Single-phase immersion liquid cooling temperature stratification evaluation and perturbation liquid injection equalization control system

Abstract

The application discloses a single-phase immersion liquid cooling temperature stratification evaluation and disturbance liquid injection equalization control system and relates to the technical field of liquid cooling circulation structures.The single-phase immersion liquid cooling temperature stratification evaluation and disturbance liquid injection equalization control system comprises a reflux liquid multi-point liquid inlet dynamic monitoring module, which collects liquid cooling branch state data; an immersion liquid temperature stratification monitoring module, which evaluates the thermal distribution unevenness degree and adjusts a disturbance liquid injection strategy; a liquid flow distribution equalization evaluation module, which analyzes the liquid flow distribution and heat dissipation matching degree and optimizes a flow direction coverage strategy; a cooling path dynamic determination and switching module, which evaluates the cooling state and thermal load change and regulates and controls a circulation path; and an immersion liquid cooling cabinet local disturbance self-balancing regulation and control module, which guides local convection and adjusts a disturbance direction.The application solves the problem that, after the reflux of the cooling liquid branch, the cold and hot layers are seriously stacked, local heat accumulation in the immersion liquid cooling cabinet is caused, and the chip heat dissipation efficiency is affected.

Description

Technical Field

[0001] This invention relates to the field of liquid cooling circulation structure technology, specifically to a single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system. Background Technology

[0002] With the expansion of high-density server deployments and the continuous improvement of data center energy efficiency standards, the role of circulating cooling structures in immersion liquid cooling systems is becoming increasingly prominent. Currently, single-phase immersion liquid cooling technology is widely used in the thermal management of high-power devices. It mainly achieves orderly circulation of liquid between the heating and heat dissipation zones by constructing a closed liquid pool, setting a main cooling loop, and configuring a liquid drive device, thereby ensuring that the heat from the internal components of the server is promptly removed and released to the external heat exchange unit.

[0003] For example, the invention patent with publication number CN114423264B discloses a single-phase immersion liquid cooling system and liquid cooling method. The liquid cooling system includes a main supply pipe and a main return pipe, which form a circulation loop. Multiple liquid cooling cabinets are connected in parallel to the circulation loop through sub-pipes for supply and return. In this embodiment, a flow regulation unit is provided on the sub-pipe for supply. The flow regulation unit includes a differential pressure detection device and a flow regulation device, which are respectively connected to the control unit. The differential pressure detection device is used to detect the differential pressure of the coolant flowing through it, and the flow regulation device is used to regulate the flow rate of the coolant flowing through it. Thus, for a certain branch of the supply sub-pipe, the control unit can adjust the flow rate of the coolant flowing into the supply sub-pipe according to the differential pressure of the supply sub-pipe. In this way, not only is the flow rate of the coolant on each parallel branch in the circulation loop adjusted as needed, but the adjustment accuracy is also high, avoiding the imbalance of coolant flow.

[0004] For example, invention patent CN109952003B discloses a data center liquid cooling system, relating to the field of data center heat dissipation device technology. It includes a cabinet, servers, cold plates, a cooling device, a liquid storage tank, and a refrigerant pump. The cabinet contains a single-phase refrigerant, and the servers are immersed in the single-phase refrigerant and connected to the cabinet. The cold plates are located inside the cabinet and connected to it. The cold plates are sequentially connected to the cooling device, the liquid storage tank, and the refrigerant pump to form a circulation loop. This invention combines indirect contact liquid cooling, single-phase direct immersion liquid cooling, and natural cooling. The single-phase refrigerant does not need to circulate under the drive of a pump; instead, it circulates based on changes in temperature and density. While the refrigerant circulates under the drive of a refrigerant pump, its viscosity is lower than that of the single-phase refrigerant, making it easier to flow and reducing pump losses. This reduces pump losses caused by single-phase refrigerants, lowers the data center's PUE, and improves the data center's greenness.

[0005] However, existing single-phase immersion liquid cooling systems generally suffer from the following technical bottlenecks: their cooling path structures mostly employ static main loop configurations, lacking real-time response mechanisms to fluid temperature gradients, flow velocity distributions, and localized heat accumulation, making it difficult to cope with fluctuations in cooling demands caused by dynamic changes in server heat load. During high-load operation, localized dead zones or thermal stratification occur in the liquid flow, leading to decreased heat exchange efficiency and severe heat retention in some areas. Furthermore, existing systems mostly circulate coolant at fixed frequencies or constant flow rates, failing to implement differentiated adjustments for heat dissipation pressure in different areas, thus limiting further improvements in overall energy efficiency and operational stability.

[0006] To address the aforementioned issues, there is an urgent need for a single-phase immersion liquid cooling temperature stratification assessment and disturbance-induced liquid injection equalization control system. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a single-phase immersion liquid cooling temperature stratification assessment and disturbance-based liquid injection equalization control system, which solves the problem of severe hot and cold layering after coolant branch return, leading to localized heat accumulation in the immersion liquid cooling cabinet and affecting chip heat dissipation efficiency.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system, comprising a multi-point dynamic monitoring module for reflux liquid, used to collect coolant state data, flow state data, and path thermal state data in the current liquid cooling branch, and preprocessing the collected coolant state data, flow state data, and path thermal state data to construct a standardized liquid cooling operation characteristic dataset; an immersion liquid temperature stratification monitoring module, used to assess the degree of uneven heat distribution of the coolant in the vertical direction based on the standardized liquid cooling operation characteristic dataset, and dynamically adjust the local disturbance injection strategy based on the assessment results; and a flow distribution equalization assessment module, used to assess the uniformity of the liquid cooling operation based on the standardized liquid cooling operation... The feature dataset analyzes the matching degree between the current liquid flow distribution and heat dissipation, and drives the optimization of the flow direction coverage strategy based on the analysis results. The cooling path dynamic judgment and switching module is used to comprehensively evaluate the current cooling state and heat load changes based on the evaluation results of the uneven heat distribution and the analysis results of the matching degree between liquid flow distribution and heat dissipation, and intelligently adjusts the circulation path and heat exchange intervention mode based on the evaluation results. The immersion liquid cooling cabinet local disturbance self-balancing control module is used to judge the evolution trend of the thermal layer structure based on the temperature profile and liquid flow distribution state, guide local convection through bottom disturbance liquid injection, and dynamically adjust the disturbance direction and intensity in combination with temperature rise feedback and distribution response results. At the same time, it is combined with flow direction relaxation control to improve the local diffusion effect of the disturbed liquid.

[0009] Further, the specific steps for collecting coolant state data, liquid flow state data, and path thermal state data in the current liquid cooling branch are as follows: Key parameters in the liquid cooling branch are collected in a structured manner; coolant state data is collected, including: coolant temperature of each layer in the submerged liquid cooling cabinet, liquid level height of the temperature sensor, and coolant temperature at the top and bottom of the submerged liquid cooling cabinet. Simultaneously, the number of sensor layers, temperature difference, and height difference between adjacent layers are recorded within the current collection period. The top and bottom coolant temperatures of the submerged liquid cooling cabinet represent the high-temperature limit and low-temperature benchmark in the liquid thermal distribution within the current period, forming a representative temperature difference reference dataset. Within multiple consecutive collection periods, the top and bottom coolant temperatures are processed into a time series, and several effective data points are extracted based on the processed time series. The stable range of temperature difference change within the cycle is used as the average value of the stable range and is recorded as the normalized baseline value of temperature difference under the current operating stage. Liquid flow status data is collected, including instantaneous flow velocity values, continuous flow velocity change data, and injection response time for each region. The average liquid flow velocity in the branch return path is also recorded. A flow velocity fluctuation time series is constructed by using the continuous flow velocity change data per unit time in each chip region within the immersion liquid cooling cabinet. The uniformity of coolant distribution within the region is obtained by comprehensively evaluating the degree of fluctuation of liquid in the region within a unit time and the consistency of the disturbance response, combined with the injection response time of the disturbance injector. Path thermal status data is collected, including the effective heat dissipation area of ​​the chip in each region, the coolant temperature in the return branch, the coolant temperature at the outlet of the immersion liquid cooling cabinet, and the chip temperature rise rate. The total number of effective monitoring areas within the current cycle is also recorded.

[0010] Furthermore, the specific steps for preprocessing the collected coolant state data, fluid flow state data, and path thermal state data to construct a standardized liquid cooling operation feature dataset are as follows: Preprocessing the collected coolant state data, fluid flow state data, and path thermal state data to complete coordinate standardization, including the correspondence between liquid level height and temperature sensors, and eliminating temperature sequence abrupt changes and inter-layer measurement point misalignments caused by sensor drift and short-term communication interruptions; Based on the distribution characteristics of coolant temperature in each layer, and combined with the actual pipe layout structure of the immersion liquid cooling cabinet and coolant flow direction constraints, performing differential repair on temperature gradient reversal and abnormal static intervals to restore continuous stability. A defined temperature profile is established. Before preprocessing, the path thermal state data is classified and mapped based on the mapping relationship between the chip region topology and the injection point. For chip regions with multiple overlapping cooling coverage, heat dissipation area splitting is performed. To address data segment conflicts caused by the parallel operation of multiple injection units during liquid cooling operation, a data window synchronization mechanism is introduced in the preprocessing stage. This mechanism prioritizes retaining data from regions with continuously increasing flow fluctuations and removes duplicate response fields, improving data consistency under multi-region joint processing. All preprocessed coolant state data, fluid flow state data, and path thermal state data are normalized to construct a standardized liquid cooling operation feature dataset.

