Layered microchannel liquid cooling method and system based on magnetic interface

Abstract

The application discloses a layered micro-channel liquid cooling method and system based on a magnetic suction interface and relates to the technical field of micro-channel liquid cooling structures.The layered micro-channel liquid cooling method and system based on the magnetic suction interface comprises the following steps of S1, collecting and preprocessing temperature characteristic data and fluid flow velocity and pressure data, and constructing a standardized heat load trend data set;S2, evaluating the cooling liquid supply state of each layer and dynamically adjusting the liquid supply rhythm of the corresponding layer;S3, analyzing the cooling demand intensity of each layer and driving the working frequency variation of the branch pump of the corresponding layer; and S4, evaluating the coordination state of the liquid supply, heat dissipation and liquid return process of each layer and dynamically adjusting the liquid cooling control strategy.The application solves the problem that the existing liquid cooling distribution path cannot realize dynamic flow adjustment according to the heat load difference of each layer of equipment in a multi-layer server structure, resulting in uneven distribution of cooling resources and serious energy waste.

Description

Technical Field

[0001] This invention relates to the field of microchannel liquid cooling structure technology, specifically to a layered microchannel liquid cooling method and system based on a magnetic interface. Background Technology

[0002] As the scale of high-performance computing equipment, data centers, and artificial intelligence training clusters continues to expand, the power consumption density and heat flux of electronic components are constantly increasing. The demand for rack-level thermal management is gradually upgrading from traditional air cooling or single-pipe liquid cooling to efficient, layered, and intelligent liquid cooling pathways. In practical applications, multi-layer server structures can directly liquid cool processors, storage modules, and high-speed interfaces at different levels through layered cold plate microchannels. Simultaneously, combined with multi-source sensors such as temperature, flow rate, and pressure, real-time monitoring of heat load changes, liquid supply status, and return flow within each layer enables dynamic monitoring of the rack's internal operating status across all levels and throughout the entire process. Especially in multi-layer structures, the focus of operation and maintenance has gradually expanded from the overall heat dissipation capacity to a joint analysis of the heat exchange efficiency of each layer's cold plates, the stability of the evaporation-condensation link, and the rationality of inter-layer liquid supply resource allocation.

[0003] For example, the invention patent with publication number CN114126357B relates to the field of two-phase liquid-cooled microchannels, specifically disclosing a two-phase liquid-cooled microchannel and its manufacturing method, as well as a server. The two-phase liquid-cooled microchannel includes a microchannel body with a porous structure inside. The height of the porous structure decreases from the inlet to the outlet of the microchannel body, and the upper surface curve of the porous structure is parabolic, causing the gas phase space inside the microchannel to increase from the inlet to the outlet. This invention can avoid the phenomenon of a sharp increase in gas phase flow velocity and pressure, eliminate gas blockage, reduce pressure oscillations in two-phase flow, enhance heat transfer, and improve system stability.

[0004] For example, invention patent CN109068538B discloses a liquid-cooled heat sink structure based on diamond microfluidic channels, including a diamond substrate and a diamond epitaxial layer stacked together. Several diamond microfluidic channels with a certain interval are arranged within the diamond substrate, and the diamond epitaxial layer is used to bond the object to be cooled to its surface. This invention also discloses a method for manufacturing this heat sink, solving the problem that traditional heat sinks cannot be used in extreme environments such as high temperature, high radiation, and strong corrosion.

[0005] However, current liquid cooling systems mostly rely on external distribution pipes or centralized water distribution with a fixed ratio for flow allocation. This makes it difficult to accurately adjust the flow based on the dynamic differences in heat load across different layers. Furthermore, they lack the capability to perform closed-loop coordination analysis of the entire supply-dissipation-return process using real-time measurement data. This results in over- or under-supply of cooling resources in some layers, increasing energy consumption and reducing overall heat dissipation efficiency. In addition, the lack of adaptive control mechanisms within each layer addresses temperature fluctuations, flow resistance changes, and return delays caused by variations in workload during operation further limits the level of refined management of tiered liquid cooling systems.

[0006] To address the above issues, there is an urgent need for a layered microchannel liquid cooling method and system based on magnetic interfaces. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a layered microchannel liquid cooling method and system based on magnetic interfaces. This solves the problem in multi-layer server structures where existing liquid cooling distribution paths cannot dynamically adjust the flow rate according to the differences in thermal load of each layer of equipment, resulting in uneven distribution of cooling resources and serious energy waste.

[0009] Technical solution

[0010] To achieve the above objectives, the present invention provides the following technical solution: a layered microchannel liquid cooling method and system based on a magnetic interface, comprising: S1, collecting temperature characteristic data and fluid flow rate and pressure data during the operation of each layer of the server, and preprocessing the collected temperature characteristic data and fluid flow rate and pressure data to construct a standardized heat load trend dataset; S2, based on the standardized heat load trend dataset, evaluating the coolant supply status of each layer in conjunction with the coolant flow rate, and dynamically adjusting the coolant supply rhythm of the corresponding layer based on the evaluation results; S3, based on the standardized heat load trend dataset, performing a demand analysis on the cooling demand intensity of each layer, and driving the corresponding layer branch pump operating frequency to change to match the coolant supply output based on the demand analysis results; S4, using the supply status evaluation results and demand analysis results as input, evaluating the coordination status of the coolant supply to heat dissipation to coolant return process of each layer, and dynamically adjusting the liquid cooling control strategy based on the evaluation results.

[0011] Further, the specific steps for collecting temperature characteristic data and fluid flow rate and pressure data during the operation of each layer of servers are as follows: Collect temperature characteristic data during the operation of each layer of servers. The temperature characteristic data includes: real-time surface temperature of the two-phase cold plates of each layer, inlet coolant temperature, outlet coolant temperature, and peak hot spot temperature. Simultaneously calculate and record the average temperature of the two-phase cold plates of each layer, the instantaneous temperature difference between the inlet and outlet, and the temperature standard deviation. Calculate the temperature difference between the inlet and outlet coolant temperatures at each moment to form a temperature difference sequence. Extract the difference between the maximum and minimum temperature values ​​from the temperature difference sequence, and record it as the temperature fluctuation intensity. Collect fluid flow rate and pressure data during the operation of each layer of servers. The fluid flow rate and pressure data includes: return flow rate at the condenser return port, inlet coolant flow rate at the two-phase cold plates of each layer, outlet coolant flow rate, inlet coolant pressure, and outlet coolant pressure. Simultaneously calculate and record the coolant pressure difference between the inlet and outlet.

[0012] Furthermore, the specific steps for preprocessing the collected temperature feature data and fluid velocity and pressure data to construct a standardized heat load trend dataset are as follows: For temperature feature data, the measurement deviations of different sensors are corrected using initial calibration coefficients; for fluid velocity and pressure data, extreme values ​​are denoised to eliminate abnormal data points caused by instantaneous fluctuations and acquisition jitter, and linear interpolation is uniformly used to fill in timestamp misalignment data caused by temporary packet loss; the temperature feature data and fluid velocity and pressure data that have completed synchronous correction are uniformly normalized to construct a standardized heat load trend dataset.

[0013] Furthermore, the specific steps for evaluating the coolant supply status of each layer based on the standardized heat load trend dataset and combined with the coolant flow rate are as follows: divide the coolant flow rate value at the inlet of the current layer's two-phase cold plate by the coolant pressure difference between the inlet and outlet to obtain the flow adaptation value; take the reciprocal of the average temperature of the current layer's two-phase cold plates to obtain the reciprocal of the average temperature; multiply the temperature fluctuation intensity by the corresponding temperature change modulation coefficient to obtain the equivalent thermal fluctuation intensity value; add one to the equivalent thermal fluctuation intensity value and take its reciprocal to obtain the thermal disturbance attenuation value; multiply the flow adaptation value, the reciprocal of the average temperature, and the thermal disturbance attenuation value, add one, and take the logarithm to obtain the coolant supply adequacy evaluation value.

