Computer heat dissipation intelligent control method and system based on multi-sensor fusion

Abstract

This application discloses a computer heat dissipation intelligent control method and system based on multi-sensor fusion, relating to the field of sensor network chips. The method includes: collecting attitude angles and temperatures; converting attitude angles into gravity projection vectors; generating an initial heat distribution map and partition index by weighted mapping of temperature directions; extracting partition temperature features according to the partition index and calculating the aggregation index in combination with the gravity vector to determine candidate hotspot coordinates and temperature change statistics; comparing the change value with the change threshold, and constructing a neighborhood to extract historical windows if the value exceeds the threshold; calculating the temperature rise slope and neighborhood correlation coefficient to determine the aggregation start point and labeling the partition; extracting the neighborhood sequence and decomposing it according to the gravity vector to obtain the trend vector and migration direction; extrapolating the partition predicted temperature based on the reference temperature; calculating the predicted increment and comparing it with the enhancement threshold; and selecting the fan group to output the fan speed adjustment amount according to the migration direction if the value exceeds the threshold. This avoids the problems of local high temperature omission, heat dissipation resource misallocation, and response lag caused by neglecting attitude factors in existing solutions.

Description

Technical Field

[0001] This invention relates to the field of sensor network chip technology, and more specifically, to a computer heat dissipation intelligent control method and system based on multi-sensor fusion. Background Technology

[0002] In the field of modern electronic devices, the intelligent development of computer heat dissipation technology is particularly crucial. With the continuous improvement of portable device performance, heat dissipation directly affects the operational stability and lifespan of these devices. How to achieve efficient and precise heat management in complex usage environments has become a critical issue that the industry urgently needs to address. Especially in scenarios where users frequently move or adjust the device's position, the adaptability of the heat dissipation system is of paramount importance.

[0003] However, many current heat dissipation control methods often lack sufficient dynamic response capabilities when dealing with uneven heat distribution caused by changes in equipment posture. Existing solutions focus more on the overall monitoring of the internal temperature of the equipment, but ignore the specific impact of external environmental factors on heat distribution. In particular, when the equipment is tilted or the placement angle changes, the heat distribution will change significantly due to gravity. This neglect means that heat dissipation measures cannot intervene in local high-temperature areas in a timely manner, thereby affecting equipment performance and user experience.

[0004] The impact of device orientation changes on heat distribution has become a core challenge. Gravity causes hot air inside a device to concentrate in a specific direction when tilted, forming localized high-temperature areas. This concentration is not instantaneous but gradually intensifies over time. Without accurately capturing the location and trend of this heat concentration, effective measures cannot be taken before high temperatures develop. For example, when a user places a laptop on a tilted desk or lap, hot air may concentrate on one side of the device due to gravity, causing a rapid temperature rise in that area, while the heat dissipation resources of other areas are not fully utilized. This uneven heat distribution and the complexity of its temporal evolution present a pressing technical challenge that needs to be addressed.

[0005] Therefore, the key issue of this study is how to accurately identify the location of heat accumulation caused by gravity in dynamic scenarios of equipment attitude changes and predict its development trend in a timely manner. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: A computer heat dissipation intelligent control method based on multi-sensor fusion includes: S101. Collect the device's attitude angle and internal temperature data at each location, and convert the attitude angle into a gravity projection vector; perform directional weighted mapping on the temperature data based on the gravity projection vector to generate an initial heat distribution map and its spatial partition index. S102. Extract the temperature features of the zones from the initial heat distribution map based on the spatial zone index, and calculate the clustering index by combining the gravity projection vector; determine the coordinates of the candidate hotspot areas based on the clustering index, and output the temperature change statistics corresponding to the coordinates of the candidate hotspot areas; S103. Compare the temperature change statistics with the preset change threshold. If the temperature exceeds the preset change threshold, determine the spatial neighborhood using the coordinates of the candidate hotspot area and extract the historical temperature window within the spatial neighborhood. Calculate the temperature rise slope and neighborhood correlation coefficient based on the historical temperature window to determine the heat accumulation starting point and mark its corresponding partition. S104. Extract the neighborhood temperature sequence within the partition of the heat accumulation starting point, decompose the neighborhood temperature change according to the gravity projection vector, and obtain the heat development trend vector and the hot spot migration direction. S105. Using the reference temperature of each zone as the base value, the heat development trend vector as the extrapolation term, and the hot spot migration direction as the directional constraint, time extrapolation is performed according to the spatial zone index to obtain the zone temperature prediction value; the difference between the zone temperature prediction value and the reference temperature is used to obtain the prediction increment, and the prediction increment is compared with the preset enhancement threshold. If it exceeds the preset enhancement threshold, the corresponding fan group is determined according to the hot spot migration direction and the fan speed adjustment amount is generated.

[0007] Furthermore, methods for generating an initial heat distribution map and establishing a spatial partition index include: Collect equipment attitude angle and internal temperature data at various locations, record them according to a unified time base and establish a corresponding relationship with timestamps; Establish a device coordinate system based on the device attitude angle and transform the geographic gravity vector to obtain the gravity projection vector; A temperature acquisition location vector is established based on the gravity projection vector, and the projection coefficient is calculated and normalized to obtain the direction weight. Temperature data is weighted by direction and mapped to the equipment coordinate system to form a direction-weighted mapping result; The mapping results are weighted and summed, and the weights are normalized to generate an initial heat distribution map and establish a spatial partition index.

[0008] Furthermore, methods for calculating the clustering index include: Based on the spatial partition index, the grid temperature characterization values ​​of the initial heat distribution map are read and the partition grid sets are collected according to the partition identifier to calculate the partition temperature characteristics; Based on the temperature characteristics of the partition, the gravity projection vector at the same sampling time is read to determine the geometric center coordinates of the partition and calculate the partition orientation coefficient; Based on the temperature characteristics of the partition, the aggregation amount within the partition, the aggregation amount at the partition boundary, and the aggregation amount over time are determined. The aggregation amount within the partition, the aggregation amount at the partition boundary, and the aggregation amount over time are multiplied by the partition direction coefficient, and then summed according to the weight parameter table to obtain the aggregation index. The weight parameter table is preset and includes a first weight, a second weight, and a third weight, which correspond to the aggregation amount within the partition, the aggregation amount at the partition boundary, and the aggregation amount over time, respectively.

[0009] Furthermore, methods for outputting statistical values ​​of temperature changes corresponding to coordinates include: At the same sampling time, the clustering index of all partitions is obtained and the partition with the largest clustering index is determined as the candidate hotspot partition; Within the candidate hotspot partition, candidate hotspot grid cells are determined based on the initial heat distribution map, and the grid center coordinates of the candidate hotspot grid cells are determined as the coordinates of the candidate hotspot region. Under the difference threshold constraint, the coordinates of multiple grid center coordinates are weighted and summed according to the temperature characterization value to obtain the coordinates of the candidate hotspot region. Based on the coordinates and spatial partition index of the candidate hotspot region, the partition identifier and candidate hotspot grid cell of the candidate hotspot region are determined. The set of neighboring grid cells is determined based on the neighborhood radius threshold or the number of adjacent layers threshold. The temperature characterization value sequence is read from the initial heat distribution map at multiple consecutive sampling times and the temperature change statistics are output.

[0010] Furthermore, methods for capturing historical temperature windows within a spatial neighborhood include: Obtain the temperature change statistics associated with the coordinates of the candidate hotspot area, select the maximum temperature increment in the spatial neighborhood as the comparison value and compare it with the preset change threshold; When the comparison result exceeds the change threshold, the candidate hot spot grid cell corresponding to the coordinate is taken as the center, and the set of neighboring grid cells is expanded according to the neighborhood radius threshold or the number of adjacent layers threshold to form a spatial neighborhood. The spatial neighborhood is then limited to the grid cells of the partition where the coordinate is located and the grid cells of the boundary of the adjacent partition according to the spatial partition index. The temperature characterization value sequence is read from historical sampling records based on spatial neighborhood, and the starting time is determined by the window length threshold. The sequence from the starting time to the current sampling time is then extracted to obtain the historical temperature window.