[0011] Further, the specific steps for evaluating the unevenness of coolant heat distribution in the vertical direction based on a standardized liquid cooling operation characteristic dataset are as follows: Real-time measurement data from temperature sensors on each layer inside the immersion liquid cooling cabinet are collected to construct a vertical temperature profile sequence within the current acquisition period. The temperature gradient at the current location is represented by the ratio of the temperature difference between each two adjacent layers to the height difference. The temperature gradient with the largest absolute value among all adjacent measurement points is extracted as the local maximum gradient of the profile. Simultaneously, by calculating the interquartile range of all temperature gradients, the upper limit of the acceptable gradient fluctuation range is obtained and recorded as the profile fluctuation reference value. The calculation is then performed sequentially for each adjacent layer... The interlayer temperature gradient is obtained by dividing the temperature difference between the layers by the difference in liquid level height between the temperature sensors of adjacent layers, and taking the absolute value of the ratio of the temperature difference between the layers to the difference in liquid level height between the temperature sensors. The average interlayer temperature value is obtained by adding the interlayer temperature gradient values ​​of each layer in the sensor and dividing by the number of sensor layers. The normalized ratio of the upper and lower temperature differences is obtained by subtracting the temperature of the bottom coolant from the temperature of the top coolant of the immersion liquid-cooled cabinet and then dividing by the normalized temperature difference benchmark value. The gradient offset ratio is obtained by dividing the maximum local gradient of the profile by the profile fluctuation benchmark value. The temperature layer difference risk value is obtained by multiplying the average interlayer temperature value, the normalized ratio of the upper and lower temperature differences, and the gradient offset ratio.

[0012] Furthermore, the specific steps of dynamically adjusting the local disturbance injection strategy based on the evaluation results are as follows: real-time comparison of the current temperature difference risk value with the temperature difference risk threshold: when the temperature difference risk value is less than or equal to the temperature difference risk threshold, maintain the current branch return state unchanged, maintain the existing liquid inlet configuration and flow direction, continue to perform multi-point injection, monitor the change range of temperature acquisition values ​​of each layer, and update the data acquisition and calculation of temperature measurement points of each layer at a fixed period; when the temperature difference risk value is greater than the temperature difference risk threshold, immediately close the branch pipeline valve, switch the main circulation cooling path, simultaneously pause the current liquid inlet injection operation, activate the bottom micro-disturbance injector to perform periodic low-speed disturbance injection, and reconstruct the internal injection path mapping relationship of the immersion liquid cooling cabinet.

[0013] Further, the specific steps for analyzing the matching degree of current liquid flow distribution and heat dissipation based on the standardized liquid cooling operation characteristic dataset are as follows: A dynamic index set of coolant distribution uniformity is constructed by calculating the ratio of continuous flow velocity change data to liquid injection response time in each effective monitoring area within the current cycle; subsequently, in the dynamic index set of coolant distribution uniformity, a normalized reference value for the distribution consistency level within the liquid cooling branch in the current cycle is obtained using the median extraction method, and this value is denoted as the coolant distribution uniformity reference value; the instantaneous flow velocity value, the effective heat dissipation area of ​​the chip, and the corresponding coolant density correction factor of the current area are multiplied to obtain the regional liquid kinetic energy value of the current area; the ratio of the coolant distribution uniformity of each area to the coolant distribution uniformity reference value is subtracted from this value and then multiplied by the distribution deviation weighting factor to obtain the distribution uniformity adjustment value; the regional liquid kinetic energy value is divided by the corresponding distribution uniformity adjustment value to obtain the regional balanced effective liquid flow value; the regional balanced effective liquid flow values ​​of each effective monitoring area within the current cycle are accumulated to obtain the liquid flow uniformity distribution evaluation value for the current cycle.

[0014] Furthermore, the specific steps for optimizing the flow direction coverage strategy based on the analysis results are as follows: Based on the calculation results of the liquid flow equilibrium distribution evaluation value, the flow direction control strategy is dynamically adjusted according to the fluctuation trend of the real-time liquid flow equilibrium distribution evaluation value: When the liquid flow equilibrium distribution evaluation value shows an upward trend in three or more consecutive cycles, and the increase does not exceed the fluctuation amplitude threshold, it indicates that the coolant flow structure tends to be stable and has a high degree of matching with the heat load. The current flow distribution structure is frozen and path reconstruction is paused, entering the observation enhancement stage: the sampling density of the micro-disturbance area around the liquid inlet point is expanded, and micro-low flow velocity disturbance injection tests are performed periodically to verify whether the current steady state is a true steady state and avoid local false steady state failure; when the liquid flow equilibrium distribution evaluation value increases beyond the fluctuation amplitude threshold in two cycles, it is determined to be liquid flow equilibrium distribution. When the flow coverage structure suddenly becomes unbalanced, all current injection points are interrupted, the main circulation path is reset, the coolant self-balancing startup mechanism is activated, and gradient recovery is performed based on the average flow velocity of each injection point. At the same time, a delayed pressure boosting strategy is used to avoid sudden pressure impact. During the recovery process, the spatial distribution and convergence speed of the chip temperature difference are monitored in real time. When turbulent areas are detected, the injection angle and flow path are forcibly rearranged. When the liquid flow balance distribution evaluation value shows a downward trend for five consecutive cycles, and the chip temperature decreases synchronously, it is determined that the current coolant layout has reached a relatively optimal balance structure. The frequency of disturbance nozzles is reduced, the holding time of the main pipeline valve is extended, and a micro-pressure difference weak disturbance mechanism is introduced to maintain the boundary heat transfer activity. At the same time, a flexible flush is automatically performed every ten cycles to achieve flow redistribution through short-time large-volume injection to prevent the accumulation of thermal inertia.

[0015] Furthermore, the specific steps for comprehensively evaluating the current cooling state and heat load changes using the assessment results of uneven heat distribution and the analysis results of the matching degree between liquid flow distribution and heat dissipation as input are as follows: obtain the temperature difference risk value and the liquid flow equilibrium distribution assessment value; calculate the absolute value of the difference between the coolant temperature in the current return branch and the coolant temperature at the outlet of the immersion liquid cooling cabinet; multiply the temperature difference risk value, the liquid flow equilibrium distribution assessment value, and the absolute value of the difference between the coolant temperature in the return branch and the coolant temperature at the outlet of the immersion liquid cooling cabinet to obtain the temperature difference enhancement load term; add the chip temperature rise rate to the reciprocal of the average liquid flow velocity in the current branch return path and then add one to obtain the liquid flow attenuation penalty value; divide the temperature difference enhancement load term by the liquid flow attenuation penalty value and then take the logarithm to obtain the path switching determination value at the current moment.

[0016] Furthermore, the specific steps for intelligently controlling the circulation path and heat exchange intervention method based on the evaluation results are as follows: Real-time comparison of the current path switching judgment value with the path switching threshold, where the path switching threshold includes a first switching threshold and a second switching threshold: When the path switching judgment value is less than or equal to the second switching threshold, maintain the current branch return path operation status unchanged, maintain the flow distribution structure and guiding direction settings at each liquid inlet point, pause path switching execution, and only periodically sample the difference between the return liquid temperature and the chip temperature, while recording the diffusion trend of the liquid on the cooling surface, and then execute the judgment again in the next cycle; When the path switching judgment value is greater than the second switching threshold and less than or equal to the first switching threshold, reduce the opening of the branch pipeline valve and gradually increase it. Maintaining pressure in the main circulation channel, the two cooling paths operate in parallel for short periods. This ensures stable chip heat exchange efficiency while establishing a flow field buffer area and initiating a dynamic path transfer recording mechanism. Based on the trend of the path switching judgment value in the next cycle, it determines whether to fully switch to the main circulation channel. When the path switching judgment value exceeds the first switching threshold, a path switching command is immediately executed, closing the branch return channel and switching to the main circulation path for forced cooling. Simultaneously, the return liquid direction and inlet point weight configuration are reset, all disturbance injection operations are suspended, and the main channel liquid velocity balance constraint logic is activated. This ensures high-flow-rate cooling while limiting concentrated injection into overloaded areas, improving heat exchange efficiency and preventing localized temperature rise at the chip edge due to high-flow-rate impact.

[0017] Furthermore, the specific steps for judging the evolution trend of the thermal layer structure based on temperature profile and liquid flow distribution, guiding local convection through bottom disturbance injection, dynamically adjusting the disturbance direction and intensity in conjunction with temperature rise feedback and distribution response results, and simultaneously improving the local diffusion effect of the disturbed liquid with flow relaxation control are as follows: During the period when the branch return path is open, the internal temperature profile data of the immersion liquid cooling cabinet collected by the vertical temperature sensor array is continuously monitored, and analyzed together with the liquid flow equilibrium distribution evaluation value and path switching judgment value to determine whether there is a trend of thermal layer structure stabilization. When it is detected that the top liquid temperature of the immersion liquid cooling cabinet is higher than that of the bottom for three consecutive cycles, and the multi-layer temperature difference changes converge, the micro-sprayers deployed at the bottom of the immersion liquid cooling cabinet are automatically activated. Without altering the structure of the main cooling path and return path, the flow meter and lateral directional channel intermittently inject disturbed coolant at a minimum flow rate, inducing slow local liquid convection, causing the coolant to rise and the hot liquid to move laterally. After each round of disturbance injection, the temperature rise rate of the chip hot zone and the temperature difference trend of the multi-layer immersion liquid-cooled cabinet are analyzed. When the temperature difference converges significantly and the liquid flow distribution index shows an upward trend, the current disturbance frequency and direction are maintained. When the liquid flow distribution index remains unchanged, the structure enters an adaptive adjustment state, constructing a new disturbance path by rotating the disturbance nozzle angle and changing the injection position, and making short-term fine adjustments in conjunction with the inlet guide direction. Simultaneously, the main channel flow rate change response is briefly delayed during the disturbance process to ensure that the disturbed liquid can complete the initial upward diffusion in the local area.