[0014] Furthermore, the specific steps for dynamically adjusting the liquid supply rhythm of the corresponding layer based on the supply status assessment results are as follows: Real-time comparison of the liquid supply sufficiency assessment value and sufficiency assessment threshold of the current layer two-phase cold plate: When the liquid supply sufficiency assessment value is less than or equal to the sufficiency assessment threshold, the branch pump at the corresponding crossbeam is automatically driven to increase the extraction flow rate, increase the coolant flow rate distributed to the current layer microchannel through the main channel of the column, and simultaneously adjust the contact pressure of the current layer magnetic interface to maintain stable contact of the liquid metal contact, while triggering the top condenser to accelerate the switching of the return path; When the liquid supply sufficiency assessment value is greater than the sufficiency assessment threshold, the operating frequency of the current layer branch pump is reduced to reduce the inflow of coolant, and the return channel is switched to the buffer branch of the liquid storage tank, while maintaining the conduction state of the steam pipe to maintain the overall balance of circulation in each layer.

[0015] Furthermore, the specific steps for demand analysis of cooling demand intensity for each layer based on the standardized heat load trend dataset are as follows: Divide the collected coolant pressure difference by the outlet coolant velocity to obtain the actual flow resistance value of the two-phase cold plate microchannel; simultaneously record and calculate the average flow resistance value of each layer's two-phase cold plate; divide the difference between the peak hotspot temperature and the average temperature by the sum of the temperature standard deviation plus one to obtain the thermal drive intensity value; square the thermal drive intensity value, add one, and take the logarithm to obtain the thermal drive response value; calculate the absolute value of the difference between the actual flow resistance value and the average flow resistance value, multiply it by the flow resistance fluctuation adjustment factor, and then take the square root to obtain the flow resistance offset term; divide the pressure fluctuation compensation factor by the coolant pressure difference plus one to obtain the pressure difference compensation term; subtract the flow resistance offset term from the thermal drive response value and add it to the pressure difference compensation term to obtain the branch pump target frequency value.

[0016] Furthermore, the specific steps for driving the corresponding layer branch pump's operating frequency change based on the demand analysis results to match the liquid supply output are as follows: Based on the calculation result of the branch pump's target frequency value, the target frequency value of the branch pump is automatically sent to the diaphragm branch pump at the connection position between the crossbeam and the column, directly driving the diaphragm branch pump to perform pulse liquid supply according to the branch pump; as the liquid supply frequency is continuously updated, the dynamic change of the flow velocity in the coolant inlet pipe is collected in real time and used to automatically adjust the pressure balance of the return path to ensure that the gas-liquid separation process in the steam pipeline remains stable; during the operation of the branch pump, when the target frequency value of the branch pump continues to rise, the layer where the current two-phase cold plate is located is identified as a high heat load area, and the condenser is linked to open the auxiliary return liquid sub-channel to accelerate the top The system improves the condensation efficiency of the steam, reduces heat buildup, and records real-time temperature changes in the coolant outlet pipe. It also dynamically updates the base liquid supply rate of the main circulation pump based on these changes and redistributes the liquid flow rate from the storage tank. When the target frequency value of the branch pumps continues to decrease, the system shortens the duration of each branch pump's single start-up and automatically adjusts the instantaneous flow distribution ratio of each layer based on real-time fluid velocity and pressure data of the two-phase cold plates. Under the condition that the total flow rate of the main channel remains constant, the system adjusts the actual liquid supply of each layer, reducing the liquid supply to the two-phase cold plate layers where the instantaneous temperature difference between the inlet and outlet is less than the temperature difference threshold, and increasing the liquid supply to the two-phase cold plate layers where the instantaneous temperature difference between the inlet and outlet is greater than or equal to the temperature difference threshold, thus eliminating differences in coolant velocity between layers.

[0017] Furthermore, the specific steps for evaluating the coordination status of the process from liquid supply to heat dissipation to liquid return in each layer, using the supply status assessment results and demand analysis results as input, are as follows: Add one to the instantaneous temperature difference between the inlet and outlet of the two-phase cold plate, take the natural logarithm, and then multiply it by the square of the value after subtracting the target frequency value of the branch pump to obtain the coordination drive strength value; square the return liquid flow rate value at the condenser return port, divide it by one, add the inlet coolant pressure at the inlet of the two-phase cold plate, add it to the liquid supply sufficiency assessment value, add one, and take the natural logarithm to obtain the coordination absorption capacity value; divide the coordination drive strength value by the coordination absorption capacity value to obtain the closed-loop coordination assessment value.

[0018] Furthermore, the specific steps for dynamically adjusting the liquid cooling control strategy based on the evaluation results are as follows: Real-time comparison of the current closed-loop coordination evaluation value with the coordination evaluation threshold, which includes a first coordination threshold and a second coordination threshold: When the closed-loop coordination evaluation value is less than or equal to the second coordination threshold, the high-frequency emergency liquid supply mode of the two-phase cold plate branch pump is immediately triggered, and the main bypass branch of the condenser return liquid channel is simultaneously opened, forcibly guiding the liquid return to preferentially enter the crossbeam microchannel. At the same time, the cold plate number, temperature difference value, pressure difference value, and return liquid speed are recorded in the abnormal data buffer area and marked as a structural instability warning; when the closed-loop coordination evaluation value is greater than the second coordination threshold and less than or equal to the first coordination threshold... When the value is maintained, the single start-up duration of the branch pump is dynamically adjusted while maintaining the current target frequency of the branch pump. The distribution ratio of the basic liquid supply of the main circulation pump is also updated in conjunction with the adjustment, so that the actual flow rate is enhanced while maintaining stable operation. At the same time, the working status monitoring process of the two-phase cold plate is activated, and the changes in the operating rhythm are transmitted back to the main control module in real time. When the closed-loop coordination assessment value is greater than the first coordination threshold, the branch pump is switched to the low load regulation mode, the flow distribution ratio is reduced, and the effective cross-section of the condenser return liquid channel is reduced simultaneously to guide more steam heat flow to the adjacent two-phase cold plate. At the same time, the current two-phase cold plate is marked as the liquid supply oversaturation area, and the core data of the process from liquid supply to temperature rise to liquid return are recorded.

[0019] The second aspect of this invention provides a layered microchannel liquid cooling system based on a magnetic interface, comprising: a multi-layer heat load dynamic acquisition module, used to acquire temperature characteristic data and fluid flow rate and pressure data during the operation of each layer of servers, and to preprocess the acquired temperature characteristic data and fluid flow rate and pressure data to construct a standardized heat load trend dataset; a branch flow response evaluation module, used to evaluate the supply status of coolant in each layer based on the standardized heat load trend dataset and the coolant flow rate, and to dynamically adjust the supply rhythm of the corresponding layer based on the supply status evaluation results; a branch pump control and coordination module, used to perform demand analysis on the cooling demand intensity of each layer based on the standardized heat load trend dataset, and to drive the operating frequency of the corresponding layer branch pump to change to match the supply output based on the demand analysis results; and a closed-loop supply regulation feedback module, used to evaluate the coordination status of the process from supply to heat dissipation to return to liquid in each layer with the supply status evaluation results and demand analysis results as input, and to dynamically adjust the liquid cooling control strategy based on the evaluation results.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) The layered microchannel liquid cooling method and system based on magnetic interface achieves tool-free connection between multi-layer cold plates and liquid supply pipeline by adopting a magnetic liquid cooling interface that can be quickly connected and disassembled, thereby reducing the time cost of maintenance and replacement of parts.