[0011] Furthermore, methods for determining the starting point of heat accumulation and labeling its corresponding zone include: Based on the historical temperature window, the temperature sequence of candidate hot spot grid cells corresponding to the coordinates of candidate hot spot regions is extracted. The temperature rise slope sequence is calculated step by step according to the temperature difference between adjacent sampling times and the sampling interval. The maximum value of the temperature rise slope sequence is taken to determine the temperature rise slope. Based on the historical temperature window and spatial neighborhood, the temperature sequence of each neighborhood grid unit is extracted. The temperature sequence of the candidate hot spot grid unit is used as the reference sequence. The reference sequence and each neighborhood temperature sequence are subjected to mean removal and scale normalization to obtain a standardized sequence. The correlation coefficient is calculated by summing the point-by-point products and dividing by the number of sampling points to obtain the neighborhood correlation coefficient set. The neighborhood correlation coefficient is obtained by weighted average normalization by the inverse distance. Based on the temperature rise slope sequence, the first slope threshold exceeding the threshold sampling time is determined within the historical temperature window as the starting time. Based on the starting time, the temperature increment is calculated in the spatial neighborhood and compared with the increment threshold. The candidate starting point is determined by combining the correlation coefficient and the correlation threshold. The first grid cell in the sorted order is selected according to the temperature increment to determine the heat accumulation starting point. The partition to which the heat accumulation starting point belongs is marked based on the spatial partition index.

[0012] Furthermore, methods for extracting neighborhood temperature sequences include: Obtain the heat accumulation starting point and its corresponding partition identifier, and read the spatial partition index to obtain the set of partition grid cells corresponding to the partition; Based on the partitioned grid cell set, with the starting grid cell as the center, the adjacency expansion is performed according to the adjacency layer threshold and the adjacency relationship of the shared boundary of the grid cells to obtain the starting point neighborhood grid cell set; The boundary grid cells are determined based on the set of grid cells in the neighborhood of the starting point. The adjacent boundary grid cells that share the boundary with the boundary grid cells but have different partition identifiers are merged into the spatial partition index to obtain the extended set of grid cells in the neighborhood of the starting point. The temperature sequence is formed by reading the temperature characterization values ​​of the neighboring grid cells at the unified timestamp of the extended starting point. The temperature characterization value sequence of the starting point grid cells is then read to form the starting point temperature sequence.

[0013] Furthermore, methods for determining the direction of hotspot migration include: Based on the gravity projection vector corresponding to the same sampling time of the neighborhood temperature sequence, the gravity projection vector is represented as a three-axis component vector in the device coordinate system, including the first axis projection component, the second axis projection component, and the third axis projection component. The magnitude is calculated based on the first axis projection component, the second axis projection component, and the third axis projection component and compared with the magnitude threshold. When the magnitude is not less than the magnitude threshold, the projection component of each axis is divided by the magnitude to obtain the gravity direction unit vector. When the magnitude is less than the magnitude threshold, the third axis positive direction unit vector is determined as the gravity direction unit vector. The displacement vector of each neighboring grid cell is calculated based on the set of neighboring grid cells of the starting point and the unit vector of gravity direction. The projection is obtained by calculating the dot product of the displacement vector and the unit vector of gravity direction. The lateral displacement vector is obtained by subtracting the product of the projection and the unit vector of gravity direction from the displacement vector. The starting time and the current time are determined based on the neighborhood temperature sequence, and the temperature change of each neighborhood grid cell is calculated. The gravity weight is obtained by taking the projection as non-negative and normalizing it, the lateral weight is obtained by normalizing the magnitude of the lateral displacement vector, and the gravity temperature change component and the lateral temperature change component are obtained by weighting and summing the temperature change based on the gravity weight and the lateral weight respectively. The lateral composite vector is obtained by normalizing the lateral displacement vector with the non-negative values ​​of temperature change and then normalizing it to obtain the lateral unit direction. The heat development trend vector is obtained by multiplying the temperature change component under gravity with the unit vector of gravity and adding the result of multiplying the lateral temperature change component with the lateral unit direction. The direction of hotspot migration is determined by normalizing the direction of the heat development trend vector.

[0014] Furthermore, the methods for generating the fan speed adjustment include: The initial heat distribution map is read according to the spatial partition index and the average temperature of each partition is calculated. The average temperature of each partition is determined as the reference temperature of each partition. Based on the reference temperature of each zone, the extrapolation time step is determined according to the preset time step threshold. The total extrapolation increment is obtained based on the magnitude of the heat development trend vector and the extrapolation time step. The distribution coefficient is obtained by taking the non-negative and normalizing the dot product of the displacement vector from the geometric center of the zone to the heat accumulation starting point and the hot spot migration direction. The total extrapolation increment is distributed and added to the reference temperature of each zone to obtain the zone temperature prediction value. The predicted increment is obtained by subtracting the predicted temperature value of each zone from the baseline temperature of each zone, and compared with the enhancement threshold to obtain the set of zones exceeding the threshold. The fan group corresponding to the zone exceeding the threshold is determined based on the fixed association table of zone identifier and fan group identifier. When the set of zones exceeding the threshold includes multiple zones, the fan group corresponding to the zone with the largest dot product is selected. Based on the difference between the predicted increment and the enhancement threshold, the fan speed adjustment amount is output segment by segment according to the speed increment threshold table.

[0015] A computer heat dissipation intelligent control system based on multi-sensor fusion is used to execute a computer heat dissipation intelligent control method based on multi-sensor fusion. The system includes: The mapping module is used to collect the device's attitude angle and internal temperature data at various locations, convert the attitude angle into a gravity projection vector, and perform directional weighted mapping on the temperature data based on the gravity projection vector to generate an initial heat distribution map and its spatial partition index. The positioning module extracts the temperature features of the zones from the initial heat distribution map based on the spatial zone index, and calculates the clustering index by combining the gravity projection vector; it determines the coordinates of candidate hotspot areas based on the clustering index, and outputs the temperature change statistics corresponding to the coordinates of the candidate hotspot areas; The annotation module is used to compare the temperature change statistics with the preset change threshold. When the temperature exceeds the preset change threshold, the spatial neighborhood is determined by the coordinates of the candidate hotspot area, and a historical temperature window is extracted within the spatial neighborhood. Based on the historical temperature window, the temperature rise slope and the neighborhood correlation coefficient are calculated to determine the heat accumulation starting point and annotate its corresponding partition. The direction generation module is used to extract the neighborhood temperature sequence within its respective partition around the heat accumulation starting point, decompose the neighborhood temperature change according to the gravity projection vector, and obtain the heat development trend vector and the hot spot migration direction. The speed adjustment module is used to perform time extrapolation based on the spatial partition index, using the reference temperature of each partition as the base value, the heat development trend vector as the extrapolation term, and the hot spot migration direction as the directional constraint, to obtain the partition temperature prediction value; the difference between the partition temperature prediction value and the reference temperature is used to obtain the prediction increment, and the prediction increment is compared with the preset enhancement threshold. If it exceeds the preset enhancement threshold, the corresponding fan group is determined according to the hot spot migration direction and the fan speed adjustment amount is generated.