[0018] The present invention has the following beneficial effects:

[0019] (1) The single-phase immersion liquid cooling temperature stratification assessment and disturbance injection balance control system can realize dynamic switching and fine control of liquid flow direction and injection path by constructing a linkage control mechanism between the main circulation path and the branch return path, effectively improving the coverage response speed of coolant in high heat load area and avoiding local heat accumulation.

[0020] (2) The single-phase immersion liquid cooling temperature stratification assessment and disturbance liquid injection balance control system, by introducing the coupling analysis method of the risk value of temperature layer difference of immersion liquid cooling cabinet and the evaluation value of liquid flow balance distribution, can perceive the evolution trend of liquid hot and cold layer structure in real time, identify the thermal inertia section in immersion liquid cooling cabinet caused by the lag of return path in advance, and ensure stable temperature flow circulation.

[0021] (3) The single-phase immersion liquid cooling temperature stratification assessment and disturbance injection balance control system, by introducing a periodic low-speed flushing process under cyclic stable state, maintains the consistency and plasticity of the flow direction inside the channel, effectively suppresses the accumulation of thermal inertia caused by long-term steady-state operation, and improves the heat dissipation stability of the server under high load continuous operation.

[0022] (4) The single-phase immersion liquid cooling temperature stratification assessment and disturbance injection balance control system can automatically identify the cold liquid floating lag zone and hot liquid accumulation zone during the cooling cycle by introducing a joint constraint mechanism of temperature profile fluctuation benchmark value and local temperature difference threshold, drive the disturbance structure fine adjustment, and realize the microscale redistribution of the circulation flow direction.

[0023] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0024] Figure 1 This is a structural diagram of the single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system of the present invention;

[0025] Figure 2 This is a line graph of the path switching determination value involved in this invention;

[0026] Figure 3 This is a flowchart of the single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system of the present invention;

[0027] Figure 4 This is a schematic diagram of the single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system of the present invention.

[0028] In the diagram, 1. Immersion liquid cooling cabinet; 2. Branch pipeline valves; 3. Main pipeline valve; 4. Main liquid outlet pipeline; 5. Pump; 6. Heat exchanger; 7. Branch pipeline. Detailed Implementation

[0029] 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.

[0030] Please see Figures 1-4This invention provides a technical solution: a single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system, including a multi-point dynamic monitoring module for reflux liquid, used to collect coolant state data, flow state data, and path thermal state data in the current liquid cooling branch, and preprocess the collected coolant state data, flow state data, and path thermal state data to construct a standardized liquid cooling operation characteristic dataset; an immersion liquid temperature stratification monitoring module, used to assess the degree of unevenness of coolant heat distribution in the vertical direction based on the standardized liquid cooling operation characteristic dataset, and dynamically adjust the local disturbance injection strategy based on the assessment results; and a flow distribution equalization assessment module, used to assess the standardized liquid cooling operation characteristic data... The system includes a collection module that analyzes the current liquid flow distribution and heat dissipation matching degree, and drives the optimization of flow direction coverage strategy based on the analysis results; a cooling path dynamic judgment and switching module that uses the heat distribution unevenness assessment results and the liquid flow distribution and heat dissipation matching degree analysis results as inputs to comprehensively evaluate the current cooling state and heat load changes, and intelligently adjusts the circulation path and heat exchange intervention mode based on the assessment results; and an immersion liquid cooling cabinet local disturbance self-balancing control module that judges the evolution trend of the thermal layer structure based on the temperature profile and liquid flow distribution state, guides local convection through bottom disturbance liquid injection, and dynamically adjusts the disturbance direction and intensity in combination with temperature rise feedback and distribution response results, while improving the local diffusion effect of the disturbed liquid with flow direction relaxation control.

[0031] Specifically, the steps for collecting coolant state data, fluid flow state data, and path thermal state data in the current liquid cooling branch are as follows:

[0032] Key parameters in the liquid cooling branch are collected in a structured manner, including coolant status data. This data includes the coolant temperature of each layer within the immersion liquid cooling cabinet 1, the liquid level height of the temperature sensors, and the coolant temperatures at the top and bottom of the immersion liquid cooling cabinet 1. Simultaneously, the number of sensor layers, the temperature difference between adjacent layers, and the height difference are recorded within the current acquisition period.

[0033] The top and bottom coolant temperatures of the immersion liquid-cooled cabinet 1 represent the high-temperature limit and low-temperature benchmark in the liquid heat distribution during the current cycle, forming a representative temperature difference reference dataset. In multiple consecutive acquisition cycles, the top and bottom coolant temperatures are processed into time series, and the stable intervals of temperature difference changes in the most recent effective cycles are extracted based on the processed time series. The mean of the stable intervals is recorded as the normalized benchmark value of temperature difference under the current operating stage.

[0034] Collect fluid flow status data, including: instantaneous flow velocity values, continuous flow velocity change data and injection response time in each region, and record the average fluid flow velocity in the branch return path;

[0035] The flow rate fluctuation time series is constructed by using the continuous flow rate change data of each chip area in the immersion liquid cooling cabinet 1 within a unit time. At the same time, the fluctuation degree of the liquid in the area within a unit time and the consistency of the disturbance response are comprehensively evaluated by combining the liquid injection response time of the disturbance injector to obtain the uniformity of coolant distribution in the area.

[0036] Collect path thermal status data, which includes: the effective heat dissipation area of ​​the chip in each region, the coolant temperature in the return branch, the coolant temperature at the outlet of the immersion liquid cooling cabinet 1, and the chip temperature rise rate. At the same time, record the total number of effective monitoring areas in the current cycle.

[0037] In this implementation scheme, comprehensive perception of three core parameters of the heat flow path during the liquid cooling cycle is achieved. A representative temperature difference reference dataset is constructed based on the coolant temperature gradient to extract heat distribution characteristics. The uniformity of coolant distribution is calculated by jointly calculating the flow velocity fluctuation and the consistency of the injection response to characterize the flow stability in local areas. At the same time, the heat transfer efficiency between paths is evaluated by the chip heat dissipation area and the liquid temperature rise difference, providing a high-precision input basis and dynamic response support for subsequent temperature difference risk identification, liquid flow optimization and control, and path switching strategies.

[0038] Specifically, the collected coolant state data, fluid flow state data, and path thermal state data are preprocessed to construct a standardized liquid cooling operation feature dataset. The specific steps are as follows: the collected coolant state data, fluid flow state data, and path thermal state data are preprocessed, and a layered identification index is introduced while retaining the original physical meaning. The coordinate standardization processing, including the correspondence between liquid level height and temperature sensor, is completed, and a layered thermal data mapping structure is constructed. At the same time, a fast window sliding comparison is performed on the temperature abrupt points that occur during the acquisition to eliminate temperature sequence abrupt changes and interlayer measurement point misalignment data caused by sensor drift and short-term communication interruption, so as to ensure the continuity and reliability of multi-cycle temperature trends.

[0039] Based on the temperature distribution characteristics of the coolant in each layer, and combined with the actual pipe layout structure and coolant flow direction constraints of the immersion liquid cooling cabinet 1, a distribution constraint matrix is ​​established to help determine the abnormal temperature area. The difference repair operation is performed on the detected temperature gradient reversal and abnormal static intervals, prioritizing the preservation of structural continuity and heat conduction direction consistency, and restoring a continuous and stable temperature profile that reflects the true flow-heat coupling state.

[0040] Before preprocessing, the path thermal state data is classified and mapped according to the mapping relationship between the chip area topology and the injection point. The coverage validity is verified by regional heat flux analysis. For chip areas with multiple cross-cooling coverage, heat dissipation area splitting is performed to ensure that the energy ownership of each hot zone is clear and the response path is independent in the subsequent processing.

[0041] To address data segment conflicts caused by the parallel operation of multiple injection units during liquid cooling, a sliding data window synchronization mechanism based on sampling time sequence is introduced in the preprocessing stage. This mechanism combines the response time difference within the region to resolve conflicts, prioritizing the retention of data from regions with continuously increasing flow fluctuations and eliminating duplicate response fields, thereby improving data consistency and response feature integrity under multi-region joint processing.

[0042] Finally, all preprocessed coolant state data, fluid flow state data, and path thermal state data are normalized to unify the numerical scale and distribution pattern, and a standardized liquid cooling operation characteristic dataset is constructed.

[0043] In this implementation plan, the collected coolant state data, fluid flow state data, and path thermal state data are standardized and preprocessed to construct a liquid cooling operation characteristic dataset that is structurally continuous, accurately distributed, and highly comparable. This improves the temperature profile reproduction, flow response consistency, and accuracy of thermal load attribution, providing high-quality data support for subsequent thermal layer structure identification, local disturbance control, and intelligent adjustment of circulation paths.