[0023] (2) The layered microchannel liquid cooling method and system based on magnetic interface, by arranging diaphragm branch pumps at the connection between the beam and the column, and combining the real-time temperature, flow rate and pressure data of each layer, can realize dynamic liquid supply adjustment across layers and improve the accuracy of heat dissipation resource allocation.

[0024] (3) The layered microchannel liquid cooling method and system based on magnetic interface can respond in a timely manner to the risk of decreased evaporation efficiency or heat accumulation by evaluating the closed-loop coordination of the entire liquid supply-heat dissipation-liquid return chain, thus ensuring the thermal stability of the multi-layer server operation.

[0025] (4) The layered microchannel liquid cooling method and system based on magnetic interface reduces the phenomenon of over-supply or under-supply of coolant by combining the layered liquid supply adequacy assessment with target frequency control, thereby reducing energy consumption and extending equipment service life.

[0026] 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

[0027] Figure 1 This is a flowchart of the layered microchannel liquid method based on a magnetic interface according to the present invention;

[0028] Figure 2 This is a structural diagram of the layered microchannel liquid system based on a magnetic interface according to the present invention;

[0029] Figure 3 This is a line graph of the closed-loop coordination evaluation values ​​involved in this invention;

[0030] Figure 4 This is a detailed diagram of the gas-liquid separation involved in this invention;

[0031] Figure 5 This is a flowchart illustrating the operational process involved in this invention;

[0032] Figure 6 This is a simplified diagram of the device involved in the present invention;

[0033] Figure 7 This is a simplified diagram of the internal structure involved in the present invention.

[0034] In the diagram, 1 is the condenser; 2 is the liquid storage tank; 3 is the column; 4 is the two-phase cold plate; 5 is the chip; 6 is the coolant inlet pipe; 7 is the coolant outlet pipe; and 8 is the steam pipe. Detailed Implementation

[0035] 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.

[0036] Please see Figures 1-7 This invention provides a technical solution: a layered microchannel liquid cooling method and system based on a magnetic interface, comprising: S1, collecting temperature characteristic data and fluid flow rate and pressure data during the operation of each layer of servers, and preprocessing the collected temperature characteristic data and fluid flow rate and pressure data to construct a standardized heat load trend dataset; S2, based on the standardized heat load trend dataset, evaluating the coolant supply status of each layer in conjunction with the coolant flow rate, and dynamically adjusting the coolant supply rhythm of the corresponding layer based on the evaluation results; S3, based on the standardized heat load trend dataset, performing a demand analysis on the cooling demand intensity of each layer, and driving the corresponding layer branch pump operating frequency to change to match the coolant supply output based on the demand analysis results; S4, using the supply status evaluation results and demand analysis results as input, evaluating the coordination status of the coolant supply to heat dissipation to coolant return process of each layer, and dynamically adjusting the liquid cooling control strategy based on the evaluation results.

[0037] Specifically, the steps for collecting temperature characteristic data and fluid flow rate and pressure data during the operation of each layer of servers are as follows: Temperature characteristic data for each layer of servers during operation is collected. This includes: real-time surface temperature obtained by temperature sensors installed on the surface of the two-phase cold plates 4 on each layer; inlet and outlet coolant temperatures obtained by temperature probes built into the coolant inlet pipe 6 and coolant outlet pipe 7; and peak temperatures of hot spots obtained by a high-sensitivity sensor array deployed in the hot spot area. During the data collection process, the average temperature of each layer of two-phase cold plates 4, the instantaneous temperature difference between the inlet and outlet of the cold plates, and the temperature standard deviation are simultaneously calculated and recorded to reflect the uniformity and fluctuation characteristics of the temperature distribution.

[0038] Based on the continuous updating of temperature characteristic data, the instantaneous temperature difference between the inlet coolant temperature and the outlet coolant temperature at each moment is calculated, and the instantaneous temperature differences are arranged in chronological order to form a temperature difference sequence. In the temperature difference sequence, the difference between the maximum temperature value and the minimum temperature value is extracted, and the difference is recorded as the temperature fluctuation intensity, which is used to characterize the magnitude of heat load change in the cooling circuit during the operating cycle.

[0039] Fluid velocity and pressure data are collected during the operation of each server layer. The fluid velocity and pressure data include: the return flow velocity value obtained by the flow velocity sensor installed at the condenser return port; the inlet and outlet coolant flow velocity values ​​obtained by the flow meter on the coolant inlet pipe 6 and coolant outlet pipe 7 of each layer's two-phase cold plate 4; and the inlet and outlet coolant pressures obtained by the pressure sensors on the coolant inlet pipe 6 and coolant outlet pipe 7. The coolant pressure difference between the inlet and outlet is calculated and recorded simultaneously with the data collection, providing a direct basis for subsequent flow resistance analysis and coolant supply adjustment.

[0040] In this implementation scheme, during the operation of the server layered liquid cooling, the temperature distribution and coolant flow characteristics of each layer of two-phase cold plate 4 are acquired and quantified in real time. By continuously collecting temperature characteristic data and fluid flow rate and pressure data, and extracting the temperature fluctuation intensity and coolant pressure difference, accurate input basis is provided for subsequent liquid supply adequacy assessment, flow resistance analysis and branch pump frequency control, thereby supporting fine adjustment and closed-loop stable control under dynamic heat load conditions.

[0041] Specifically, the preprocessing of collected temperature characteristic data and fluid velocity and pressure data to construct a standardized heat load trend dataset involves the following steps: For temperature characteristic data, measurement deviations generated by different sensors during long-term operation are corrected using initial calibration coefficients to ensure that temperature readings from the cold plate surface, inlet and outlet positions, and hot spot detection points are comparable under the same benchmark; For fluid velocity and pressure data, sudden spikes and discrete outliers are denoised after data acquisition to eliminate invalid data points caused by instantaneous flow field disturbances, sensor jitter, and electronic interference; When timestamps are missing due to network transmission and packet loss, linear interpolation is used to fill the gaps on the time axis to maintain the continuity and time alignment of each data sequence. After the above corrections and filling, all temperature characteristic data and fluid velocity and pressure data are uniformly normalized within the same numerical range to convert them into dimensionless standardized values, thereby constructing a standardized heat load trend dataset. This provides consistent input for subsequent liquid supply adequacy assessment, target frequency calculation, and closed-loop coordination analysis.

[0042] In this implementation scheme, the raw temperature characteristic data and fluid velocity and pressure data are preprocessed to ensure uniformity, continuity, and comparability. Initial calibration corrects sensor deviations, eliminating systematic errors between different measuring points and ensuring temperature and fluid information are within the same benchmark framework. Noise reduction and outlier removal effectively shield invalid points caused by instantaneous fluctuations, jitter interference, and data loss, guaranteeing data stability and accuracy. Linear interpolation of timestamps maintains the alignment of multi-source data in the time dimension, enabling different data sequences to correspond synchronously within the same period. Finally, unified normalization maps temperature and fluid pressure and velocity to the same numerical scale, forming a standardized heat load trend dataset. This heat load trend dataset serves as the foundational input for subsequent liquid supply adequacy assessment, target frequency control, and closed-loop coordination analysis, ensuring the control algorithm maintains consistency in real-time performance and accuracy, effectively supporting the dynamic adjustment and stable operation of the stratified liquid cooling method.