[0016] Compared with related technologies, the present invention has the following beneficial effects: By synchronously collecting device attitude angles and internal temperature data at multiple locations under the same time reference, the attitude angles are converted into gravity projection vectors in the device coordinate system. Based on this, the temperature data is subjected to directional weighted mapping to generate an initial heat distribution map and spatial partition index. This makes the temperature field no longer an overall average or a single-point monitoring, but a spatial heat expression explicitly constrained by the direction of gravity. Thus, in dynamic scenarios where the device tilts and placement angles change frequently, it can stably reflect the spatial non-uniformity caused by the accumulation of hot air along the direction of gravity, avoiding the problems of local high temperature omission, heat dissipation resource misallocation, and response lag caused by neglecting attitude factors in existing solutions.

[0017] Based on an initial heat distribution map and spatial partition index, this invention extracts partition temperature features and calculates the aggregation index using gravity projection vectors. Candidate hotspot partitions are first derived by weighting the aggregation amount within the partition, the aggregation amount at the partition boundary, the aggregation amount over time, and the partition direction coefficient. Then, the data is refined to grid coordinates within each partition, and temperature change statistics bound to the coordinates are output. When the change reaches the trigger condition, a spatial neighborhood is further constructed using coordinates, and a historical temperature window is extracted. The correlation coefficient between the temperature rise slope and the neighborhood is derived. The starting point of heat aggregation and its corresponding partition are determined using slope triggering, incremental constraints, and consistent correlation criteria. This effectively solves the pain point of not being able to accurately capture the location of heat aggregation and the process of its gradual intensification over time: it not only locates where the heat is, but also where it starts to rise and whether it is spatially synchronously expanding. This significantly reduces false triggers and missed triggers caused by attitude changes and provides a traceable starting point and partition label for subsequent directional heat dissipation.

[0018] This invention extracts the neighborhood temperature sequence around the heat accumulation starting point and decomposes the neighborhood temperature change into gravity-directed and lateral components using the unit vector of gravity direction as a unified benchmark. These components are then synthesized into a heat development trend vector and normalized to obtain the hotspot migration direction. Subsequently, using the reference temperature of each zone as the base value, the heat development trend vector as the extrapolation term, and the hotspot migration direction as the directional constraint, the predicted temperature value and predicted increment of each zone are derived according to the spatial zone index. When the predicted increment reaches the enhanced trigger condition, the corresponding fan group is selected based on the fixed association between the zone and the fan group and the migration direction to generate the speed adjustment amount. This transforms the heat dissipation control from the passive temperature tracking of the existing scheme to a feedforward intervention for hotspot migration. It can increase the air volume of high-risk zones in advance before local high temperatures form, while avoiding ineffective acceleration of zones in non-migration directions. This improves the efficiency of heat dissipation resource utilization and temperature stability, and solves the problems of lack of dynamic response capability, inability to intervene in local high temperature areas in a timely manner, and impact on performance and user experience. Attached Figure Description

[0019] Figure 1 A flowchart illustrating the steps of the intelligent computer heat dissipation control method based on multi-sensor fusion provided by the present invention; Figure 2 A schematic diagram of a computer heat dissipation intelligent control system based on multi-sensor fusion provided by the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1 Please see Figure 1 As shown, this embodiment provides a computer heat dissipation intelligent control method based on multi-sensor fusion, which mainly includes: S101. Collect the device attitude angle and internal temperature data at each location, and convert the attitude angle into a gravity projection vector; perform directional weighted mapping on the temperature data based on the gravity projection vector to generate an initial heat distribution map and its spatial partition index.

[0022] Collect equipment attitude angles and internal temperature data at various locations to obtain a common input for attitude conditions and temperature fields. Methods for collecting equipment attitude angles include: The device, such as a laptop, has an attitude acquisition device fixedly arranged inside the laptop body, which outputs the device attitude angle. The device attitude angle is used to characterize the tilt state of the computer relative to the geographical direction of gravity, and the device attitude angle is continuously output at a preset sampling period.

[0023] Methods for collecting internal temperature data at various locations include: Temperature acquisition devices are deployed in multiple locations within the computer, covering at least the areas adjacent to heat sources, air duct inlets, air duct outlets, heat sinks, cavity transition areas, and edge cavity areas of the chassis. Each temperature acquisition device outputs the internal temperature data for its corresponding location. To ensure that the device's attitude angle corresponds to the internal temperature data at each location at any given time, both types of data are recorded according to a unified time reference, and a timestamp is used to establish the correspondence between attitude angle and temperature data. The purpose of acquiring the device's attitude angle is to provide input for subsequent gravity direction representation, and the purpose of acquiring the internal temperature data at each location is to provide a distribution basis for subsequent spatial mapping, thereby enabling subsequent processing to simultaneously utilize both attitude and temperature data.

[0024] The device attitude angle is converted into a gravity projection vector, transforming the angle into a vector usable for spatial direction calculations. Based on the acquired device attitude angle, a device coordinate system is established, including a first axis along the long side of the device body, a second axis along the short side of the device body, and a third axis perpendicular to the plane of the device body. The geographic gravity direction is represented as a vector in the device coordinate system, and the vector is transformed according to the device attitude angle to obtain the gravity projection vector. The gravity projection vector includes projection components on the first axis, second axis, and third axis (i.e., the first axis projection component, second axis projection component, and third axis projection component below), and each projection component corresponds one-to-one with the device attitude angle. The purpose of converting the device attitude angle into a gravity projection vector is to enable subsequent direction-weighted mapping to directly reference the projection components of gravity in each axis of the body, thereby introducing the attitude influence into the temperature space expression in a calculable form.

[0025] A weighted mapping is performed on temperature data based on the gravity projection vector to form a weighted temperature representation consistent with the attitude direction. Based on the gravity projection vector and the internal temperature data at each location, a positional correspondence between each temperature acquisition device and the equipment coordinate system is established, resulting in a temperature acquisition location vector. The projection coefficient of the temperature acquisition location vector in the direction of the gravity projection vector is calculated and normalized to obtain the directional weight, ensuring that the directional weight is consistent with the direction of the gravity projection vector and can be directly calculated from the projection coefficient. The internal temperature data at each location is multiplied by the corresponding directional weight to obtain the directional weighted temperature value. This directional weighted temperature value is then mapped to the spatial location in the equipment coordinate system according to the temperature acquisition location vector, forming the directional weighted mapping result. To ensure the mapping result can be continuously expressed in space, for acquisition locations with adjacency relationships, their directional weighted temperature values ​​are weighted and converged according to the adjacency relationship. The convergence weight is obtained by normalizing the inverse of the distance between adjacent acquisition locations. The reason for performing directional weighted mapping is to transform the internal temperature data at each location from only including temperature magnitude to a weighted temperature including the influence of gravity direction projection, thereby providing a temperature input with directional conditions for subsequent mapping.

[0026] Methods for calculating projection coefficients include: The gravity projection vector is represented as a vector in the same device coordinate system. The temperature acquisition position vector is composed of the three-axis coordinates of the temperature acquisition device relative to the origin of the device coordinate system. The gravity projection vector is obtained by converting the device attitude angle and includes the three-axis projection components. This ensures that subsequent projection calculations are performed in the same coordinate system and have repeatable geometric meaning.

[0027] For any temperature acquisition location vector, calculate its dot product with the gravity projection vector, and then calculate the square of the magnitude of the gravity projection vector. Divide the dot product by the square of the magnitude to obtain the projection coefficient. The projection coefficient represents the magnitude of the component of the acquisition location in the direction of the gravity projection vector. To avoid confusion of directional meaning due to the sign, the projection coefficient can be converted to a non-negative quantity according to a preset directional convention. For example, take the positive direction of the third axis of the device coordinate system as the reference. If the projection coefficient is negative, take its opposite number to ensure that the projection coefficient is consistent with the strength of the influence along the gravity projection vector when used for weighting.

