CN-121145642-B - Optimization method of integrated micro-module based on complex power distribution

CN121145642BCN 121145642 BCN121145642 BCN 121145642BCN-121145642-B

Abstract

The invention discloses an optimization method of an integrated micro-module based on complex power distribution, which relates to the technical field of integrated circuit thermal management, and comprises the steps of constructing a power distribution feature matrix based on a structured data set through multi-physical field coupling analysis, and generating micro-channel geometric parameters by adopting a gradient descent algorithm; the method comprises the steps of constructing a structural heat conduction efficiency mapping model, obtaining heat resistance characteristic change data of a flow channel section by combining geometric parameters of a micro flow channel, outputting a heat conduction efficiency optimization matrix, extracting heat resistance distribution characteristics in the heat conduction efficiency optimization matrix, generating a collaborative optimization parameter instruction by combining the heat resistance characteristic change data of the flow channel section, executing the collaborative optimization parameter instruction to obtain heat power distribution characteristics, calculating equivalent heat resistance gradient, and obtaining a heat dissipation efficiency regulation instruction through dynamic mapping. According to the invention, by constructing the structural heat conduction efficiency mapping model, the real-time prediction from the geometric parameters to the thermal properties of the micro-channels is realized, the efficiency is improved, and the global performance is ensured to be optimal.

Inventors

LU DONGYUE
PENG XINGJIE
LIU DINGZHU

Assignees

北京摩纳科技有限责任公司

Dates

Publication Date: 20260508
Application Date: 20250911

Claims (6)

1. An optimization method of an integrated micro module based on complex power distribution is characterized by comprising the following steps of, Collecting monitoring operation data and obtaining a structured data set through preprocessing; the monitoring operation data comprise equipment performance indexes, environment parameters and equipment operation states; the performance index of the equipment is mainly obtained by actively inquiring the voltage and the current of the equipment through monitoring agents deployed on a server or through SNMP and JMX protocols; The environmental parameters are collected by a temperature sensor, a vibration sensor or an environmental monitoring sensor which are deployed in the equipment or a machine room; the running state of the equipment obtains the running time, the reference temperature value and the reference power of the equipment by using the application and log file, analyzing the network flow and calling an API (application program interface); the structured dataset comprises an ambient temperature value, a power value, a spatiotemporal feature tensor, and an equipment operating parameter; The environment temperature value is obtained by removing abnormal values of the average value and the standard deviation of the environment parameters through data cleaning and scale conversion through normalization processing; the power value comprises active power and reactive power, and the acquired voltage and current are converted into voltage and current values with actual physical units through normalization processing; The space-time characteristic tensor is obtained by integrating and structurally expressing monitoring operation data of different space position sensors along with time variation; the equipment operation parameters are obtained by performing data cleaning on the equipment operation state to process missing values and obvious abnormal points and performing data integration and normalization processing; constructing a power distribution feature matrix by the structured data set through multi-physical field coupling analysis, and generating micro-channel geometric parameters by adopting a gradient descent algorithm; A structural heat conduction efficiency mapping model is constructed, geometric parameters of a micro-channel are input to obtain heat resistance characteristic change data of the cross section of the channel, and a heat conduction efficiency optimizing matrix is output, specifically comprising the following steps of, Carrying out feature extraction on geometric parameters of the micro-channel by adopting a tensor decomposition method to generate a geometric feature tensor; Generating a fluid temperature field and a structural thermal stress field through bidirectional thermal coupling analysis based on geometric feature tensors, and generating a global energy flow density distribution field through a spatial gradient algorithm; calculating grid equivalent thermal resistance according to the global energy flux density distribution field to generate a thermal resistance characteristic distribution field, and generating a statistical feature vector through multi-order moment statistics; acquiring a structural heat conduction efficiency mapping model by constructing a heat resistance geometrical efficiency correlation network based on a heat resistance characteristic distribution field and a statistical feature vector; Inputting geometric parameters of a micro-channel into a structural heat conduction efficiency mapping model to obtain heat resistance characteristic change data of a channel section, and adopting a multi-target particle swarm optimization algorithm to perform non-dominant sorting and output a heat conduction efficiency optimization matrix; Extracting thermal resistance distribution characteristics in the thermal conduction efficiency optimization matrix, and generating a collaborative optimization parameter instruction by combining the thermal resistance characteristic change data of the flow passage section; executing the collaborative optimization parameter instruction to obtain the thermal power distribution characteristics, calculating the equivalent thermal resistance gradient, obtaining the heat dissipation efficiency regulation instruction through dynamic mapping, specifically comprising the following steps, Executing the collaborative optimization parameter instruction to generate temperature field distribution data, and calculating a space temperature gradient by adopting a recursive least square method to acquire thermal power distribution characteristics; based on the thermal power distribution characteristics, calculating an equivalent thermal resistance gradient by combining temperature field distribution data; And fusing the thermal power distribution characteristics and the equivalent thermal resistance gradient into a thermal state joint characteristic vector, and acquiring a heat dissipation efficiency regulation instruction through an optimization learning algorithm.
2. The method for optimizing integrated micro-modules based on complex power distribution of claim 1, wherein the monitoring operation data comprises equipment performance index, environment parameters and equipment operation state; the preprocessing includes data cleaning, normalization processing and space-time coding.
3. The method for optimizing integrated micro-modules based on complex power distribution of claim 2 wherein the structured dataset comprises ambient temperature values, power values, spatiotemporal feature tensors, and device operating parameters; The structured dataset is analyzed by multiple physical fields to construct a power distribution feature matrix, which comprises the following steps, Inputting the structured dataset into a multi-physical field coupling equation to perform transient solution, and outputting global physical field distribution data; Based on the global physical field distribution data, energy flow density vectors are calculated, and a power distribution feature matrix is constructed through spatial discretization.
4. The method for optimizing integrated micro-modules based on complex power distribution of claim 3, wherein said micro-channel geometric parameters are generated by extracting statistical characteristics in a power distribution characteristic matrix, calculating power distribution uniformity, and performing topology optimization.
5. The method for optimizing an integrated micro-module based on complex power distribution as claimed in claim 4, wherein the extracting the thermal resistance distribution characteristics in the thermal conduction efficiency optimization matrix is analyzing a gradient change rule of the thermal resistance distribution characteristics by adopting an asymmetric characteristic extraction algorithm, and extracting the thermal resistance distribution characteristics in the thermal conduction efficiency optimization matrix.
6. The method for optimizing integrated micro-modules based on complex power distribution of claim 5, wherein the step of generating a collaborative optimization parameter instruction by combining the flow passage section thermal resistance characteristic change data and the thermal resistance distribution characteristics comprises the following steps of, Performing countermeasure fusion on the thermal resistance distribution characteristics and the thermal resistance characteristic change data of the flow channel section to construct a thermal resistance cooperative characteristic map; and extracting space-time coupling thermal resistance distribution in the thermal resistance collaborative feature map, and generating a collaborative optimization parameter instruction by positioning a thermal conduction critical path.