[0043] Specifically, based on a standardized heat load trend dataset, the supply status assessment of each layer's coolant supply status, combined with coolant flow rate, involves the following steps: dividing the current layer's two-phase cold plate 4 inlet coolant flow rate by the coolant pressure difference between the inlet and outlet to obtain the flow adaptation value; taking the reciprocal of the current layer's two-phase cold plate 4 temperature average to obtain the reciprocal of the temperature average; multiplying the temperature fluctuation intensity by the corresponding temperature change modulation coefficient to obtain the equivalent thermal fluctuation intensity value; adding one to the equivalent thermal fluctuation intensity value and taking its reciprocal to obtain the thermal disturbance attenuation value; and multiplying the flow adaptation value, the reciprocal of the temperature average, and the thermal disturbance attenuation value together, adding one, and taking the logarithm to obtain the coolant supply adequacy assessment value.

[0044] The formula for calculating the fluid supply adequacy assessment value is:

[0045] ;

[0046] In the formula, It represents the inlet coolant flow rate of the two-phase cold plate 4 in the current layer, which is used to characterize the liquid supply capacity of the layer per unit time. It is a direct representation of the liquid cooling distribution state and is derived from the miniature flow rate sensor at the branch port of the main channel of column 3. It represents the coolant pressure difference between the inlet and outlet of the two-phase cold plate in the current layer, which is used to reflect the flow resistance of the microchannel in this layer. It is an important reference value for the actual flow capacity of the coolant and is derived from the miniature differential pressure sensor installed at the inlet and outlet of the two-phase cold plate 4. This represents the average temperature of the current two-phase cold plate 4, which is used to quantify the current thermal accumulation level and cooling difficulty of the structure. It is an important parameter reflecting the heat dissipation pressure of the two-phase cold plate 4 and is derived from the temperature sensors deployed on the surface of the two-phase cold plate in this layer. It represents the intensity of temperature fluctuation of the current layer two-phase cold plate 4 within the sliding time window, which is used to measure the trend of heat load change of the layer and is a sensitive factor for judging the necessity of dynamic adjustment of liquid supply. It is derived from the calculation of continuous data sequence changes of the surface temperature sensor of the two-phase cold plate 4. The temperature change modulation coefficient, ranging from 0.5 to 2, is used to dynamically adjust the response weight of temperature changes to the degree of interference in the cooling process. It is derived from the stability performance index of the temperature curve under the current two-phase cold plate thermal load state, primarily based on the data evolution characteristics of the surface temperature sensors of each layer of cold plates. The specific calculation process is as follows: First, the temperature change sequence of each layer of two-phase cold plate 4 within the sliding time window is extracted, and its slope, fluctuation frequency, and short-term peak change trend are analyzed to construct a current fluctuation state profile. Then, the temperature curve benchmark template formed during historical steady-state operation is called, including fluctuation amplitude, oscillation rhythm, and temperature difference distribution pattern, to compare the deviation of the current data. Finally, a comprehensive evaluation is performed by combining the deviation amplitude, disturbance frequency, and synchronization mismatch between multiple layers of cold plates to generate the temperature change modulation coefficient. When the current temperature change exhibits drastic jumps and high-frequency instability, the temperature change modulation coefficient is increased to enhance the response sensitivity to temperature disturbances and strengthen the adaptability of stratified cooling. If the temperature fluctuation is stable and the trend is predictable, the temperature change modulation coefficient is decreased to avoid interfering with the normal adjustment rhythm, thereby maintaining the temperature stability adjustment accuracy and anti-disturbance capability of the cooling channel in a dynamic environment.

[0047] In this implementation scheme, a dynamic index is constructed to measure the rationality of the coolant distribution state between multiple layers of cold plates. This formula comprehensively considers the coolant flow rate per unit time, the pressure difference between the cold plate inlet and outlet, the average temperature of the cold plate surface, the intensity of temperature fluctuations, and a control factor for adjustment sensitivity. It accurately reflects the matching degree between the current layer's coolant supply capacity and the thermal load response. In practical applications, if the coolant supply sufficiency assessment value of a certain cold plate is significantly lower than that of other layers, it indicates a risk of insufficient cooling; conversely, a high value may indicate over-cooling. The coolant supply sufficiency assessment value not only serves as a direct feedback parameter for dynamically adjusting the coolant diversion ratio and channel opening, but also enhances the robustness to short-term thermal disturbances by introducing a fluctuation modulation factor, avoiding false triggering of control logic. This enables adaptive optimization control under high thermal load and frequent fluctuation scenarios, significantly improving response efficiency, stability, and adjustment accuracy.

[0048] Specifically, the steps for dynamically adjusting the liquid supply rhythm of the corresponding layer based on the supply status assessment results are as follows:

[0049] The system compares the current liquid supply sufficiency assessment value of the two-phase cold plate 4 with the sufficiency assessment threshold in real time. By continuously monitoring the matching degree between the liquid supply status of different cold plate layers and the actual cooling demand, it dynamically judges whether there is insufficient or excessive coolant distribution. When the liquid supply sufficiency assessment value is less than or equal to the sufficiency assessment threshold, it indicates that there is a risk of insufficient coolant supply to the two-phase cold plate 4. The branch pump set in the crossbeam structure is immediately and automatically driven to improve the extraction capacity of coolant at the inlet of the cold plate of that layer. This helps to increase the liquid supply from the main channel of column 3 to the microchannel of the target layer, thereby improving the cooling capacity of that layer. At the same time, in order to ensure the continuity and adhesion of the liquid metal cooling medium with high thermal conductivity at the magnetic attraction interface, the control module synchronously adjusts the contact pressure of the magnetic attraction interface of the current layer to enhance the stability of the interface contact. In addition, in order to cooperate with the increase of coolant flow, the system triggers the accelerated switching logic of the return liquid path in the top condenser 1 to accelerate the heat recovery and condensation cooling efficiency in the high-temperature area.

[0050] When the coolant supply adequacy assessment value exceeds the adequacy assessment threshold, it indicates that the coolant supply to the current layer has met the demand. The operating frequency of the branch pump is actively reduced to decrease the amount of coolant injected, thus avoiding energy waste and local overcooling. At the same time, the control logic switches the coolant return path to the buffer branch connected to the liquid storage tank 2, making the return process more gentle and reducing overall pressure fluctuations. In addition, to maintain the smoothness and balance of the entire multi-layer circulation path, the steam pipe 8 remains open to ensure that the steam flow after gas-liquid separation is not blocked, further ensuring the stable conduction of heat distribution and the synergistic heat exchange efficiency between the upper and lower layers.

[0051] This implementation scheme achieves dynamic closed-loop control and distributed adjustment of the coolant supply status of each cold plate layer in the layered liquid cooling structure, ensuring precise and on-demand distribution of coolant among different layers, thereby improving the overall heat exchange efficiency and operational stability. By comparing the coolant supply sufficiency assessment value and sufficiency assessment threshold of each cold plate layer in real time, it can promptly identify whether the coolant supply is insufficient or redundant. When the coolant supply is insufficient, the local pumping rate is automatically increased, the adhesion of the liquid metal contacts is enhanced, and the condensate return path is optimized to quickly replenish the cooling capacity; while when the coolant supply is excessive, the flow is actively limited and the liquid is guided back to the buffer channel to maintain overall thermal balance and flow stability. This mechanism significantly enhances the adaptive response capability of the liquid cooling device to load changes, avoiding uneven heating and local overheating problems.