[0028] Methods for obtaining directional weights by normalizing the projection coefficient set include: The projection coefficients corresponding to all temperature acquisition locations at the same time are used to form a projection coefficient set. The maximum and minimum values ​​of the projection coefficient set are calculated. For any projection coefficient, linear normalization is performed by subtracting the minimum value from the projection coefficient and then dividing by the difference between the maximum and minimum values ​​to obtain a directional weight between zero and one. This ensures that the directional weights remain comparable under different orientations and arrangement conditions. When the difference between the maximum and minimum values ​​is less than a preset difference threshold, to avoid the denominator being too small and causing abnormal directional weights, the difference between the maximum and minimum values ​​can be replaced with the difference threshold, thereby ensuring that the numerical value of the directional weight calculation process is controllable.

[0029] Methods for obtaining direction-weighted temperature values ​​include: A one-to-one correspondence is established between the directional weights and the internal temperature data at each location, and this correspondence is used for directional weighted mapping. Each directional weight is bound to the internal temperature data at each location of the corresponding temperature acquisition device by the temperature acquisition device number or by the temperature acquisition location vector index. Then, the directional weights are used as multiplication factors to weight the internal temperature data at each location to obtain the directional weighted temperature value. The directional weighted temperature value is used as the input for generating the initial heat distribution map, so that the temperature space representation can be adjusted synchronously with the change of the device's attitude angle.

[0030] An initial heat distribution map and its spatial partition index are generated. The directional weighted mapping results are organized into a spatially referable partition. Based on the directional weighted mapping results, a spatial grid covering the internal cavity and air duct area of ​​the device is constructed in the device coordinate system. The spatial grid is divided into multiple grid cells. For each grid cell, the directional weighted temperature values ​​falling into that grid cell are aggregated. The aggregation method adopts weighted summation and weight normalization so that each grid cell obtains a corresponding grid temperature characterization value, thereby forming the initial heat distribution map. Each spatial partition is assigned a unique partition identifier, and a correspondence table between grid cells and partition identifiers is established. The correspondence table is the spatial partition index. The spatial partition index is used to enable subsequent processing to perform statistics and location on the initial heat distribution map by partition. The initial heat distribution map is used to characterize the spatial heat distribution state inside the computer.

[0031] Furthermore, based on the spatial grid, adjacent grid cells are merged into spatial partitions according to the fuselage structure boundary, air duct connectivity, and heat dissipation component coverage. The purpose of generating an initial heat distribution map and its spatial partition index is to transform discretely collected temperature data into a gridded heat representation under gravity projection vector constraints and provide directly referable partition indexes, thereby providing a unified spatial benchmark for subsequent partition-level temperature statistics and hotspot location.

[0032] S102. Extract the temperature features of the partitions from the initial heat distribution map based on the spatial partition index, and calculate the clustering index by combining the gravity projection vector; determine the coordinates of the candidate hotspot areas based on the clustering index, and output the temperature change statistics corresponding to the coordinates.

[0033] Based on the spatial partition index, the temperature features of each partition are extracted from the initial heat distribution map. Specifically, this includes: reading the grid temperature characterization value of each grid cell in the initial heat distribution map and reading the correspondence between grid cells and partition identifiers in the spatial partition index; grouping grid cells belonging to the same partition into partition grid sets according to the partition identifier; calculating the partition temperature features for each partition grid set, including the partition temperature mean, partition temperature maximum, partition temperature minimum, partition temperature dispersion, partition boundary temperature difference, and partition temperature change rate. The partition temperature dispersion is obtained by taking the square root of the average of the squares of the temperature deviations from the partition temperature mean within the partition grid set; the partition boundary temperature difference is obtained by summing the absolute values ​​of the temperature differences between the partition boundary grid cell and its adjacent partition boundary grid cells and normalizing by the boundary logarithm; when the initial heat distribution map includes data from consecutive sampling times, the partition temperature change rate is further obtained. The partition temperature change rate is obtained by subtracting the partition temperature mean from adjacent sampling times and dividing by the sampling interval. The purpose of extracting the partition temperature features is to transform the temperature expression at the grid cell level into a statistical expression at the partition level, so that the subsequent clustering index calculation has a unified statistical object and can directly reuse the spatial partition index.

[0034] The aggregation index is calculated by combining gravity projection vectors. Specifically, this includes: reading the gravity projection vectors corresponding to the same sampling time of the obtained partition temperature features; for each partition, determining the geometric center coordinates of the partition based on the spatial partition index, and calculating the partition orientation coefficient using the geometric center coordinates and the gravity projection vector. The partition orientation coefficient is the absolute value of the dot product of the partition geometric center position vector and the gravity projection vector, normalized to a value between zero and one; based on this, the maximum partition temperature minus the partition temperature mean is defined as the partition internal aggregation quantity, the partition boundary temperature difference is defined as the partition boundary aggregation quantity, and the absolute value of the partition temperature change rate is defined as the partition temporal aggregation quantity; the partition internal aggregation quantity, partition boundary aggregation quantity, and partition temporal aggregation quantity are multiplied by the partition orientation coefficient, and then weighted and summed according to the weight parameter table to obtain the aggregation index. The weight parameter table is a pre-stored fixed parameter table, and the same set of weight parameters is used for all partitions. The purpose of calculating the aggregation index is to introduce gravity projection vectors to form a partition quantification index related to attitude direction based on the obtained partition temperature features, so that the subsequent candidate hotspot location is constrained by both partition temperature features and attitude direction.

[0035] The weight parameter table is preset and includes at least a first weight, a second weight, and a third weight, which correspond to the aggregation amount within the partition, the aggregation amount at the partition boundary, and the aggregation amount over time in the partition, respectively. The weight parameter table can be indexed by equipment model, radiator layout, and air duct shape.

[0036] The coordinates of candidate hotspot regions are determined based on the aggregation index, specifically including: obtaining the aggregation index of all partitions within the same sampling time and determining the partition with the largest aggregation index as the candidate hotspot partition; within the candidate hotspot partition, selecting the grid cell with the largest temperature characterization value as the candidate hotspot grid cell based on the initial heat distribution map, and determining the grid center coordinates of the candidate hotspot grid cell as the coordinates of the candidate hotspot region.

[0037] Furthermore, when multiple grid cells within a candidate hotspot partition satisfy the condition that the difference between their temperature characterization value and the maximum temperature of the partition does not exceed a threshold, the coordinates of the grid center of the multiple grid cells that meet the condition are weighted and summed according to their temperature characterization values ​​to obtain the coordinates of the candidate hotspot region. The purpose of determining the coordinates of the candidate hotspot region is to further refine the obtained candidate hotspot partition (determined by the aggregation index) to the grid cell location, thereby providing a clear spatial center for the subsequent output of temperature change statistics corresponding to the coordinates.

[0038] The output includes temperature change statistics corresponding to the coordinates, specifically: after obtaining the coordinates of the candidate hotspot region, determining the partition identifier to which the coordinates belong based on the spatial partition index, and identifying the candidate hotspot grid cell corresponding to the coordinates in the initial heat distribution map; determining the neighborhood grid cell set based on the neighborhood radius threshold or the number of adjacent layers threshold, with the candidate hotspot grid cell as the center; reading the temperature characterization value sequence of the candidate hotspot grid cell and the neighborhood grid cell set from the initial heat distribution map at multiple consecutive sampling times, and calculating the temperature change statistics, which include: the temperature increment of the candidate hotspot grid cell, the temperature rise slope of the candidate hotspot grid cell, the average temperature increment of the neighborhood grid cell set, and the maximum temperature increment of the neighborhood grid cell set. The temperature increment is obtained by subtracting the temperature at the start of the time window from the temperature at the current sampling time, and the temperature rise slope is obtained by dividing the temperature difference between adjacent sampling times by the sampling interval; and establishing a correlation between the temperature change statistics and the coordinates of the candidate hotspot region and outputting them. The purpose of outputting temperature change statistics is to bind the coordinates of candidate hotspot regions with the quantified results of temperature changes over time at those coordinates and in their neighborhood as the same output object, so that subsequent steps can directly perform threshold comparisons on the temperature change statistics and use the coordinates as the spatial center for further processing.