[0044] Specifically, the steps for evaluating the unevenness of coolant heat distribution in the vertical direction based on a standardized liquid cooling operation characteristic dataset are as follows: Real-time measurement data from temperature sensors on each layer inside the immersion liquid cooling cabinet 1 are collected to construct a vertical temperature profile sequence within the current acquisition period. The temperature gradient at the current location is represented by the ratio of the temperature difference between each adjacent layer to the height difference. The temperature gradient with the largest absolute value among all adjacent measurement points is extracted as the local maximum gradient of the profile. Simultaneously, by calculating the interquartile range of all temperature gradients, the upper limit of the acceptable gradient fluctuation range is obtained and recorded as the profile fluctuation reference value. The interquartile range of each adjacent layer is then calculated sequentially. The interlayer temperature gradient value is obtained by dividing the temperature difference between the two adjacent layers by the difference in liquid level height between the temperature sensors. The absolute value of the ratio of the temperature difference between the two layers is taken. The interlayer temperature gradient values ​​of each layer in the sensor are added together and divided by the number of sensor layers to obtain the average interlayer temperature value. The top coolant temperature of the immersion liquid-cooled cabinet 1 is subtracted from the bottom coolant temperature and then divided by the temperature difference normalization benchmark value to obtain the upper and lower temperature difference normalization ratio. The maximum local gradient of the profile is divided by the profile fluctuation benchmark value to obtain the gradient offset ratio. The average interlayer temperature value, the upper and lower temperature difference normalization ratio, and the gradient offset ratio are multiplied to obtain the temperature layer difference risk value.

[0045] The formula for calculating the risk value of temperature difference is:

[0046] ;

[0047] In the formula:

[0048] P represents the temperature difference risk value; The number of sensor layers indicates the total number of temperature acquisition points vertically arranged in the immersion liquid cooling cabinet 1. It is used to calculate the temperature gradient distribution in segments and is derived from the structural design and probe layout of the immersion liquid cooling cabinet 1. Indicates the first The coolant temperature of each layer is used to calculate the temperature difference between adjacent heights and is the basic data for constructing the vertical temperature profile. It originates from the [missing information - likely a specific data point or source]. The temperature sensor of the layer collects values ​​in real time; Indicates the first The liquid level height where the layer temperature sensor is located is used to calculate the liquid level distance between each two layers, and then derive the temperature gradient distribution. This is derived from the static height parameter table of the internal structure of the immersion liquid cooling cabinet 1. The top coolant temperature is used to indicate the temperature status of the uppermost liquid layer and is derived from the real-time measurement value of the temperature sensor in the top area of ​​the immersion liquid cooling cabinet 1. The bottom coolant temperature is used to indicate the temperature status of the liquid at the bottom layer. It is derived from the real-time measurement value of the temperature sensor in the bottom area of ​​the immersion liquid cooling cabinet 1. This represents the temperature difference normalization benchmark value, used to normalize the temperature difference between the top and bottom, so that the risk values ​​under different working conditions are comparable. It is derived from the average value of the steady-state range of the coolant temperature difference in the immersion liquid cooling cabinet 1 during the initial use. It represents the local maximum gradient of the profile and is used to quantify the temperature abrupt change phenomenon that occurs in a certain height segment of the immersion liquid-cooled cabinet 1. It is an important basis for determining the spatial location of thermal stratification. The reference value for profile fluctuation is used to measure the maximum acceptable temperature change gradient inside the immersion liquid-cooled cabinet 1 within the same monitoring period. It is a relative reference value for judging whether the temperature disturbance exceeds the normal interlayer transition range. It is derived from the statistical fluctuation range of the set of temperature gradient values ​​of all adjacent measuring points within the monitoring period and is automatically calculated.

[0049] In this implementation scheme, by comprehensively analyzing the temperature difference distribution, profile gradient change and its fluctuation stability over time between different vertical layers in the immersion liquid cooling cabinet 1, it is possible to identify whether the liquid thermal stratification structure is in different evolution states of continuous aggravation, slow convergence and critical equilibrium, and then determine whether the coolant is stuck at the top and the bottom coolant is insufficiently compensated due to insufficient return path structure and local disturbance.

[0050] Specifically, the steps for dynamically adjusting the local disturbance injection strategy based on the evaluation results are as follows:

[0051] The system compares the current temperature difference risk value with the temperature difference risk threshold in real time to dynamically identify the evolution trend of internal heat distribution in the immersion liquid-cooled cabinet 1 and drive the adaptive adjustment of the liquid injection strategy accordingly.

[0052] When the temperature difference risk value is less than or equal to the temperature difference risk threshold, the current thermal stratification structure is determined to be within a controllable range. The current branch return state remains unchanged, the existing liquid inlet configuration and flow direction are maintained, and multi-point liquid injection continues to be performed to maintain the uniformity of liquid flow. During this process, the data acquisition and calculation tasks of temperature measurement points of each layer of the immersion liquid cooling cabinet are updated at fixed intervals, and the change range of temperature acquisition values ​​of each layer is monitored in real time to dynamically assess whether the micro-fluctuation of temperature difference has an upward trend.

[0053] When the temperature difference risk value exceeds the temperature difference risk threshold, it indicates that a thermal stagnation structure has formed inside the immersion liquid cooling cabinet 1. Immediately close the branch pipe valve 2, switch the main circulation cooling path, and prevent high-temperature liquid that has not been treated by the heat exchanger from flowing back into the main cooling circuit. Simultaneously, suspend the liquid injection operation at the current inlet point to block the potential heat conduction backflow path. Then, activate the micro-disturbance injector located at the bottom of the immersion liquid cooling cabinet 1, and perform periodic low-speed disturbance injection without affecting the main cooling channel to guide the coolant to float and the local hot liquid to move laterally. At the same time, reconstruct the state of all injection channels in the immersion liquid cooling cabinet 1, reconstruct the injection path mapping relationship based on the temperature level distribution, and prioritize the allocation of disturbance channels to the temperature difference concentration area, thereby improving the thermal layer structure breaking efficiency.

[0054] In this implementation scheme, the temperature stratification status inside the immersion liquid-cooled cabinet 1 is dynamically monitored and the evolution trend of the thermal layer structure is determined in real time. Based on the comparison results of the temperature difference risk value and the preset threshold, the cooling path switching and disturbance liquid injection strategy are adaptively adjusted to ensure the balanced heat distribution in the liquid-cooled environment and prevent the formation of high-temperature stagnation areas, thereby improving the thermal control response efficiency and stability under different operating stages.

[0055] Specifically, the steps for analyzing the matching degree between the current liquid flow distribution and heat dissipation based on a standardized liquid cooling operation characteristic dataset are as follows: A dynamic index set for coolant distribution uniformity is constructed by calculating the ratio of continuous flow velocity changes to injection response time in each effective monitoring area within the current cycle. Then, in this dynamic index set, a median extraction method is used to obtain a normalized reference value for the distribution consistency level within the liquid cooling branch in the current cycle, which is denoted as the coolant distribution uniformity reference value. The instantaneous flow velocity value, the effective heat dissipation area of ​​the chip, and the corresponding coolant density correction factor in the current area are multiplied to obtain the regional liquid kinetic energy value for the current area. The ratio of the coolant distribution uniformity of each area to the coolant distribution uniformity reference value is subtracted from this value and multiplied by the distribution deviation weighting factor to obtain the distribution uniformity adjustment value. The regional liquid kinetic energy value is divided by the corresponding distribution uniformity adjustment value to obtain the regional balanced effective liquid flow value. The regional balanced effective liquid flow values ​​of each effective monitoring area within the current cycle are accumulated to obtain the liquid flow uniformity distribution evaluation value for the current cycle.

[0056] The formula for calculating the fluid flow equilibrium distribution evaluation value is:

[0057] ;