[0052] Specifically, based on the standardized heat load trend dataset, the specific steps for demand analysis of cooling demand intensity for each layer are as follows: Divide the collected coolant pressure difference by the outlet coolant velocity to obtain the actual flow resistance value of the microchannel of the two-phase cold plate 4, and simultaneously record and calculate the average flow resistance value of the two-phase cold plate 4 for each layer; Divide the difference between the peak temperature of the hot spot and the average temperature by the sum of the temperature standard deviation plus one to obtain the thermal drive intensity value; Square the thermal drive intensity value, add one, and take the logarithm to obtain the thermal drive response value; Calculate the absolute value of the difference between the actual flow resistance value and the average flow resistance value, multiply it by the flow resistance fluctuation adjustment factor, and then take the square root to obtain the flow resistance offset term; Divide the pressure fluctuation compensation factor by the value of the coolant pressure difference plus one to obtain the pressure difference compensation term; Subtract the flow resistance offset term from the thermal drive response value and add it to the pressure difference compensation term to obtain the target frequency value of the branch pump.

[0053] The formula for calculating the target frequency value of the branch pump is:

[0054] ;

[0055] In the formula, This represents the peak temperature of the hotspots monitored by the server, used to quantify the most extreme heat dissipation pressure. It is an important input parameter for assessing the degree of cooling stress and is extracted from the maximum value in the array of thermal sensors. This represents the average temperature of the current two-phase cold plate 4, which is used to quantify the current thermal accumulation level and cooling difficulty of the structure. It is an important parameter reflecting the heat dissipation pressure of the two-phase cold plate and is derived from the temperature sensors installed on the surface of the two-phase cold plate 4. The standard deviation of temperature in region 4 of the two-phase cold plate is used to quantify the dispersion of temperature distribution. It is an important indicator for judging the concentration of local hot spots and is derived from the square root of the statistical variance of multiple temperature acquisition points within the sliding window. This represents the actual flow resistance value within the microchannel of the cold plate, reflecting the magnitude of the resistance encountered by the liquid when passing through the channel. It is an important variable for adjusting the flow adaptability and is derived from the calculation results of the pressure difference between the inlet and outlet of this layer and the flow rate ratio. It represents the average flow resistance value, which is used as a reference for flow resistance and is an important indicator for judging the current flow deviation intensity. It is derived from the moving average of the actual flow resistance value. It represents the coolant pressure difference between the inlet and outlet of the two-phase cold plate 4 in the current layer, which is used to reflect the flow resistance of the microchannel in this layer. It is an important reference value for the actual flow capacity of the coolant and comes from the miniature differential pressure sensors installed at the inlet and outlet of the two-phase cold plate 4. The pressure fluctuation compensation factor, ranging from 0.8 to 1.5, is used to compensate for pressure response fluctuations caused by coolant flow disturbances resulting from the alternation of hot and cold liquids within the channel. It originates from the relationship between the peak difference and duration of the coolant pressure fluctuation amplitude between the inlet and outlet of the two-phase plates. Specifically, the calculation first extracts the micro-pressure sensor data sequence of the two-phase cold plates within the sliding time window, calculating its local peak value, trough value, and period frequency to characterize the instantaneous pressure fluctuation pattern. Then, it retrieves typical pressure oscillation patterns under historical stable cooling conditions, including low-frequency stable intervals, high-frequency burst segments, and edge transition regions. Next, it dynamically aligns the current fluctuation with historical patterns, comparing their amplitude gradient, periodic trend, and disturbance symmetry, and considering environmental factors such as the complexity of the cooling channel structure and the branch flow deviation coefficient to generate a comprehensive compensation coefficient. If the current channel pressure fluctuations are frequent, severe, and long-lasting, the pressure fluctuation compensation factor increases to enhance the branch pump frequency's ability to regulate impact loads. If the fluctuation amplitude is low, rhythmic, and tends towards a steady state, the pressure fluctuation compensation factor decreases to reduce ineffective compensation and improve the tracking accuracy and energy efficiency of the coolant supply control for a stable state. The flow resistance fluctuation adjustment factor, ranging from 0.5 to 1.2, controls the dynamic coupling effect of micro-flow resistance changes on the target frequency calculation value. It originates from the deviation and rate of change between the current actual flow resistance of the microchannel group and the historical sliding window mean. In the specific calculation, firstly, a historical flow resistance baseline sequence for each cooling branch is constructed, extracting the steady-state mean, maximum gradient, and fluctuation boundary as a reference template. Then, the real-time flow resistance value and its sliding rate of change of the current channel are collected, and the degree of matching between it and the reference template in terms of mean difference, variation rate, and peak abrupt change point is evaluated. Combining the current channel load distribution, cold plate heat flux density, and upstream coolant supply capacity, adjustment weights are assigned to form the flow resistance adjustment coupling factor for branch pump frequency adjustment. When the flow resistance changes drastically, the trend is unstable, and it significantly deviates from the historical mean, the flow resistance fluctuation adjustment factor increases to improve adjustment sensitivity and quickly respond to unstable situations such as microchannel blockage and structural abrupt changes. If the current flow resistance change is gradual, with good predictability and consistency, the flow resistance fluctuation adjustment factor approaches 1 to maintain adjustment stability and avoid frequent fluctuations interfering with the main circulation.

[0056] In this implementation scheme, the core function of achieving precise control of layered liquid supply regulation in the magnetic microchannel liquid cooling structure is to comprehensively assess the heat load concentration, cooling channel flow resistance characteristics, and dynamic fluctuation disturbance level of the current cold plate layer based on multi-source state parameters, and output precise branch pump operating frequency values ​​accordingly to match the actual cooling needs of the current layer. The formula first characterizes the concentration and deviation of the heat load distribution by combining the difference between the peak temperature of the server hotspot and the average temperature of the corresponding cold plate area with the temperature variance information of that area, serving as an important basis for assessing potential overheating risks. Then, it introduces the deviation of the channel flow resistance value from its historical moving average to measure the obstruction state of the current channel in coolant transport, and combines this with the hydraulic pressure difference between the cold plate inlet and outlet to reflect the microscopic flow resistance inside the channel. Furthermore, two dynamic adjustment factors are embedded: a pressure fluctuation compensation factor reflecting the severity of cooling pressure disturbance and a flow resistance fluctuation adjustment factor assessing the dynamic stability of the channel flow resistance, enhancing the formula's ability to perceive and adapt to unsteady operating scenarios. Overall, this formula establishes a branch pump frequency modulation modeling path that integrates heat, pressure, resistance, and disturbance, which can effectively improve the liquid supply response flexibility of the liquid cooling channel to local abnormal conditions, thereby ensuring the dynamic steady-state operation of the cooling balance of multi-layer cold plates and the overall thermal management efficiency.

[0057] Specifically, the steps for driving the corresponding layer branch pump's operating frequency variation based on the demand analysis results to match the liquid supply output are as follows: Based on the calculation results of the branch pump's target frequency value, the target frequency value of the branch pump is automatically sent to the diaphragm branch pump at the connection position between the crossbeam and the column 3, directly driving the diaphragm branch pump to perform pulse-type liquid supply control according to the branch pump's target frequency, ensuring that the liquid supply response of the current cold plate layer is highly matched with the heat load state; As the liquid supply frequency is continuously updated, the instantaneous flow rate value of the coolant inlet pipe 6 is collected in real time by the high-speed micro flow rate sensor, constructing a dynamic change sequence of the flow rate in the inlet channel, and using this as a reference parameter, automatically adjusting the opening of the throttling component in the return path to maintain the pressure gradient balance in the return channel, ensuring that the gas-liquid separation process in the steam pipe 8 maintains a stable stratified distribution under different load conditions, and avoiding the decrease in cooling efficiency caused by gas-liquid misalignment.