[0039] S103. Compare the temperature change statistics with the preset change threshold. If the temperature exceeds the preset change threshold, determine the spatial neighborhood using the coordinates of the candidate hotspot area and extract the historical temperature window within the spatial neighborhood. Calculate the temperature rise slope and neighborhood correlation coefficient based on the historical temperature window, determine the heat accumulation starting point, and mark its corresponding partition.

[0040] The temperature change statistics are compared with a preset change threshold. If the temperature change exceeds the preset change threshold, a spatial neighborhood is determined based on the coordinates of the candidate hotspot area. A historical temperature window is then extracted within this spatial neighborhood. Specifically, this involves: obtaining the temperature change statistics associated with the coordinates of the candidate hotspot area; selecting the maximum temperature increment within the spatial neighborhood as the comparison value; comparing the comparison value with the preset change threshold; when the comparison result is the preset change threshold, expanding the neighborhood grid cell set by the candidate hotspot grid cell corresponding to the coordinates of the candidate hotspot area according to a neighborhood radius threshold or an adjacent layer number threshold to form a spatial neighborhood; and limiting the spatial neighborhood to the grid cells of the partition where the coordinates of the candidate hotspot area are located and the boundary grid cells of its adjacent partitions based on the spatial partition index; subsequently, reading the temperature characterization value sequence of each grid cell within the spatial neighborhood from the historical sampling records; determining the start time using a window length threshold; and extracting the sequence from the start time to the current sampling time as the historical temperature window; using the change threshold as a trigger condition, generating the spatial neighborhood and historical temperature window only when the temperature change reaches the trigger condition, and ensuring that the historical temperature window maintains a consistent spatial range with the coordinates and spatial partition index of the candidate hotspot area.

[0041] The calculation of the temperature rise slope and neighborhood correlation coefficient based on the historical temperature window includes: extracting the temperature sequence of candidate hotspot grid cells corresponding to the coordinates of candidate hotspot regions within the historical temperature window; dividing the temperature difference between adjacent sampling times by the sampling interval to obtain the time-by-time temperature rise slope sequence; and taking the maximum value of the temperature rise slope sequence as the temperature rise slope. Simultaneously, extracting the temperature sequence of each neighborhood grid cell within the spatial neighborhood; using the candidate hotspot grid cell temperature sequence as a reference sequence, calculating the correlation coefficient between each neighborhood grid cell temperature sequence and the reference sequence within the historical temperature window to obtain a neighborhood correlation coefficient set; then, using the reciprocal of the distance from each neighborhood grid cell to the coordinates of the candidate hotspot region as a weight and normalizing it, calculating a weighted average of the neighborhood correlation coefficient set to obtain the neighborhood correlation coefficient. The historical temperature window provides the time change (temperature rise slope) and spatial synchronization (neighborhood correlation coefficient) that can be directly used for threshold comparison, providing quantitative conditions within the same window for subsequent determination of the heat accumulation starting point.

[0042] Methods for obtaining the set of neighborhood correlation coefficients include: Using the candidate hotspot grid cell corresponding to the coordinates of the candidate hotspot region as the reference point, the temperature characterization value sequence of the grid cell at each sampling time is read from the historical temperature window as the reference sequence; the set of neighboring grid cells is determined based on the spatial neighborhood, and the temperature characterization value sequence of each neighboring grid cell at the same sampling time is read one by one as the neighboring grid cell temperature sequence, ensuring that the timestamps of the two types of sequences correspond one-to-one.

[0043] The mean of the reference sequence is calculated and subtracted point by point to obtain the mean-reduced reference sequence. The mean of the temperature sequence of each neighboring grid cell is calculated and subtracted point by point to obtain the mean-reduced neighborhood sequence. The sum of squares of the mean-reduced reference sequence and the mean-reduced neighborhood sequence are calculated and the square root is taken to obtain the scale quantity. The mean-reduced sequence is divided by the corresponding scale quantity to obtain the standardized sequence, so that the subsequent correlation coefficient calculation is not affected by the absolute temperature level.

[0044] For any neighborhood grid cell, its normalized sequence is multiplied point by point with the reference normalized sequence at each sampling time and summed to obtain the sum of products within the window; the sum of products is divided by the number of sampling points within the historical temperature window to obtain the correlation coefficient between the neighborhood grid cell and the reference sequence; the above calculation is repeated for all neighborhood grid cells to obtain the set of neighborhood correlation coefficients, and each correlation coefficient is associated with the corresponding neighborhood grid cell identifier and output.

[0045] Determine the starting point of heat accumulation and label its corresponding partition; specifically, this includes: based on the temperature rise slope sequence, locate the starting time within the historical temperature window. The location method is: identify the first sampling time in the temperature rise slope sequence that satisfies the condition that the temperature rise slope exceeds the slope threshold in the time-by-time comparison as the starting time; at the starting time, calculate the temperature increment of each grid cell in the spatial neighborhood, compare the temperature increment with the increment threshold, and at the same time compare the correlation coefficient corresponding to the grid cell with the correlation threshold. Select the grid cell that simultaneously satisfies the condition that the temperature increment exceeds the threshold and the correlation coefficient exceeds the threshold as the candidate starting point.

[0046] When there is more than one candidate starting point, the grid cells are sorted from largest to smallest according to the temperature increment at the start time. The grid cell with the highest temperature increment is selected as the starting point for heat accumulation. Then, the correspondence between grid cells and partition identifiers in the spatial partition index is read, and the starting point for heat accumulation is mapped to the partition identifier, thereby marking the partition to which the starting point for heat accumulation belongs. With the coordinates of the candidate hotspot region determined and the historical temperature window captured, the starting point for heat accumulation is determined by using slope threshold, increment threshold and related threshold to form continuous conditions. The partition to which the starting point belongs is then output through the spatial partition index, so that subsequent steps can directly use the starting point for heat accumulation and its partition as input.

[0047] S104. Extract the neighborhood temperature sequence within the partition of the heat accumulation starting point, decompose the neighborhood temperature change according to the gravity projection vector, and obtain the heat development trend vector and the hot spot migration direction.

[0048] The process involves extracting the neighborhood temperature sequence around the heat accumulation starting point within its respective partition. Specifically, this includes: obtaining the heat accumulation starting point and its partition identifier, and reading the spatial partition index to obtain the partition grid cell set corresponding to the partition; taking the starting point grid cell corresponding to the heat accumulation starting point as the center, performing adjacency expansion within the partition grid cell set according to the adjacency layer threshold to obtain the starting point neighborhood grid cell set, where the adjacency relationship is determined by the shared boundary of the grid cells.

[0049] Furthermore, if there are boundary grid cells in the set of grid cells in the neighborhood of the starting point (whose adjacent grid cells have different partition identifiers), then the adjacent partition boundary grid cells that share the boundary with the boundary grid cell and have different partition identifiers will be incorporated into the set of grid cells in the neighborhood of the starting point according to the spatial partition index. This will allow the set of grid cells in the neighborhood of the starting point to simultaneously cover the adjacent grid cells on both sides of the boundary of its respective partition. The temperature characterization values ​​of the set of grid cells in the neighborhood of the starting point are read from the historical sampling records at a unified timestamp at the continuous sampling time, forming a neighborhood temperature sequence that corresponds one-to-one with the grid cell identifier. The temperature characterization value sequence of the starting point grid cell is read synchronously as the starting point temperature sequence. Under the premise that the heat accumulation starting point and its corresponding partition have been determined, a time-series temperature data set constrained by the spatial partition index is formed, so that subsequent decomposition processing can be carried out in the same spatial range and under the same timestamp reference, and provide a traceable neighborhood time-series input for outputting the hotspot migration direction.