[0058] In the formula: C represents the fluid flow uniformity distribution assessment value; n represents the total number of effective monitoring areas; The instantaneous flow velocity value of the j-th region is used to characterize the change in the flow velocity of the coolant in that region. It is a basic parameter for analyzing the local fluid mobility and is derived from the multi-point miniature flow velocity sensors deployed inside the immersion liquid cooling cabinet 1. This represents the effective heat dissipation area of ​​the chip in region j, used to estimate the required contact range of the cooling liquid in this region. It is an input parameter for evaluating the distribution characteristics of cooling demand and is derived from the analysis results of the comparison between the real-time heat flux response and the physical geometric projection on the chip surface. This indicates the uniformity of coolant distribution in region j, reflecting the flow rate fluctuation and coverage diffusion of the liquid in this region per unit time. It is a reference quantity for identifying whether there are dead zones or local accumulation. This represents a reference value for coolant distribution uniformity, used to characterize the consistency level of the overall fluid flow distribution within the current cycle. It serves as a normalized benchmark for measuring the degree of deviation in a single area. It is derived from a dynamic index set for coolant distribution uniformity constructed by calculating the ratio of continuous flow velocity change data to injection response time in each effective monitoring area within the current cycle. Subsequently, the median extraction method is used to obtain a normalized reference value for the distribution consistency level within the liquid cooling branch in the current cycle from the dynamic index set for coolant distribution uniformity. The coolant density correction factor for region j ranges from 0.85 to 1.15. It is used to correct for differences in the inertial response and pressure drop performance of different coolant types under the same flow structure. It is derived from the ratio between the instantaneous density of the target cooling medium under current temperature and pressure conditions and the standard density reference value. Specifically, the local temperature and pressure data of the coolant in each monitoring region are first collected, and the actual density value of that region is calculated using the refrigerant density state equation. Then, the standard density reference value of the medium under rated operating conditions is extracted from the material performance database. The ratio of these two values ​​is the coolant density correction factor. The coolant density correction factor reflects the degree of density drift of the coolant in the current operating environment compared to standard operating conditions. When the actual density is significantly higher than the standard value, it means that there is local stagnation in the coolant, and the coolant density correction factor needs to be increased; conversely, it indicates that the medium's diffusivity is enhanced, and the coolant density correction factor should be decreased to improve the overall distribution elasticity. The distribution deviation weight factor, ranging from 0.5 to 2, is used to adjust the calibration weights for the spatial consistency of fluid velocity and the degree of multi-point distribution deviation within the monitoring area. It originates from the structural difference analysis results between velocity vectors in each region during the current monitoring period. In practice, instantaneous velocity vectors are first extracted from each region and mapped to a unified coordinate reference frame. Spatial vector clustering is then performed to identify the concentration of flow direction and the trend of deviation direction. Subsequently, the velocity center offset distance, azimuth angle difference, and velocity amplitude difference between each cluster are calculated, forming a velocity distribution deviation tensor. Finally, combining the deviation weight distribution between each region and the mainstream direction, the overall velocity field deviation concentration index is calculated and denoted as the distribution deviation weight factor. When there is severe unevenness in fluid distribution, large velocity differences between regions, or a significant imbalance trend between hot and cold spots, a larger value is used for the distribution deviation weight factor. Conversely, when the fluid distribution is basically uniform, minor local differences are acceptable, and no significant adjustment is needed, a smaller value is used to reduce sensitivity to minor deviations, maintain control stability, and avoid flow field disturbances caused by excessive adjustments.

[0059] This implementation scheme quantifies the distribution of coolant within different chip regions of the liquid-cooled server to identify the compatibility and flow field balance between the current fluid flow structure and the thermal load. By integrating instantaneous flow velocity changes, disturbance response delays, and flow velocity fluctuation standard deviations in each region per unit time, the scheme dynamically characterizes the coolant coverage consistency, disturbance response stability, and flow balance, thereby constructing a comprehensive index reflecting the overall fluid flow synergy. The fluid flow balance distribution assessment value serves as the core criterion for dynamic optimization of cooling paths, adjustment of injection strategies, and identification of abnormal states. It helps determine whether there are imbalances such as flow velocity deviation, insufficient cold zone coverage, and local blockages in the current liquid cooling structure, providing accurate data support for subsequent disturbance guidance, path switching, and cooling reconfiguration strategies, thereby improving the adaptive control capability and overall heat dissipation efficiency under complex thermal load changes.

[0060] Specifically, the steps for optimizing the flow coverage strategy based on the analysis results are as follows:

[0061] Based on the calculation results of the fluid flow equilibrium distribution evaluation value, the flow direction control strategy within the liquid cooling structure is dynamically adjusted according to the fluctuation trend of the fluid flow equilibrium distribution evaluation value in the time series. A complete process from data acquisition to preprocessing to fluid flow equilibrium distribution evaluation value calculation to result caching within a fixed time window is defined as one cycle.

[0062] When the fluid flow equilibrium distribution assessment value steadily increases over three or more consecutive cycles, and the increase in each cycle does not exceed the set fluctuation threshold, it indicates that the coolant distribution within the structure is becoming more balanced, the flow path adequately covers the heat source area, and the flow field disturbance is low with a high thermal load response matching degree. At this point, the current flow distribution structure is frozen, path reconstruction and valve control adjustment are paused, and the observation enhancement phase begins: the sampling density of the micro-disturbance area around the inlet point is expanded, more refined boundary velocity disturbance information is obtained through high-frequency sampling, and micro-low velocity disturbance injection tests are periodically performed to verify whether there is a hidden false steady-state risk in the surface steady state, ensuring that local flow is not masked by stagnant water and thermal inertia response.

[0063] When the fluid flow balance distribution assessment value shows a sudden increase within two consecutive cycles, and the increase exceeds the fluctuation threshold, it is determined that a sudden imbalance has occurred in the fluid flow coverage structure. All existing injection point operations are immediately interrupted, the main circulation cooling path is reconstructed, and the coolant self-balancing startup mechanism is activated. This mechanism is a rapid recovery process based on adaptive flow rate adjustment and injection channel reconstruction, specifically including three stages: The initial stage shuts down all injection modules and branch return paths to ensure a zero-injection steady state; the transition stage starts the main channel circulation at a safe minimum flow rate and sorts the injection points according to their average flow rate during the operating cycle, prioritizing the activation of injection points with the closest target average to achieve gradient recovery; the final stage introduces a delayed pressure increase strategy: after restarting each injection point, the corresponding flow rate is increased only after a one-cycle delay to avoid local flow field impacts and pipeline fluctuation rebounds caused by sudden pressure. Simultaneously, the spatial distribution and convergence speed of the chip temperature difference are monitored in real time. Once a continuous fluctuation in temperature difference and abnormal flow rate are detected in a certain area, the injection angle and flow path of that area are immediately rearranged to ensure that the newly restored structure has good dynamic adaptability and heat exchange stability.

[0064] When the liquid flow equilibrium distribution evaluation value continuously decreases over five consecutive cycles, and the temperature of each chip core area decreases synchronously, it indicates that the current liquid cooling structure has reached a relatively optimal equilibrium heat dissipation state. This reduces the trigger frequency of all disturbing liquid injection nozzles and extends the holding time of valves 2 in each branch pipe of the main cooling channel to reduce flow field disturbances caused by frequent switching. Simultaneously, a micro-pressure difference weak disturbance mechanism is introduced: subtle periodic pressure fluctuations maintain the activity and micro-circulation capacity of the boundary heat exchange surface, preventing the formation of a thermal inertia layer due to excessive stabilization. To prevent the accumulation of residual internal heat energy due to a fixed coolant flow direction during steady-state operation, a flexible flushing process is automatically triggered every ten cycles. While maintaining the liquid cooling path structure unchanged, a short-duration, large-volume injection is used to slowly redistribute the coolant, thereby achieving directional adjustment of the flow direction and re-energizing the equilibrium performance, ensuring the continuous heat dissipation efficiency and structural elasticity of the liquid cooling structure during operation.

[0065] In this implementation scheme, an adaptive liquid cooling flow direction control mechanism is constructed based on the dynamic trend of the liquid flow equilibrium distribution evaluation value. This mechanism enables the circulating cooling process to respond promptly to changes in the stability of the coolant distribution structure, sudden changes in the flow field, and local imbalances. This allows for precise control of the heat load matching degree and timely judgment of the cooling path reconstruction. When the liquid flow equilibrium distribution evaluation value shows a steady upward trend, freezing the current flow distribution structure, increasing the micro-disturbance observation density, and performing low-intensity disturbance verification helps identify local pseudo-steady states and stabilize the overall heat dissipation structure. When the liquid flow equilibrium distribution evaluation value suddenly increases and exceeds the threshold, the abnormal channel is quickly interrupted and the liquid injection path is reconstructed through a self-balancing activation mechanism, effectively avoiding a sudden drop in cooling efficiency and the risk of heat accumulation on the chip. When the liquid flow equilibrium distribution evaluation value continues to decrease and is accompanied by a synchronous decrease in temperature, a steady-state maintenance mode is entered. Micro-pressure difference disturbances and a flexible flushing process are introduced to maintain boundary heat transfer activity and prevent the accumulation of thermal inertia, comprehensively improving the robustness and stability of the liquid cooling circulation structure under dynamic heat dissipation scenarios.

[0066] Specifically, the steps for comprehensively evaluating the current cooling state and heat load changes using the results of the thermal unevenness assessment and the analysis of the matching degree between liquid flow distribution and heat dissipation as input are as follows: Obtain the temperature difference risk value and the liquid flow equilibrium distribution assessment value; calculate the absolute value of the difference between the current coolant temperature in the return branch and the coolant temperature at the outlet of the immersion liquid cooling cabinet 1; multiply the temperature difference risk value, the liquid flow equilibrium distribution assessment value, and the absolute value of the difference between the coolant temperature in the return branch and the coolant temperature at the outlet of the immersion liquid cooling cabinet 1 to obtain the temperature difference enhancement load term; add the chip temperature rise rate to the reciprocal of the average liquid flow velocity in the current branch return path and then add one to obtain the liquid flow attenuation penalty value; divide the temperature difference enhancement load term by the liquid flow attenuation penalty value and then take the logarithm to obtain the path switching judgment value at the current moment.

[0067] The formula for calculating the path switching determination value is:

[0068] ;

[0069] In the formula, Indicates the path switching determination value; This indicates the risk value for temperature difference between layers; This represents the evaluation value for the uniform distribution of fluid flow. This indicates the current coolant temperature in the return branch, used to assess the remaining heat absorption capacity of the liquid in this path. It is a key variable for determining whether there is potential for secondary cooling and is derived from the temperature probe data collected by the probe embedded in the return channel of the branch. The value represents the coolant temperature at the outlet of the immersion liquid cooling cabinet 1. It is used to form a difference with the return liquid temperature and reflects the intensity of temperature transfer between cooling paths. It is derived from the measurement results of the real-time temperature sensor at the outlet of the immersion liquid cooling cabinet 1. This represents the average liquid flow velocity in the current branch return path, which reflects the smoothness of the liquid flow in the return channel and is an important factor in judging the return efficiency. It is derived from the sequence of flow velocity sensors installed in the branch pipe 7. It represents the rate of temperature rise of the chip and is used to quantify the severity of the current change in the chip's thermal load. It is an important parameter for determining whether there is a short-term trend of lag in heat dissipation and is derived from the time difference value of the chip surface temperature sensor over two consecutive cycles.