[0058] During the operation of the branch pump, when the target frequency value of the branch pump continues to rise, the layer where the current two-phase cold plate 4 is located is identified as a high heat load area. The control loop of the top condenser 1 is linked to actively open its built-in auxiliary liquid return sub-channel. By shortening the condensation path delay time, the steam condensation efficiency is improved, and the rate of condensate fall from the top is further accelerated to alleviate the local heat accumulation trend. At the same time, the high-response temperature sensor installed in the coolant outlet pipe 7 is used to record the real-time outlet temperature change, forming a continuous temperature fluctuation sequence. Based on this, the basic liquid supply rate of the main circulation pump is dynamically updated to build a dynamic adjustment benchmark for the current cooling path. In addition, based on the real-time feedback of the main circulation pump's adjustment output, the flow ratio of liquid in the liquid storage tank 2 in the multi-branch path will be dynamically redistributed to form an overall scheduling strategy for stratified liquid supply.

[0059] When the target frequency value of the branch pump continues to decrease, it identifies that the current layer's thermal load is approaching a stable and low-load state, automatically shortens the single start-up duration of the branch pump, prevents excessive fluid supply from causing local hydraulic fluctuations, and automatically adjusts the instantaneous flow distribution ratio of each layer based on the real-time fluid velocity and pressure data of the two-phase cold plates 4 in each layer: under the condition that the total flow rate of the main channel remains unchanged, according to the relationship between the instantaneous temperature difference at the inlet and outlet and the temperature difference threshold, it reduces the fluid supply of the two-phase cold plate layer where the instantaneous temperature difference at the inlet and outlet is less than the temperature difference threshold, while increasing the fluid supply of the two-phase cold plate layer where the instantaneous temperature difference at the inlet and outlet is greater than or equal to the temperature difference threshold, thereby dynamically eliminating the difference in coolant flow rate between layers and improving the overall fluid supply balance and thermal response coordination between multiple cooling paths.

[0060] This implementation scheme achieves dynamic liquid supply regulation and multi-layer heat load balance management based on the target frequency value of the branch pumps. By using the calculated target frequency value of the branch pumps to control the pulse liquid supply behavior of the diaphragm branch pumps in the layered cooling circuit, the flow rate of the coolant is closely linked to changes in heat load. Simultaneously, during the liquid supply frequency update process, changes in the flow rate within the coolant inlet channel are sensed in real time, thereby adjusting the pressure in the return path to ensure the stability of the vapor-liquid separation process. Furthermore, by identifying the trend of the target frequency value, this step can also help identify local high heat load areas, thereby activating the top condenser's auxiliary return mechanism, improving condensation efficiency, and alleviating local heat accumulation. Conversely, when the heat load decreases, by shortening the liquid supply time and adjusting the actual liquid supply ratio of each layer, the allocation of cooling resources among multiple layers is dynamically optimized, achieving a spatially balanced scheduling of coolant flow rate, ultimately improving the thermal control accuracy and operational stability of the entire liquid cooling structure.

[0061] Specifically, using the supply status assessment results and demand analysis results as inputs, the coordination status of the process from liquid supply to heat dissipation to liquid return in each layer is assessed through the following steps: The instantaneous temperature difference between the inlet and outlet of the two-phase cold plate 4 is incremented by one and its natural logarithm is taken. This logarithm is then multiplied by the square of the value obtained by subtracting the target frequency value of the branch pump to obtain the coordination drive strength value. The square of the return flow rate at the return port of condenser 1 is divided by one and the inlet coolant pressure at the inlet of the two-phase cold plate 4 is added. This square is then added to the liquid supply sufficiency assessment value and its natural logarithm is taken to obtain the coordination absorption capacity value. The coordination drive strength value is divided by the coordination absorption capacity value to obtain the closed-loop coordination assessment value.

[0062] The formula for calculating the closed-loop coordination assessment value is:

[0063] ;

[0064] In the formula, This indicates the assessment value for fluid supply adequacy. Indicates the target frequency value of the branch pump; It represents the instantaneous temperature difference between the inlet and outlet of the two-phase cold plate, used to measure whether there is significant heat accumulation in the cold plate at this moment. It is derived from the difference in real-time readings of two temperature sensors at the inlet and outlet of the cold plate. The return liquid flow rate value at the condenser return liquid port is used to determine whether the steam can be successfully condensed and returned. It is derived from the liquid flow rate sensor at the top return liquid channel. This indicates the inlet coolant pressure at the inlet of the two-phase cold plate 4, used to identify whether there is blockage or abnormal flow resistance in the channel, and is derived from the pressure sensor reading at the inlet end.

[0065] In this implementation example, Example 1 has the following instantaneous temperature difference between inlet and outlet set to 0.65, a target frequency of 0.21 for the branch pump, a fluid supply adequacy assessment value of 1.02, a return flow rate of 0.31, and an inlet coolant pressure of 0.42; Example 2 has the following instantaneous temperature difference between inlet and outlet set to 0.72, a target frequency of 0.16 for the branch pump, a fluid supply adequacy assessment value of 1.08, a return flow rate of 0.35, and an inlet coolant pressure of 0.40; Example 3 has the following instantaneous temperature difference between inlet and outlet set to 0.78, a target frequency of 0.13 for the branch pump, a fluid supply adequacy assessment value of 1.15, a return flow rate of 0.38, and an inlet coolant pressure of 0.37; Example 4 has the following instantaneous temperature difference between inlet and outlet set to 0.69 and a target frequency of 0.18 for the branch pump. The following examples illustrate the coolant supply adequacy assessment values: Example 5: Inlet / outlet instantaneous temperature difference set at 0.88, branch pump target frequency at 0.11, coolant supply adequacy assessment value at 1.11, return flow rate at 0.36, inlet coolant pressure at 0.36; Example 6: Inlet / outlet instantaneous temperature difference set at 0.95, branch pump target frequency at 0.10, coolant supply adequacy assessment value at 1.20, return flow rate at 0.39, inlet coolant pressure at 0.35; Example 7: Inlet / outlet instantaneous temperature difference set at 0.81, branch pump target frequency at 0.15, coolant supply adequacy assessment value at 1.09, return flow rate at 0.34, inlet coolant pressure at 0.38. The closed-loop coordination assessment values ​​for each example are calculated and are shown in Table 1.

[0066] Table 1. Closed-Loop Coordination Evaluation Values

[0067]

[0068] like Figure 3 The table shown is a data table of closed-loop coordination evaluation values ​​provided in this application example. (See Table 1 and...) Figure 3It can be seen that among the coolant distribution control tasks in this batch, Example 6 has the highest closed-loop coordination evaluation value, indicating that its dynamic coupling relationship among the instantaneous temperature difference between inlet and outlet, the target frequency of the branch pump, the sufficiency of coolant supply, the return flow rate, and the inlet coolant pressure is the most stable. This demonstrates that the cooling control strategy has a fast response to heat load, accurate flow distribution, smooth return path, and the most coordinated overall regulation, making it suitable as an ideal benchmark mode for coolant supply regulation. Examples 5 and 7 are also relatively high in coordination evaluation value, indicating that these two control parameter configurations also have strong energy efficiency synergy capabilities, and can maintain stable operation and precise regulation of the cooling system in actual operation. In contrast, Example 1 has the lowest closed-loop coordination evaluation value. Although the sufficiency of coolant supply and the inlet pressure are at an intermediate level, the branch pump frequency is high while the heat exchange temperature difference is relatively small, resulting in insufficient cooling response per unit frequency increase. This indicates that the current control configuration has a certain lag in energy response and fluid regulation, and is not suitable as a primary control strategy. Example 4 also has a relatively low value, indicating that its regulation coordination needs to be optimized. In summary, the closed-loop coordination evaluation value reflects the degree of nonlinear coupling and real-time adjustment capability among the variables. A higher evaluation value indicates that the parameter configuration can achieve precise coordinated control between liquid supply and return, temperature difference, and frequency without introducing additional energy consumption, which is beneficial for improving the dynamic adaptability and response sensitivity of the entire cooling process. The line graph of the closed-loop coordination evaluation value clearly reveals the coordination differences of the seven examples under the current configuration, providing quantitative support for subsequent cooling path optimization.