[0050] The neighborhood temperature change is decomposed using the gravity projection vector to obtain the heat development trend vector and the hotspot migration direction. Specifically, this includes: obtaining the gravity projection vector corresponding to the same sampling time in the neighborhood temperature sequence, and normalizing the gravity projection vector according to its magnitude to obtain the gravity direction unit vector; for each neighborhood grid cell in the set of neighborhood grid cells of the starting point, taking the difference between its grid center coordinates and the grid center coordinates of the starting point to obtain the displacement vector, and calculating the projection of the displacement vector onto the gravity direction unit vector, the projection being the dot product of the displacement vector and the gravity direction unit vector; calculating the lateral displacement vector of the displacement vector in the plane perpendicular to the gravity direction unit vector, the lateral displacement vector being the displacement vector minus the product of the projection and the gravity direction unit vector; in the time dimension, the starting time and the current time are determined based on the neighborhood temperature sequence read in the first step, and the temperature change is calculated for each neighborhood grid cell, the temperature change being the current temperature minus the starting temperature.

[0051] Methods for obtaining the unit vector of gravity direction by normalizing the gravity projection vector according to its magnitude include: Read the gravity projection vector corresponding to the same sampling time of the neighboring temperature sequence. The gravity projection vector is represented as a three-axis component vector in the device coordinate system, including the first axis projection component, the second axis projection component and the third axis projection component, and corresponds one-to-one with the device coordinate system, as the input vector for subsequent normalization.

[0052] The first axis projection component, the second axis projection component, and the third axis projection component are squared and summed, and the square root of the sum is taken to obtain the modulus length. The modulus length is compared with the modulus length threshold to obtain a determination result of whether it is less than the modulus length threshold.

[0053] When the magnitude is not less than the magnitude threshold, the projection components of each axis of the gravity projection vector are divided by the magnitude to obtain the gravity direction unit vector; when the magnitude is less than the magnitude threshold, the gravity direction unit vector is set to the preset unit vector and output.

[0054] The decomposition results are obtained using the following method: The projected values ​​are non-negative to obtain the gravity-following weights, and all gravity-following weights are normalized. The normalized gravity-following weights are used to perform a weighted summation of the temperature changes of each neighboring grid cell to obtain the gravity-following temperature change component. The magnitude of the lateral displacement vector is used as the lateral weight and normalized. The normalized lateral weights are used to perform a weighted summation of the temperature changes of each neighboring grid cell to obtain the lateral temperature change component.

[0055] Furthermore, the lateral displacement vectors of each neighboring grid cell are normalized and weighted according to the non-negativity of their temperature changes to obtain a lateral composite vector. The lateral composite vector is then normalized to obtain the lateral unit direction. The component of temperature change along gravity is multiplied by the unit vector of gravity direction, and the component of temperature change in the lateral direction is multiplied by the lateral unit direction. The two are then added together to obtain the heat development trend vector. The normalized direction of the heat development trend vector is then used as the hotspot migration direction.

[0056] Based on the established neighborhood temperature sequence and the set of neighborhood grid cells of the starting point, a gravity projection vector is introduced as a unified directional reference. The neighborhood temperature change is decomposed into gravity component and lateral component, and further synthesized into a heat development trend vector and hot spot migration direction. This makes the output include both the intensity of change and spatial orientation, which facilitates the use of the hot spot migration direction as a directional constraint for partition-level time extrapolation and fan parameter adjustment in subsequent steps.

[0057] S105. Using the reference temperature of each zone as the base value, the heat development trend vector as the extrapolation term, and the hot spot migration direction as the directional constraint, time extrapolation is performed according to the spatial zone index to obtain the zone temperature prediction value; the difference between the zone temperature prediction value and the reference temperature is used to obtain the prediction increment, and the prediction increment is compared with the enhancement threshold. If the threshold is exceeded, the corresponding fan group is determined according to the hot spot migration direction and the fan speed adjustment amount is generated.

[0058] Using the baseline temperature of each zone as the base value, and with the heat development trend vector as the extrapolation term and the hotspot migration direction as the directional constraint, time extrapolation is performed according to the spatial zone index to obtain the predicted zone temperature value. Specifically, this includes: obtaining the set of grid cells corresponding to each zone based on the heat development trend vector and the hotspot migration direction, and reading the spatial zone index; calculating the mean temperature of each zone from the initial heat distribution map according to the spatial zone index at the current sampling time, and determining the mean temperature of each zone as the baseline temperature of each zone, with each zone's baseline temperature corresponding one-to-one with the zone identifier; and determining the extrapolation time based on a preset time step threshold. The time step is determined, and the number of prediction time steps is determined based on the prediction time step threshold; the magnitude of the heat development trend vector is multiplied by the extrapolation time step to obtain the total extrapolation increment; for each partition, the dot product of the displacement vector from the geometric center of the partition to the heat accumulation starting point and the hot spot migration direction is taken, and the non-negativity is taken to obtain the direction projection coefficient, and the direction projection coefficient of all partitions is normalized to obtain the allocation coefficient; the total extrapolation increment is allocated to each partition according to the allocation coefficient to obtain the extrapolation increment of each partition; the extrapolation increment of each partition is added to the reference temperature of each partition to obtain the predicted temperature value of the partition, and the predicted temperature value of the partition is bound to the partition identifier for output.

[0059] The baseline temperature of each zone is used as the starting point to ensure that the predicted object is consistent with the spatial zone index; the heat development trend vector is used to provide the source of extrapolation increments; and the hot spot migration direction is used to constrain the allocation coefficients so that the extrapolation increments are distributed in the intervals according to the migration direction, thereby obtaining the zone-level zone temperature prediction values.

[0060] The predicted increment is obtained by subtracting the predicted temperature value of each zone from the baseline temperature. The predicted increment is then compared with the enhancement threshold. If the threshold is exceeded, the corresponding fan group is determined based on the hotspot migration direction, and the fan speed adjustment amount is generated. Specifically, for each zone, the predicted increment is obtained by subtracting the predicted temperature value of the zone from the baseline temperature of each zone. The predicted increment is compared with the enhancement threshold to obtain the set of zones exceeding the threshold. The fan group is determined based on the hotspot migration direction. The determination method includes: establishing a fixed association table between zone identifiers and fan group identifiers based on the spatial zone index and the air duct connectivity relationship. The fixed association table is used to characterize which fan group forms the covering airflow for each zone.

[0061] When the set of overthreshold partitions includes multiple partitions, calculate the dot product of the displacement vector of each overthreshold partition and the hotspot migration direction, and select the fan group corresponding to the overthreshold partition with the largest dot product as the corresponding fan group.

[0062] Furthermore, when generating the fan speed adjustment amount, the difference between the predicted increment and the enhancement threshold is calculated. The difference is then segmented and mapped to the speed increment threshold table to output the fan speed adjustment amount. When multiple over-threshold partitions are mapped to the same fan group, the fan speed adjustment amount corresponding to the partition with the largest difference is taken as the fan speed adjustment amount for that fan group. The purpose is to determine the partition range that needs to be adjusted by comparing the predicted increment with the enhancement threshold, and then determine the corresponding fan group by the hotspot migration direction, so that the adjustment object is consistent with the migration direction, and output the executable fan speed adjustment amount through the correspondence between the difference and the speed increment threshold table.