Description

Optimization method of integrated micro-module based on complex power distribution Technical Field The invention relates to the technical field of integrated circuit thermal management, in particular to an optimization method of an integrated micro module based on complex power distribution. Background With the rapid development of three-dimensional integrated circuits and high power density electronic devices, thermal management has become a key technology limiting performance and reliability. The micro-channel cooling technology is widely applied to the fields of integrated circuits, power modules and advanced packaging due to the high-efficiency heat dissipation capability. The traditional micro-channel design method mainly depends on parameterized scanning and empirical rules, and in recent years, a multi-physical field coupling simulation technology is introduced into the micro-channel design flow, and the collaborative analysis of a temperature field, a flow field and a stress field is realized through coupling physical fields such as fluid mechanics, heat conduction and structural mechanics. In addition, the intelligent optimization algorithm is also applied to micro-channel optimization and is used for solving the problem of multi-objective and high-dimensional design space searching. However, most of the prior art is still limited to a single physical field or static power hypothesis, and fails to adequately cope with a complex power distribution scenario that is dynamic and non-uniform in actual operation. In the prior art, two bottlenecks mainly exist, on one hand, most optimization methods assume that power is uniformly distributed or based on an idealized heat source model, real-time power data and geometric design cannot be effectively coupled, so that local overheating occurs in a high-power gradient region, and channel space is wasted in a low-power region. On the other hand, the existing optimization framework has limitation on the integration level of multi-physical field coupling and intelligent decision, and although the multi-physical field simulation tool can realize electric-thermal-force coupling analysis, the simulation flow is usually separated from the optimization algorithm, multiple iterations are needed, and the calculation cost is high. Disclosure of Invention The present invention has been made in view of the above-described problems occurring in the prior art. Therefore, the invention provides an optimization method of an integrated micro module based on complex power distribution, which solves the problems of local overheating of a high-power gradient region and low intelligent optimization decision-making efficiency. In order to solve the technical problems, the invention provides the following technical scheme: the invention provides an optimization method of an integrated micro module based on complex power distribution, which comprises the following steps of, The method comprises the steps of collecting monitoring operation data, obtaining a structured data set through preprocessing, constructing a power distribution feature matrix through multi-physical field coupling analysis based on the structured data set, generating micro-channel geometric parameters through a gradient descent algorithm, constructing a structural heat conduction efficiency mapping model, obtaining heat resistance characteristic change data of a channel section by combining the micro-channel geometric parameters, outputting a heat conduction efficiency optimizing matrix, extracting heat resistance distribution features in the heat conduction efficiency optimizing matrix, generating a collaborative optimizing parameter instruction by combining the heat resistance characteristic change data of the channel section, executing the collaborative optimizing parameter instruction to obtain heat power distribution features, calculating equivalent heat resistance gradients, and obtaining a heat dissipation efficiency regulating instruction through dynamic mapping. As a preferable scheme of the optimization method of the integrated micro-module based on complex power distribution, the monitoring operation data comprises equipment performance indexes, environment parameters and equipment operation states; the preprocessing includes data cleaning, normalization processing and space-time coding. As a preferred scheme of the optimization method of the integrated micro-module based on complex power distribution, the structured data set comprises an environment temperature value, a power value, a space-time characteristic tensor and equipment operation parameters; The power distribution characteristic matrix is constructed by multi-physical field coupling analysis based on the structured dataset, and the specific steps are as follows, Inputting the structured dataset into a multi-physical field coupling equation to perform transient solution, and outputting global physical field distribution data; Based on the glob