[0069] Specifically, the steps for dynamically adjusting the liquid cooling control strategy based on the evaluation results are as follows: During the intelligent regulation of the coolant, the current closed-loop coordination evaluation value is compared with the coordination evaluation threshold in real time to dynamically determine whether the coolant supply scheduling strategy needs to be adjusted. The coordination evaluation threshold consists of a first coordination threshold and a second coordination threshold, which correspond to the judgment boundaries of the normal deviation adjustment zone and the abnormal destability warning zone, respectively.

[0070] When the closed-loop coordination assessment value is lower than or equal to the second coordination threshold, the state is immediately judged as having extremely weak liquid supply coordination capability, which may lead to local cooling failure and structural thermal destability. At this time, the high-frequency liquid supply mode of the two-phase cold plate 4 branch pump is urgently triggered to rapidly enhance the cooling intensity, and the main bypass branch of the condenser 1 return liquid channel is opened simultaneously. The coolant is forcibly guided to flow through the crossbeam microchannel through the bypass return liquid method to achieve rapid temperature drop. At the same time, the number of the two-phase cold plate 4, the corresponding temperature difference value, pressure difference value and return liquid speed are written into the abnormal data cache area and a structural destability warning label is assigned for reference in subsequent judgment and maintenance intervention.

[0071] When the closed-loop coordination assessment value is greater than the second coordination threshold and less than or equal to the first coordination threshold, it is judged as a slight deviation but not yet destabilized. At this time, instead of directly changing the branch pump frequency, the single start-up duration of the branch pump is dynamically adjusted to improve the coordination by lengthening and shortening the liquid supply cycle. At the same time, the basic liquid supply ratio of the main circulation pump is adjusted in conjunction to enhance the actual liquid supply flow rate on the basis of stable operation. In addition, in this state, the real-time monitoring process of the target two-phase cold plate 4 is automatically activated to monitor its cooling rhythm and load response, and these status change information are sent back to the main control module to prepare for subsequent predictive regulation.

[0072] When the closed-loop coordination assessment value exceeds the first coordination threshold, the region is identified as an oversupply and oversaturated cooling state, requiring proactive reduction of redundant energy supply. At this time, the branch pump is switched to a low-load regulation mode, and the current flow distribution ratio is reduced to avoid energy waste. Simultaneously, the effective cross-section of the condenser 1 return channel is synchronously controlled to adjust the liquid phase return pressure, guiding the steam heat flow to be automatically distributed to the adjacent two-phase cold plate 4, expanding the heat load distribution area, thereby alleviating the local saturation problem. This high-assessment state will be marked as an oversaturated supply region, and its core control data from the start of supply, the temperature rise stage to the end of return will be automatically recorded for future supply strategies and parameter updates.

[0073] like Figure 4 The diagram shown is a detailed illustration of gas-liquid separation provided in this application example. It illustrates the flow and phase change process of two-phase coolant in a cooling structure, highlighting the heat transfer path and phase interface distribution characteristics of the liquid coolant as it transforms into vapor after passing through the cold plate structure. The left side of the diagram represents the liquid coolant channel, where the liquid is fed into the two-phase coolant region on the right. This region corresponds to the internal structure of the cold plate, where the liquid coolant absorbs heat and undergoes a phase change upon contact with the heat source, gradually transforming into a gaseous state. The liquid state in the middle and the vapor indicated by the arrow in the upper right corner illustrate this process: the coolant is heated into vapor after flowing through the cold plate and rises along the vapor pipe, thus achieving heat removal and transfer. The diagram emphasizes the spatial distribution and migration trend of the liquid-gas two-phase flow within the cold plate, highlighting the following key points: liquid-gas stratification, where liquid and vapor form a phase interface within the cold plate, with vapor typically escaping upwards while the liquid remains in the lower region; the phase change cooling path, where heat is carried away through two-phase flow, resulting in higher efficiency compared to single-phase liquid cooling and facilitating rapid heat dissipation in high-heat-flux areas; and the flow diagram within the structure, illustrating the complete heat transfer path from liquid inlet to vaporization and then discharge through the vapor channel, reflecting the cold plate structural design's support mechanism for phase change cooling. Overall, this diagram provides an intuitive visualization of the liquid heat absorption—vapor escape—heat transfer mechanism of the two-phase cold plate, aiding in the explanation of the key functional layout and operational logic of related heat dissipation devices in the cooling system.

[0074] like Figure 5 The diagram shown illustrates the operational process of a typical two-phase liquid cooling cycle structure, provided as an example of this application. It primarily consists of core components including a storage tank, condenser, coolant pipes, gas-liquid separator, cold plate layer, and two-phase liquid pipes, enabling efficient heat dissipation and heat flow control. Starting in the storage tank, coolant enters the cold plate layer through the coolant pipes. Upon contact with the high heat source area, part of the liquid coolant absorbs heat and vaporizes, forming a two-phase fluid containing both liquid and gaseous components. This two-phase fluid flows along the two-phase liquid pipes into the gas-liquid separator, where it undergoes phase separation. The liquid portion flows back to the storage tank for closed-loop recycling, while the gaseous portion is guided to the condenser, where it is cooled and converted back into liquid, also flowing back to the storage tank to replenish the cycle. This structure, through continuous and efficient heat exchange and gas-liquid circulation, maintains a dynamic balance in thermal management, making it particularly suitable for cooling electronic modules with high heat flux and significant thermal fluctuations. The diagram visually illustrates the fluid transfer paths and collaborative logic between the various functional components in a liquid cooling system, emphasizing the closed-loop characteristics and energy efficiency advantages of two-phase coolants in the heat absorption, transfer, and recovery processes.

[0075] like Figure 6 The diagram shown is a simplified representation of the device provided in this application, illustrating the basic components of a condensation and storage device based on a column-supported structure. The condenser 1, located at the top of the device, cools and liquefies high-temperature steam into a coolant, serving as the core heat exchange component of the condensation loop. The storage tank 2, located at the bottom, stores the liquid cooling medium flowing from the condenser, acting as a condensate collector and buffer. The column 3 provides structural support, connecting the condenser 1 and the storage tank 2, maintaining the overall stability and spatial layout of the device. The overall structure adopts a vertical frame design, with the upper condenser 1 fixed to the lower storage tank 2 via the column 3, allowing the condensate to flow naturally into the storage area under gravity, thus simplifying the condensation loop and improving heat exchange efficiency. This diagram clearly reflects the functional zoning and fluid flow path of the device in the vertical direction, providing a basic structural reference for subsequent cooling system integration and module installation.

[0076] like Figure 7The diagram shown is a simplified internal structure diagram provided in this application example, illustrating the internal component arrangement of the cooling structure. This structure adopts a three-dimensional, layered arrangement, with each layer containing a two-phase cold plate 4 on which several chips 5 are mounted. The chips 5 are tightly fitted to the two-phase cold plate 4 to achieve efficient heat conduction. Coolant enters each layer of the cold plate through the coolant inlet pipe 6, absorbs heat below the chips 5, and is then led out through the coolant outlet pipe 7, achieving the circulation of the liquid coolant. The heated coolant partially vaporizes within the cold plate, forming steam, which rises and is discharged through the vertically arranged steam pipe 8 on the right side, effectively completing the phase change transfer of heat from the chips 5 to the steam. This diagram visually demonstrates the three-dimensional heat dissipation structure of multi-layer cold plate series cooling and the coordination of liquid and gas separation channels, helping to understand the thermal management logic and fluid path planning of the two-phase cooling device in a multi-chip high-density arrangement scenario.

