment occurs when the method faces to variable heterogeneous heat flow and sudden hot spot transfer of a circuit board; (2) The existing regulation and control scheme is insufficient in topology perception, and the core topology features such as global structures, hot spot aggregation paths, heat flow migration trends and the like in a heat flow network cannot be automatically identified, so that systematic potential conflicts (such as multi-heat source coupling, local resource exhaustion and the like) are difficult to find, and therefore, the adopted resolution measures are limited to local optima; (3) The strategy matching is single, and at present, a plurality of strategies can only be matched with a static template or a preset working condition, and the awareness and evolution capability of the real-time change of the thermal management environment are lacked, so that the on-line dynamic learning and behavior migration are difficult to support, and the accuracy, the robustness and the energy consumption performance of the thermal regulation strategy are limited; (4) Feedback and self-adaptation capability are weak, and although partial schemes introduce feedback control, most of the schemes fail to realize global-oriented online self-optimization based on multi-dimensional thermal field deviation and topology state fusion. The system can not quickly correct strategies and parameters in the situations of sudden load, heat dissipation blockage and the like, and the problems of local overheating, energy consumption increase and the like are caused. Disclosure of Invention The invention aims to solve the technical problems and provides a circuit board thermal management method based on multistage cooperation. The technical scheme of the invention is realized by a circuit board thermal management method based on multistage cooperation, which comprises the following steps: S1, acquiring temperature, power density and heat flux density data of a plurality of heat source areas and a heat dissipation unit on a circuit board, and recording the spatial position and functional attribute of each sampling point by combining physical layout information of a thermal management system; s2, carrying out normalization and outlier filtering processing on the acquired thermal field data so as to eliminate the influence of transient disturbance and sensor drift on the subsequent modeling precision; S3, constructing a circuit board thermal field topological graph based on the processed thermal field data, wherein graph nodes represent heat sources or heat dissipation units, and