[0063] Further, according to the spatial partition index, the temperature characterization value of each grid cell is collected at continuous sampling times after the fan speed adjustment is executed, and the average temperature of each partition is calculated to form a feedback temperature sequence corresponding one-to-one with the partition identifier. The feedback temperature sequence and the partition temperature prediction value use the same timestamp. For each partition under the same timestamp, the partition temperature value in the feedback temperature sequence and the partition temperature prediction value are read, and the difference is obtained to obtain the partition deviation. The absolute value of the partition deviation is taken to obtain the partition deviation amount, which is bound to the partition identifier for output. For each partition, the partition deviation amount is compared with the deviation threshold to obtain the set of over-threshold partitions. When the set of over-threshold partitions is not empty, the starting point update step is entered. When the set of over-threshold partitions is empty, the current heat accumulation starting point is kept unchanged. Based on the set of over-threshold partitions, the grid cell with the largest temperature increment in the initial heat distribution map is selected as the new heat accumulation starting point, and its partition is marked according to the spatial partition index. The updated heat accumulation starting point is output, and S104 is returned to extract the neighborhood temperature sequence around the new starting point.

[0064] Example 2 See Figure 2 As shown, this embodiment discloses a computer heat dissipation intelligent control system based on multi-sensor fusion. Details not described in detail are given in Embodiment 1. The system includes: The mapping module is used to collect the device's attitude angle and internal temperature data at various locations, convert the attitude angle into a gravity projection vector, and perform directional weighted mapping on the temperature data based on the gravity projection vector to generate an initial heat distribution map and its spatial partition index. The positioning module extracts the temperature features of the zones from the initial heat distribution map based on the spatial zone index, and calculates the clustering index by combining the gravity projection vector; it determines the coordinates of candidate hotspot areas based on the clustering index, and outputs the temperature change statistics corresponding to the coordinates of the candidate hotspot areas; The annotation module is used to compare the temperature change statistics with the preset change threshold. When the temperature exceeds the preset change threshold, the spatial neighborhood is determined by the coordinates of the candidate hotspot area, and a historical temperature window is extracted within the spatial neighborhood. Based on the historical temperature window, the temperature rise slope and the neighborhood correlation coefficient are calculated to determine the heat accumulation starting point and annotate its corresponding partition. The direction generation module is used to extract the neighborhood temperature sequence within its respective partition around the heat accumulation starting point, decompose the neighborhood temperature change according to the gravity projection vector, and obtain the heat development trend vector and the hot spot migration direction. The speed adjustment module is used to perform time extrapolation based on the spatial partition index, using the reference temperature of each partition as the base value, the heat development trend vector as the extrapolation term, and the hot spot migration direction as the directional constraint, to obtain the partition temperature prediction value; the difference between the partition temperature prediction value and the reference temperature is used to obtain the prediction increment, and the prediction increment is compared with the preset enhancement threshold. If it exceeds the preset enhancement threshold, the corresponding fan group is determined according to the hot spot migration direction and the fan speed adjustment amount is generated.

[0065] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of protection of the present invention.

[0066] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A computer heat dissipation intelligent control method based on multi-sensor fusion, characterized in that, include: S101. Collect the device attitude angle and internal temperature data at each location, and convert the attitude angle into a gravity projection vector. Based on the gravity projection vector, a directional weighted mapping is performed on the temperature data to generate an initial heat distribution map and its spatial partition index; S102. Extract the temperature features of the zones from the initial heat distribution map based on the spatial zone index, and calculate the clustering index by combining the gravity projection vector; determine the coordinates of the candidate hotspot areas based on the clustering index, and output the temperature change statistics corresponding to the coordinates of the candidate hotspot areas; S103. Compare the temperature change statistics with the preset change threshold. If the temperature exceeds the preset change threshold, determine the spatial neighborhood using the coordinates of the candidate hotspot area and extract the historical temperature window within the spatial neighborhood. Calculate the temperature rise slope and neighborhood correlation coefficient based on the historical temperature window to determine the heat accumulation starting point and mark its corresponding partition. S104. Extract the neighborhood temperature sequence within the partition of the heat accumulation starting point, decompose the neighborhood temperature change according to the gravity projection vector, and obtain the heat development trend vector and the hot spot migration direction. S105. Using the baseline temperature of each zone as the base value, the heat development trend vector as the extrapolation term, and the hot spot migration direction as the directional constraint, time extrapolation is performed according to the spatial zone index to obtain the zone temperature prediction value. The predicted increment is obtained by subtracting the predicted temperature value of the zone from the baseline temperature. The predicted increment is then compared with the preset enhancement threshold. If the predicted increment exceeds the preset enhancement threshold, the corresponding fan group is determined based on the hot spot migration direction, and the fan speed adjustment amount is generated.

2. The intelligent control method for computer heat dissipation based on multi-sensor fusion according to claim 1, characterized in that, Methods for generating initial heat distribution maps and establishing spatial partitioning indexes include: Collect equipment attitude angle and internal temperature data at various locations, record them according to a unified time base and establish a corresponding relationship with timestamps; Establish a device coordinate system based on the device attitude angle and transform the geographic gravity vector to obtain the gravity projection vector; A temperature acquisition location vector is established based on the gravity projection vector, and the projection coefficient is calculated and normalized to obtain the direction weight. Temperature data is weighted by direction and mapped to the equipment coordinate system to form a direction-weighted mapping result; The mapping results are weighted and summed, and the weights are normalized to generate an initial heat distribution map and establish a spatial partition index.

3. The intelligent control method for computer heat dissipation based on multi-sensor fusion according to claim 1, characterized in that, Methods for calculating the clustering index include: Based on the spatial partition index, the grid temperature characterization values ​​of the initial heat distribution map are read and the partition grid sets are collected according to the partition identifier to calculate the partition temperature characteristics; Based on the temperature characteristics of the partition, the gravity projection vector at the same sampling time is read to determine the geometric center coordinates of the partition and calculate the partition orientation coefficient; Based on the temperature characteristics of the partition, the aggregation amount within the partition, the aggregation amount at the partition boundary, and the aggregation amount over time are determined. The aggregation amount within the partition, the aggregation amount at the partition boundary, and the aggregation amount over time are multiplied by the partition direction coefficient, and then summed according to the weight parameter table to obtain the aggregation index. The weight parameter table is preset and includes a first weight, a second weight, and a third weight, which correspond to the aggregation amount within the partition, the aggregation amount at the partition boundary, and the aggregation amount over time, respectively.

4. The intelligent control method for computer heat dissipation based on multi-sensor fusion according to claim 3, characterized in that, Methods for outputting statistical values ​​of temperature changes corresponding to coordinates include: At the same sampling time, the clustering index of all partitions is obtained and the partition with the largest clustering index is determined as the candidate hotspot partition; Within the candidate hotspot partition, candidate hotspot grid cells are determined based on the initial heat distribution map, and the grid center coordinates of the candidate hotspot grid cells are determined as the coordinates of the candidate hotspot region. Under the difference threshold constraint, the coordinates of multiple grid center coordinates are weighted and summed according to the temperature characterization value to obtain the coordinates of the candidate hotspot region. Based on the coordinates and spatial partition index of the candidate hotspot region, the partition identifier and candidate hotspot grid cell of the candidate hotspot region are determined. The set of neighboring grid cells is determined based on the neighborhood radius threshold or the number of adjacent layers threshold. The temperature characterization value sequence is read from the initial heat distribution map at multiple consecutive sampling times and the temperature change statistics are output.

5. The intelligent control method for computer heat dissipation based on multi-sensor fusion according to claim 1, characterized in that, Methods for extracting historical temperature windows within a spatial neighborhood include: Obtain the temperature change statistics associated with the coordinates of the candidate hotspot area, select the maximum temperature increment in the spatial neighborhood as the comparison value and compare it with the preset change threshold; When the comparison result exceeds the change threshold, the candidate hot spot grid cell corresponding to the coordinate is taken as the center, and the set of neighboring grid cells is expanded according to the neighborhood radius threshold or the number of adjacent layers threshold to form a spatial neighborhood. The spatial neighborhood is then limited to the grid cells of the partition where the coordinate is located and the grid cells of the boundary of the adjacent partition according to the spatial partition index. The temperature characterization value sequence is read from historical sampling records based on spatial neighborhood, and the starting time is determined by the window length threshold. The sequence from the starting time to the current sampling time is then extracted to obtain the historical temperature window.