[0070] In this implementation example, the temperature difference risk value of Example 1 is set to 0.36, the liquid flow balance distribution evaluation value is 0.68, the coolant temperature in the return branch is 43.5, the coolant temperature at the outlet is 26.1, the average liquid flow rate is 0.94, the chip temperature rise rate is 0.11, and the path switching judgment value is 2.057.

[0071] In Example 2, the temperature difference risk value was set to 0.42, the liquid flow balance distribution evaluation value was 0.75, the coolant temperature in the return branch was 44.2, the coolant temperature at the outlet was 25.7, the average liquid flow rate was 1.01, the chip temperature rise rate was 0.09, and the path switching judgment value was 2.324.

[0072] In Example 3, the temperature difference risk value was set to 0.39, the liquid flow balance distribution evaluation value was 0.72, the coolant temperature in the return branch was 41.8, the coolant temperature at the outlet was 26.4, the average liquid flow rate was 0.98, the chip temperature rise rate was 0.10, and the path switching judgment value was 2.025.

[0073] In Example 4, the temperature difference risk value was set to 0.33, the liquid flow balance distribution evaluation value was 0.65, the coolant temperature in the return branch was 42.6, the coolant temperature at the outlet was 25.9, the average liquid flow rate was 0.93, the chip temperature rise rate was 0.12, and the path switching judgment value was 1.981.

[0074] In Example 5, the temperature difference risk value was set to 0.45, the liquid flow balance distribution evaluation value was 0.79, the coolant temperature in the return branch was 45.1, the coolant temperature at the outlet was 26.0, the average liquid flow rate was 1.07, the chip temperature rise rate was 0.08, and the path switching judgment value was 2.446.

[0075] In Example 6, the temperature difference risk value was set to 0.37, the liquid flow balance distribution evaluation value was 0.70, the coolant temperature in the return branch was 43.0, the coolant temperature at the outlet was 25.8, the average liquid flow rate was 0.96, the chip temperature rise rate was 0.11, and the path switching judgment value was 2.094.

[0076] In Example 7, the temperature difference risk value was set to 0.40, the fluid flow balance distribution evaluation value to 0.67, the coolant temperature in the return branch to 42.2°C, the coolant temperature at the outlet to 25.5°C, the average fluid velocity to 0.95, the chip temperature rise rate to 0.10, and the path switching determination value to 2.146. The path switching determination values ​​for each example were calculated, as shown in Table 1.

[0077] Table 1 Path Switching Judgment Values

[0078]

[0079] like Figure 2 As shown in Table 1, this is a line graph of the path switching determination values ​​provided in this application example. Figure 2 It can be seen that the highest path switching judgment value is for Example 5, indicating that its immersion liquid cooling cabinet 1 has a high risk of temperature difference, good liquid flow balance distribution, low chip temperature rise rate, and large temperature difference between hot and cold liquids. This indicates that there is a significant matching deviation between the coolant flow structure and the heat load in this example, which is prone to imbalance of hot and cold spot distribution and backflow delay. It is necessary to prioritize adjusting the injection path and flow ratio and activate the active path reconstruction mechanism to restore dynamic balance. The lowest path switching judgment value is for Example 4. Although its chip temperature rise rate is slightly higher, the overall liquid flow distribution stability is good, the temperature difference gradient of the immersion liquid cooling cabinet 1 is relatively gentle, and the temperature difference between the coolant inlet and outlet is moderate. This indicates that the current flow path structure is relatively reasonable and has good heat load bearing capacity and flow stability. Path adjustment can be temporarily suspended, and the existing cooling layout can be maintained to reduce disturbance. The line graph of the path switching judgment value intuitively reflects the dynamic changes in the stability of liquid flow, thermal coupling coordination and path adaptability of different instances during the cooling process. The higher the value, the greater the pressure of path regulation, and the more necessary it is to intervene in dynamic reconstruction and regulation optimization in order to improve the overall cooling efficiency and stability.

[0080] Specifically, the steps for intelligently adjusting the circulation path and heat exchange intervention method based on the evaluation results are as follows: Real-time comparison of the current path switching judgment value with the path switching threshold. The path switching threshold includes a first switching threshold and a second switching threshold, used to distinguish the flow direction control level under different heat load pressures. The current path switching judgment value is jointly composed of the temperature difference risk value of the immersion liquid cooling cabinet 1, the liquid flow balance distribution evaluation value, the chip temperature rise rate, and the temperature difference between the hot and cold liquids, reflecting the comprehensive intensity of cooling structure imbalance and heat transfer pressure.

[0081] When the path switching determination value is less than or equal to the second switching threshold, it indicates that the current cooling state is still within the thermal load tolerance range. The current branch return path operation state remains unchanged, the flow distribution structure and flow direction settings of each liquid inlet point are maintained, the path switching execution is paused, and only the return liquid temperature and chip temperature difference are periodically sampled at low frequency, while the diffusion trend of the liquid on the cooling surface is recorded.

[0082] When the path switching judgment value is greater than the second switching threshold and less than or equal to the first switching threshold, it indicates that the current cooling flow field has a certain degree of instability, but has not yet exceeded the thermal load control limit. At this time, the opening of branch pipeline valve 2 is reduced and the pressure of the main circulation channel is gradually increased. This suppresses the branch flow while increasing the cooling capacity of the main channel, maintaining the two cooling paths in parallel operation for a short period, constructing a transitional cooling redundancy structure, forming a flow field buffer area, and avoiding temperature shocks caused by sudden cooling changes. A dynamic path transfer recording mechanism is also activated to continuously collect changes in the flow direction of each injection point and changes in the temperature difference on the chip surface. After the fluctuation trend of the path switching judgment value in the next cycle is clear, it is then determined whether to completely switch to the main circulation channel.

[0083] When the path switching judgment value exceeds the first switching threshold, it is determined that the cooling imbalance has reached a dangerous level. A path switching command is immediately executed, closing the branch return channel and fully switching to the main circulation path to prioritize restoring the cooling stability of the chip area. Simultaneously, the return liquid direction and inlet point weight configuration are reset, and the liquid injection ratio in each area is readjusted. All disturbance-induced liquid injection operations are suspended to prevent local disturbances from interfering with the main circulation flow direction. The main channel liquid velocity balance constraint logic is activated to dynamically constrain the liquid velocity impact behavior in high-risk areas. While ensuring high-flow-rate cooling, flow direction restrictions prevent abnormal temperature rise at the chip edges, further improving heat exchange efficiency and preventing local hotspot runaway.

[0084] In this implementation scheme, a hierarchical response management system is implemented for the dynamic adjustment process of the cooling path based on the path switching judgment value, constructing an adaptive switching control mechanism that balances cooling efficiency, flow stability, and chip thermal safety. By comparing the path switching judgment value with two thresholds in real time, the system dynamically identifies whether there is an imbalance in heat conduction and abnormal flow trends in the cooling structure, and triggers different levels of cooling path intervention operations accordingly. When the cooling state is stable, the original path is maintained and the frequency of disturbances is reduced; when moderate structural fluctuations occur, a parallel buffer path is constructed and the decision is delayed; and when the cooling imbalance is severe, forced path switching and flow reconstruction are immediately executed to ensure the heat exchange stability and cooling efficiency of the chip under high thermal load, effectively preventing performance degradation and hardware risks caused by local temperature runaway.

[0085] Specifically, the method is used to determine the evolution trend of the thermal layer structure based on temperature profile and liquid flow distribution, guide local convection through bottom disturbance injection, dynamically adjust the disturbance direction and intensity by combining temperature rise feedback and distribution response results, and enhance the local diffusion effect of the disturbed liquid by coordinating flow relaxation control. The specific steps are as follows:

[0086] During the period when the branch return path is open, the internal temperature profile data of the immersion liquid-cooled cabinet 1, collected by the vertical temperature sensor array, is continuously monitored. This data is then combined with the liquid flow equilibrium distribution evaluation value and the path switching judgment value to determine whether there is a trend of the thermal layer structure becoming stable. When the monitoring results show that the liquid temperature at the top of the immersion liquid-cooled cabinet 1 is consistently higher than that at the bottom for multiple sampling cycles, and the multi-layer temperature difference change trend along the vertical distribution gradually converges, it is determined that the liquid thermal distribution has a potential stable state. Subsequently, the disturbance injection response mechanism is activated: the micro-jet nozzles and lateral directional channels deployed at the bottom of the immersion liquid-cooled cabinet 1 are automatically used to intermittently disturb the coolant injection at the minimum flow rate without reconstructing the main cooling path and branch return structure. This induces slow local liquid convection, achieving slow rise of cold liquid and lateral movement of hot liquid, thereby promoting the natural decomposition and redistribution of the local thermal layer structure.