6. The intelligent control method for computer heat dissipation based on multi-sensor fusion according to claim 5, characterized in that, Methods for determining the starting point of heat accumulation and labeling its corresponding zone include: Based on the historical temperature window, the temperature sequence of candidate hot spot grid cells corresponding to the coordinates of candidate hot spot regions is extracted. The temperature rise slope sequence is calculated step by step according to the temperature difference between adjacent sampling times and the sampling interval. The maximum value of the temperature rise slope sequence is taken to determine the temperature rise slope. Based on the historical temperature window and spatial neighborhood, the temperature sequence of each neighborhood grid unit is extracted. The temperature sequence of the candidate hot spot grid unit is used as the reference sequence. The reference sequence and each neighborhood temperature sequence are subjected to mean removal and scale normalization to obtain a standardized sequence. The correlation coefficient is calculated by summing the point-by-point products and dividing by the number of sampling points to obtain the neighborhood correlation coefficient set. The neighborhood correlation coefficient is obtained by weighted average normalization by the inverse distance. Based on the temperature rise slope sequence, the first slope threshold exceeding the threshold sampling time is determined within the historical temperature window as the starting time. Based on the starting time, the temperature increment is calculated in the spatial neighborhood and compared with the increment threshold. The candidate starting point is determined by combining the correlation coefficient and the correlation threshold. The first grid cell in the sorted order is selected according to the temperature increment to determine the heat accumulation starting point. The partition to which the heat accumulation starting point belongs is marked based on the spatial partition index.

7. The intelligent control method for computer heat dissipation based on multi-sensor fusion according to claim 6, characterized in that, Methods for extracting neighborhood temperature sequences include: Obtain the heat accumulation starting point and its corresponding partition identifier, and read the spatial partition index to obtain the set of partition grid cells corresponding to the partition; Based on the partitioned grid cell set, with the starting grid cell as the center, the adjacency expansion is performed according to the adjacency layer threshold and the adjacency relationship of the shared boundary of the grid cells to obtain the starting point neighborhood grid cell set; The boundary grid cells are determined based on the set of grid cells in the neighborhood of the starting point. The adjacent boundary grid cells that share the boundary with the boundary grid cells but have different partition identifiers are merged into the spatial partition index to obtain the extended set of grid cells in the neighborhood of the starting point. The temperature sequence is formed by reading the temperature characterization values ​​of the neighboring grid cells at the unified timestamp of the extended starting point. The temperature characterization value sequence of the starting point grid cells is then read to form the starting point temperature sequence.

8. The intelligent control method for computer heat dissipation based on multi-sensor fusion according to claim 7, characterized in that, Methods for determining the direction of hotspot migration include: Based on the gravity projection vector corresponding to the same sampling time of the neighborhood temperature sequence, the gravity projection vector is represented as a three-axis component vector in the device coordinate system, including the first axis projection component, the second axis projection component, and the third axis projection component. The magnitude is calculated based on the first axis projection component, the second axis projection component, and the third axis projection component and compared with the magnitude threshold. When the magnitude is not less than the magnitude threshold, the projection component of each axis is divided by the magnitude to obtain the gravity direction unit vector. When the magnitude is less than the magnitude threshold, the third axis positive direction unit vector is determined as the gravity direction unit vector. The displacement vector of each neighboring grid cell is calculated based on the set of neighboring grid cells of the starting point and the unit vector of gravity direction. The projection is obtained by calculating the dot product of the displacement vector and the unit vector of gravity direction. The lateral displacement vector is obtained by subtracting the product of the projection and the unit vector of gravity direction from the displacement vector. The starting time and the current time are determined based on the neighborhood temperature sequence, and the temperature change of each neighborhood grid cell is calculated. The gravity weight is obtained by taking the projection as non-negative and normalizing it, the lateral weight is obtained by normalizing the magnitude of the lateral displacement vector, and the gravity temperature change component and the lateral temperature change component are obtained by weighting and summing the temperature change based on the gravity weight and the lateral weight respectively. The lateral composite vector is obtained by normalizing the lateral displacement vector with the non-negative values ​​of temperature change and then normalizing it to obtain the lateral unit direction. The heat development trend vector is obtained by multiplying the temperature change component under gravity with the unit vector of gravity and adding the result of multiplying the lateral temperature change component with the lateral unit direction. The direction of hotspot migration is determined by normalizing the direction of the heat development trend vector.

9. The intelligent control method for computer heat dissipation based on multi-sensor fusion according to claim 8, characterized in that, Methods for generating fan speed adjustment include: The initial heat distribution map is read according to the spatial partition index and the average temperature of each partition is calculated. The average temperature of each partition is determined as the baseline temperature of each partition. Based on the reference temperature of each zone, the extrapolation time step is determined according to the preset time step threshold. The total extrapolation increment is obtained based on the magnitude of the heat development trend vector and the extrapolation time step. The distribution coefficient is obtained by taking the non-negative and normalizing the dot product of the displacement vector from the geometric center of the zone to the heat accumulation starting point and the hot spot migration direction. The total extrapolation increment is distributed and added to the reference temperature of each zone to obtain the zone temperature prediction value. The predicted increment is obtained by subtracting the predicted temperature value of each zone from the baseline temperature of each zone, and compared with the enhancement threshold to obtain the set of zones exceeding the threshold. The fan group corresponding to the zone exceeding the threshold is determined based on the fixed association table of zone identifier and fan group identifier. When the set of zones exceeding the threshold includes multiple zones, the fan group corresponding to the zone with the largest dot product is selected. Based on the difference between the predicted increment and the enhancement threshold, the fan speed adjustment amount is output segment by segment according to the speed increment threshold table.

10. A computer heat dissipation intelligent control system based on multi-sensor fusion, characterized in that, The system is used to execute the intelligent computer heat dissipation control method based on multi-sensor fusion as described in any one of claims 1-8, and comprises: The mapping module is used to collect the device's attitude angle and internal temperature data at various locations, convert the attitude angle into a gravity projection vector, and perform directional weighted mapping on the temperature data based on the gravity projection vector to generate an initial heat distribution map and its spatial partition index. The positioning module extracts the temperature features of the zones from the initial heat distribution map based on the spatial zone index, and calculates the clustering index by combining the gravity projection vector; it determines the coordinates of candidate hotspot areas based on the clustering index, and outputs the temperature change statistics corresponding to the coordinates of the candidate hotspot areas; The annotation module is used to compare the temperature change statistics with the preset change threshold. When the temperature exceeds the preset change threshold, the spatial neighborhood is determined by the coordinates of the candidate hotspot area, and a historical temperature window is extracted within the spatial neighborhood. Based on the historical temperature window, the temperature rise slope and the neighborhood correlation coefficient are calculated to determine the heat accumulation starting point and annotate its corresponding partition. The direction generation module is used to extract the neighborhood temperature sequence within its respective partition around the heat accumulation starting point, decompose the neighborhood temperature change according to the gravity projection vector, and obtain the heat development trend vector and the hot spot migration direction. The speed adjustment module is used to perform time extrapolation based on the spatial partition index, using the reference temperature of each partition as the base value, the heat development trend vector as the extrapolation term, and the hot spot migration direction as the directional constraint, to obtain the partition temperature prediction value; the difference between the partition temperature prediction value and the reference temperature is used to obtain the prediction increment, and the prediction increment is compared with the preset enhancement threshold. If it exceeds the preset enhancement threshold, the corresponding fan group is determined according to the hot spot migration direction and the fan speed adjustment amount is generated.

Images