[0087] After each round of perturbation injection, the temperature rise rate of the chip hot zone and the dynamic evolution trend of the temperature difference in the multiple layers of the immersion liquid-cooled cabinet 1 are analyzed simultaneously. When the temperature difference convergence trend is obvious and the liquid flow distribution index shows a continuous upward trend, the current perturbation strategy is deemed effective, and the established perturbation frequency, injection direction, and injection rhythm are maintained. Conversely, when the temperature rise change after perturbation is weak and the liquid flow distribution index does not show significant improvement, the structure enters an adaptive adjustment state, actively adjusting the angle of the perturbation nozzle and the injection position, and linking the main inlet flow direction to perform short-term fine-tuning operations, constructing a new perturbation path to improve local heat exchange efficiency. At the same time, a flow relaxation control strategy is introduced during the perturbation process to briefly delay the change in the main channel flow velocity response, allowing the perturbation liquid to have sufficient initial diffusion time in the local area, ensuring that it completes an effective rising, lateral movement, and heat exchange process, and enhancing the perturbation liquid's ability to intervene in and adjust the stability of the hot layer structure.

[0088] like Figure 3The diagram shows the flowchart of the single-phase immersion liquid cooling temperature stratification assessment and disturbance-induced liquid injection equalization control system of this invention, illustrating a cooling path switching control process driven by chip thermal load. Starting with chip heating, the chip continuously releases heat during operation. After absorbing heat, the temperature of the liquid cooling medium is monitored in real time. The system determines whether the temperature of the coolant after heat absorption meets the standard. If it does, it enters a branch circuit, flowing directly back to the panel for secondary cooling, achieving rapid local heat removal and avoiding excessive energy consumption and redundant cycles. If it does not meet the standard, it enters the main circuit, where the coolant passes through a post-circulation stage and flows to the heat exchanger for further cooling, ensuring the coolant temperature returns to a reasonable range. The liquid after heat exchange re-enters the monitoring stage, forming a complete closed-loop control.

[0089] like Figure 4 The diagram shows the principle of the single-phase immersion liquid cooling temperature stratification assessment and disturbance liquid injection equalization control system of this invention, illustrating its working principle and structure. It is mainly used for heat exchange management in high-heat-load scenarios, including chips and servers, and achieves efficient cooling through immersion liquid cooling. The immersion liquid cooling cabinet 1 carries the equipment to be cooled and is completely covered by coolant, achieving uniform heat absorption. Branch pipe valves 2 control the coolant return from the cabinet to the cooling circuit via branch pipes 7. The main pipe valve 3 switches whether the coolant enters the main cooling path, controlling the switching between branch and main paths. The main outlet pipe 4 is the main channel for coolant to flow out of the cabinet and into the cooling circuit. The pump 5 provides the driving force for liquid circulation, propelling the coolant through the pipes. The heat exchanger 6 releases the absorbed heat to the external environment, achieving rapid heat removal. Branch pipes 7 form auxiliary paths for coolant return, which can be used for local circulation and secondary cooling. The entire system uses a pump to drive the coolant to flow between the submerged liquid cooling cabinet 1 and the heat exchanger 6, and combines valves to adjust the flow direction to achieve multi-path control and dynamic cooling strategy switching, ensuring that the equipment maintains stable thermal management capabilities even under high heat density operating conditions.

[0090] In this implementation scheme, during the operation of the branch return path, a joint analysis based on the temperature profile change trend and liquid flow equilibrium characteristics dynamically identifies potential risk sections within the immersion liquid-cooled cabinet 1 where the thermal layer structure tends to be stable but is not completely decoupled. When the top liquid temperature is detected to be higher than the bottom for three consecutive cycles and the temperature difference distribution tends to converge, intermittent liquid injection intervention is implemented with minimal disturbance through the bottom micro-jet nozzle and lateral directional channel. Without affecting the main circulation path and return channel structure, local slow liquid convection is triggered, achieving cold liquid floating and hot liquid lateral diffusion, thereby actively breaking the pseudo-steady-state thermal layer accumulation phenomenon. The disturbance effect is judged by real-time feedback of the chip hot zone temperature rise rate and the multi-layer temperature difference change trend of the immersion liquid-cooled cabinet 1 after disturbance liquid injection, improving local heat exchange efficiency, suppressing the continuous accumulation of high-temperature stagnant areas, and ensuring the dynamic steady-state control capability of the overall liquid cooling structure.

[0091] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0092] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A single-phase immersion liquid-cooled temperature stratification assessment and disturbance-induced liquid injection equalization control system, characterized in that: include: The multi-point liquid inlet dynamic monitoring module for reflux liquid is used to collect coolant status data, liquid flow status data, and path thermal status data in the current liquid cooling branch, and to preprocess the collected coolant status data, liquid flow status data, and path thermal status data to construct a standardized liquid cooling operation feature dataset. The immersion liquid temperature stratification monitoring module is used to assess the degree of uneven heat distribution of coolant in the vertical direction based on a standardized liquid cooling operation characteristic dataset, and dynamically adjust the local disturbance injection strategy based on the assessment results. The fluid distribution equilibrium assessment module is used to analyze the matching degree between the current fluid distribution and heat dissipation based on a standardized liquid cooling operation characteristic dataset, and drive the optimization of the flow direction coverage strategy based on the analysis results. The cooling path dynamic determination and switching module is used to comprehensively evaluate the current cooling status and heat load changes based on the evaluation results of the uneven heat distribution and the analysis results of the matching degree of liquid flow distribution and heat dissipation, and intelligently adjust the circulation path and heat exchange intervention mode based on the evaluation results. The immersion liquid-cooled cabinet local disturbance self-balancing control module is used to judge the evolution trend of the thermal layer structure based on the temperature profile and liquid flow distribution. It guides local convection through bottom disturbance liquid injection, and dynamically adjusts the disturbance direction and intensity by combining temperature rise feedback and distribution response results. At the same time, it improves the local diffusion effect of the disturbed liquid with flow relaxation control.

2. The single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system according to claim 1, characterized in that: The specific steps for collecting coolant state data, fluid flow state data, and path thermal state data in the current liquid cooling branch are as follows: Key parameters in the liquid cooling branch are collected in a structured manner, and coolant status data is collected, including: coolant temperature of each layer in the immersion liquid cooling cabinet, liquid level height of the temperature sensor, coolant temperature at the top and bottom of the immersion liquid cooling cabinet, and the number of sensor layers, temperature difference and height difference between each adjacent layer in the current collection period are recorded. The top and bottom coolant temperatures of the immersion liquid-cooled cabinet represent the high-temperature limit and low-temperature benchmark in the liquid heat distribution during the current cycle, forming a representative temperature difference reference dataset. Over multiple consecutive acquisition cycles, the top and bottom coolant temperatures are processed into time series, and based on the processed time series, a stable range of temperature difference changes over several effective cycles is extracted. The mean of the stable range is recorded as the normalized reference value of the temperature difference under the current operating stage. Collect fluid flow status data, including: instantaneous flow velocity values, continuous flow velocity change data and injection response time in each region, and record the average fluid flow velocity in the branch return path; The flow rate fluctuation time series is constructed by using the continuous flow rate change data of each chip area in the immersion liquid cooling cabinet per unit time. At the same time, the fluctuation degree of the liquid in the area per unit time and the consistency of the disturbance response are comprehensively evaluated by combining the liquid injection response time of the disturbance injector to obtain the uniformity of coolant distribution in the area. Collect path thermal status data, which includes: the effective heat dissipation area of ​​the chip in each region, the coolant temperature in the return branch, the coolant temperature at the outlet of the immersion liquid cooling cabinet, and the chip temperature rise rate. At the same time, record the total number of effective monitoring areas in the current cycle.

3. The single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system according to claim 1, characterized in that: The specific steps for preprocessing the collected coolant state data, fluid flow state data, and path thermal state data to construct a standardized liquid cooling operation feature dataset are as follows: The collected coolant state data, fluid flow state data, and path thermal state data are preprocessed to complete the coordinate standardization of the correspondence between liquid level and temperature sensor, and to eliminate temperature sequence abrupt changes and interlayer measurement point misalignment caused by sensor drift and short-term communication interruption. Based on the distribution characteristics of coolant temperature in each layer, combined with the actual pipe layout structure of the immersion liquid cooling cabinet and coolant flow direction constraints, the temperature gradient reversal and abnormal static intervals are repaired by difference to restore a continuous and stable temperature profile. Before preprocessing, the thermal state data of the path is classified and mapped according to the mapping relationship between the chip region topology and the liquid injection point. For chip regions with multiple overlapping cooling coverage, heat dissipation area splitting is performed. To address data segment conflicts caused by the parallel operation of multiple injection units during liquid cooling, a data window synchronization mechanism is introduced in the preprocessing stage to prioritize the retention of data in areas where flow fluctuations are continuously increasing and to remove duplicate response fields. All preprocessed coolant state data, fluid flow state data, and path thermal state data are normalized to construct a standardized liquid cooling operation characteristic dataset.

4. The single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system according to claim 1, characterized in that: The specific steps for evaluating the degree of uneven heat distribution of coolant in the vertical direction based on a standardized liquid cooling operation characteristic dataset are as follows: Real-time measurement data from temperature sensors on each layer inside the immersion liquid-cooled cabinet are collected to construct a vertical temperature profile sequence within the current acquisition period. The temperature gradient at the current location is represented by the ratio of the temperature difference between each two adjacent layers to the height difference. The temperature gradient with the largest absolute value among all adjacent measurement points is extracted as the local maximum gradient of the profile. At the same time, by calculating the interquartile range of all temperature gradients, the upper limit of the acceptable gradient fluctuation range is obtained and recorded as the profile fluctuation reference value. The temperature difference between each two adjacent layers is calculated sequentially and divided by the difference in liquid level between the temperature sensors of the two adjacent layers. The absolute value of the ratio of the temperature difference between the coolant and the difference in liquid level between the temperature sensors is taken to obtain the interlayer temperature gradient value. The average interlayer temperature value is obtained by summing the interlayer temperature gradient values ​​of each layer in the sensor and dividing by the number of layers in the sensor. Subtract the bottom coolant temperature from the top coolant temperature of the immersion liquid-cooled cabinet, and then divide by the temperature difference normalization benchmark value to obtain the upper and lower temperature difference normalization ratio. Divide the local maximum gradient of the profile by the profile fluctuation reference value to obtain the gradient offset ratio; The temperature difference risk value is obtained by multiplying the average interlayer temperature value, the normalization ratio of the temperature difference between the upper and lower layers, and the gradient offset ratio.

5. The single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system according to claim 1, characterized in that: The specific steps for dynamically adjusting the local disturbance injection strategy based on the evaluation results are as follows: Real-time comparison of the current temperature difference risk value with the temperature difference risk threshold: When the temperature difference risk value is less than or equal to the temperature difference risk threshold, the current branch return state remains unchanged, the existing liquid inlet configuration and flow direction are maintained, and multi-point liquid injection is continued to be performed. The change range of temperature acquisition values ​​of each layer is monitored, and the data acquisition and calculation of temperature measurement points of each layer are updated at a fixed period. When the temperature difference risk value exceeds the temperature difference risk threshold, immediately close the branch pipeline valve, switch the main circulation cooling path, simultaneously suspend the liquid injection operation at the current inlet point, activate the bottom micro-disturbance injector to perform periodic low-speed disturbance injection, and reconstruct the internal liquid injection path mapping relationship of the immersion liquid cooling cabinet.

6. The single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system according to claim 1, characterized in that: The specific steps for analyzing the matching degree between the current liquid flow distribution and heat dissipation based on the standardized liquid cooling operation characteristic dataset are as follows: By calculating the ratio of continuous flow rate change data to injection response time in each effective monitoring area within the current cycle, a dynamic index set for coolant distribution uniformity is constructed. Subsequently, in the dynamic index set for coolant distribution uniformity, the median extraction method is used to obtain the normalized reference value of the distribution consistency level within the liquid cooling branch in the current cycle, which is denoted as the coolant distribution uniformity reference value. Multiply the instantaneous flow rate of the current region, the effective heat dissipation area of ​​the chip, and the corresponding coolant density correction factor to obtain the region's kinetic energy value. Subtract the ratio of the coolant distribution uniformity of each region to the reference value of coolant distribution uniformity, and then multiply by the distribution deviation weighting factor to obtain the distribution uniformity adjustment value. Divide the regional fluid kinetic energy value by the corresponding distribution uniformity adjustment value to obtain the regional balanced effective fluid flow value. Then, sum the regional balanced effective fluid flow values ​​of each effective monitoring area in the current period to obtain the fluid flow balance distribution evaluation value for the current period.

7. The single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system according to claim 1, characterized in that: The specific steps for optimizing the flow direction coverage strategy based on the analysis results are as follows: Based on the calculation results of the fluid flow equilibrium distribution assessment value, the flow direction control strategy is dynamically adjusted according to the fluctuation trend of the real-time fluid flow equilibrium distribution assessment value: When the fluid flow equilibrium distribution evaluation value shows an upward trend for three or more consecutive cycles, and the increase does not exceed the fluctuation amplitude threshold, it indicates that the coolant flow structure tends to be stable and has a high degree of matching with the heat load. The current flow distribution structure is frozen and path reconstruction is paused, and the observation enhancement phase is entered: the sampling density of the micro-disturbance area around the liquid inlet point is expanded, and micro-low flow rate disturbance liquid injection tests are performed regularly to verify whether the current steady state is a true steady state and avoid local false steady state failure. When the fluid flow balance distribution evaluation value increases beyond the fluctuation threshold within two cycles, it is determined that the fluid flow coverage structure has suddenly become unbalanced. All current injection points are interrupted, the main circulation path is reset, the coolant self-balancing start-up mechanism is activated, and gradient recovery is performed based on the average flow rate of each injection point. At the same time, a delayed pressure boosting strategy is enabled to avoid sudden pressure impact. During the recovery process, the spatial distribution and convergence speed of the chip temperature difference are monitored in real time. When a disordered area is found, the injection angle and flow path are forcibly rearranged. When the fluid flow balance distribution evaluation value shows a downward trend over five consecutive cycles, and the chip temperature decreases synchronously, it is determined that the current coolant layout has reached a relatively optimal balance structure. The frequency of disturbance nozzles is reduced, the holding time of the main pipeline valve is extended, and a micro-pressure difference weak disturbance mechanism is introduced to maintain the boundary heat transfer activity. At the same time, a flexible flush is automatically performed every ten cycles to achieve flow redistribution through short-time large-volume liquid injection to prevent the accumulation of thermal inertia.

8. The single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system according to claim 1, characterized in that: The specific steps for comprehensively evaluating the current cooling state and heat load changes, using the assessment results of uneven heat distribution and the analysis results of the matching degree between fluid flow distribution and heat dissipation as inputs, are as follows: Obtain the temperature difference risk value and the liquid flow balance distribution assessment value, calculate the absolute value of the difference between the current coolant temperature in the return branch and the coolant temperature at the outlet of the submerged liquid cooler cabinet, and multiply the temperature difference risk value, the liquid flow balance distribution assessment value and the absolute value of the difference between the coolant temperature in the return branch and the coolant temperature at the outlet of the submerged liquid cooler cabinet to obtain the temperature difference enhancement load term; Add the chip temperature rise rate to the reciprocal of the average liquid flow rate in the current branch return path, and then add one to obtain the liquid flow attenuation penalty value. Divide the temperature difference enhancement load term by the fluid flow attenuation penalty value and then take the logarithm to obtain the path switching determination value at the current moment.

9. The single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system according to claim 1, characterized in that: The specific steps for intelligently adjusting the circulation path and heat exchange intervention method based on the evaluation results are as follows: Real-time comparison of the current path switching determination value with the path switching threshold, which includes a first switching threshold and a second switching threshold: When the path switching judgment value is less than or equal to the second switching threshold, the current branch return path operation status remains unchanged, the flow distribution structure and flow direction settings of each liquid inlet point are maintained, the path switching execution is paused, and only the return liquid temperature and chip temperature difference are periodically sampled at low frequency. At the same time, the diffusion trend of the liquid on the cooling surface is recorded, and the judgment is executed again in the next cycle. When the path switching determination value is greater than the second switching threshold and less than or equal to the first switching threshold, the opening of the branch pipeline valve is reduced and the pressure of the main circulation channel is gradually increased to keep the two cooling paths running in parallel for a short time. While ensuring the stability of the chip heat exchange efficiency, a flow field buffer area is established, and a dynamic path transfer recording mechanism is started. The path switching determination value of the next cycle is used to determine whether to switch to the main circulation channel completely. When the path switching judgment value is greater than the first switching threshold, the path switching command is executed immediately, the branch return channel is closed, and the main circulation path is switched to perform forced cooling. At the same time, the return liquid direction and inlet point weight configuration are reset, all disturbance injection operations are suspended, and the main channel liquid velocity balance constraint logic is activated. While ensuring high flow rate cooling, the concentrated injection to the overload area is restricted, the heat exchange efficiency is improved, and the chip edge local temperature rise is prevented due to high flow rate impact.

10. The single-phase immersion liquid cooling temperature stratification assessment and disturbance injection equalization control system according to claim 1, characterized in that: The specific steps for determining the evolution trend of the thermal layer structure based on temperature profile and liquid flow distribution, guiding local convection through bottom disturbance injection, dynamically adjusting the disturbance direction and intensity by combining temperature rise feedback and distribution response results, and simultaneously enhancing the local diffusion effect of the disturbed liquid through flow relaxation control are as follows: During the period when the branch return path is open, the internal temperature profile data of the immersion liquid cooling cabinet collected by the vertical temperature sensor array is continuously monitored, and the data is analyzed together with the liquid flow balance distribution evaluation value and the path switching judgment value to determine whether there is a trend of the thermal layer structure becoming stable. When the top liquid temperature of the immersion liquid-cooled cabinet is detected to be higher than that of the bottom for three consecutive cycles, and the temperature difference changes of multiple layers converge, the coolant is automatically injected intermittently at a minimum flow rate through the micro-jet nozzles and lateral directional channels deployed at the bottom of the immersion liquid-cooled cabinet without changing the structure of the main cooling passage and return path. This causes slow local liquid convection, making the cold liquid float and the hot liquid move laterally. After each round of perturbation injection, the temperature rise rate of the chip hot zone and the temperature difference change trend of the multi-layer immersion liquid-cooled cabinet are analyzed. When the temperature difference converges significantly and the liquid flow distribution index shows an upward trend, the current perturbation frequency and direction are maintained. When the liquid flow distribution index does not change, the structure enters an adaptive adjustment state. By rotating the angle of the perturbation nozzle and changing the liquid injection position, a new perturbation path is constructed, and the flow direction of the liquid inlet is adjusted in a short time. At the same time, the response of the main channel flow velocity change is briefly delayed during the perturbation process to ensure that the perturbation liquid can complete the initial upward diffusion in the local area.

Images