[0077] In this implementation plan, the current closed-loop coordination assessment value is compared with the coordination assessment threshold in real time, triggering corresponding liquid supply regulation behavior to achieve dynamic adaptation and precise adjustment of the operating state. By introducing two judgment boundaries, a first coordination threshold and a second coordination threshold, the closed-loop coordination can be divided into three typical operating conditions: low coordination state, critical coordination state, and oversaturation state, and a differentiated response strategy is executed for each state. Specifically, when the assessment value is low, the system quickly switches to a high-frequency liquid supply and main-bypass coordinated mode to prevent thermal instability; in the intermediate range, the pump start-up time and flow ratio are dynamically optimized to maintain stability; when the assessment value is high, redundant liquid supply is reduced and the heat flow path is reconstructed to avoid energy waste. Overall, this step ensures adaptive regulation capability and precise local liquid supply scheduling capability under changing operating conditions, improves the working efficiency of the cold plate, suppresses the risk of heat accumulation, and provides basic data support for subsequent operating state prediction and fault early warning.

[0078] The second aspect of this invention provides a layered microchannel liquid cooling system based on a magnetic interface, comprising: a multi-layered dynamic heat load acquisition module, used to acquire temperature characteristic data and fluid flow rate and pressure data during the operation of each layer of the server, and to preprocess the acquired temperature characteristic data and fluid flow rate and pressure data to construct a standardized heat load trend dataset; wherein, the temperature characteristic data includes the instantaneous temperature of the chip surface, the rate of change of interlayer temperature difference, and the heat dissipation delay feedback value, and the fluid flow rate and pressure data covers the instantaneous flow rate of the liquid inlet, the pressure drop gradient, and the pressure fluctuation amplitude of the branch pipes, and forms a multi-dimensional heat load evolution sequence with time continuity through normalization processing, which is used by subsequent modules to dynamically determine the cooling load status.

[0079] The branch flow response assessment module is used to assess the coolant supply status of each layer based on a standardized heat load trend dataset and coolant flow rate, and dynamically adjust the coolant supply rhythm of the corresponding layer based on the assessment results. This assessment process analyzes the matching degree between the actual coolant supply of each layer and its heat load intensity to identify whether the flow supply is sufficient, whether there is a lag response and local overcooling problems. On this basis, the opening and closing frequency and flow pulse period of each branch cold plate are adjusted in real time to achieve efficient matching between the coolant supply rhythm and heat load changes.

[0080] The branch pump control and coordination module is used to perform demand analysis on the cooling demand intensity of each floor based on a standardized heat load trend dataset, and drive the corresponding branch pump operating frequency to change to match the liquid supply output based on the demand analysis results. The demand analysis process uses the fluctuation frequency and intensity of the heat load of each floor to estimate the potential liquid supply demand for the current and next period, and adjusts the target frequency value of the branch pump accordingly to reduce energy consumption and operational impact while meeting cooling requirements, and further coordinates the dynamic load transfer process between multiple floors.

[0081] The closed-loop liquid supply regulation feedback module is used to evaluate the coordination status of the liquid supply, heat dissipation, and liquid return processes of each layer, taking the supply status assessment results and demand analysis results as inputs, and dynamically adjusts the liquid cooling control strategy based on the assessment results. This module constructs a closed-loop behavior model of liquid supply-heat exchange-liquid return in the multi-layer cooling path, and jointly judges key process nodes such as the stability of return flow rate, the synchronicity of heat flux response, and pressure difference hysteresis. If insufficient coordination is detected, the module immediately optimizes the coolant circulation rate, switches the bypass channel, and adjusts the control threshold range to ensure that the overall operation is in a highly efficient and stable state.

[0082] like Figure 2 The diagram shows the structure of a layered microchannel liquid system based on a magnetic interface, as provided in this application example. It illustrates the overall functional structure and working logic of this layered microchannel liquid cooling system based on a magnetic interface. The overall structure is arranged vertically in layers from top to bottom, with each functional module connected sequentially to form a clear operational link, reflecting a closed-loop control approach from status acquisition to regulation feedback.

[0083] At the top is the multi-layer heat load dynamic acquisition module, which is used to sense and acquire the heat load status at different levels and locations in real time, providing basic data support for subsequent flow distribution and regulation. Below it is the branch flow response evaluation module, which analyzes and evaluates the flow response characteristics of each microchannel branch based on the acquired heat load dynamic information, thereby characterizing the different responses of different branches to changes in cooling demand. Further down is the branch pump control and coordination module, which coordinates the operating status of each branch pump according to the branch flow response evaluation results, realizing flow coordination and dynamic matching between multiple branches. At the bottom is the closed-loop liquid supply regulation feedback module, which is used to provide real-time feedback correction of the regulation results, continuously correcting the liquid supply strategy through a closed-loop mechanism to ensure the stability and regulation accuracy of the liquid cooling system during operation. The structure diagram of the layered microchannel liquid system based on the magnetic interface reflects a microchannel liquid cooling workflow with the magnetic interface as the connection basis and layered sensing and coordinated regulation as the core, emphasizing the hierarchical progressive relationship and linkage control logic between multi-layer heat load sensing, branch flow evaluation, pump-level coordinated regulation and closed-loop feedback.

[0084] In this implementation scheme, the multi-layer thermal load dynamic acquisition module is responsible for continuously collecting key thermal load-related data during the actual operation of each layer of the server, including temperature changes on the chip surface, temperature difference evolution characteristics between partitions, and flow rate and pressure fluctuations of the coolant in different channels. By standardizing these data, a unified thermal load trend dataset is constructed, providing time-series and comparable basic data support for subsequent modules, which is a prerequisite for achieving dynamic coordinated control.

[0085] The branch flow response assessment module uses standardized heat load trend data, combined with the current coolant flow rate information for each layer, to quantitatively assess the coolant supply status. The assessment includes whether there are abnormal conditions such as insufficient coolant supply, cooling delay, or oversaturation. The assessment results are used to adjust the coolant supply rhythm of each layer in real time, achieving rapid cooling response and load synchronization.

[0086] The branch pump control and coordination module focuses on intelligent adjustment of the liquid supply execution layer. Based on heat load trend data, it dynamically analyzes the cooling demand intensity of servers at each layer, and then drives the corresponding branch pumps to adjust their operating frequency, ensuring a closed-loop match between the liquid supply output intensity and the cooling demand intensity. This module can significantly improve the targeting of cooling resource allocation, reduce energy consumption, and improve control accuracy.

[0087] The closed-loop liquid supply regulation feedback module integrates the liquid supply status assessment results with the cooling demand analysis results to conduct a coordinated assessment of the complete liquid supply-heat dissipation-return process for each layer. By analyzing the stability of the return flow rate and the synchronicity of the heat flux response in each layer, it determines whether the current cooling loop is operating smoothly and dynamically adjusts the liquid cooling control strategy accordingly, such as adjusting the main pump liquid supply ratio, opening and closing specific bypasses, and optimizing the operating rhythm of heat dissipation nodes, thereby ensuring optimal cooling conditions.

[0088] 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.

[0089] 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 layered microchannel liquid cooling method based on a magnetic interface, characterized in that: include: S1. Collect temperature characteristic data and fluid flow rate and pressure data during the operation of each layer of the server, and preprocess the collected temperature characteristic data and fluid flow rate and pressure data to construct a standardized heat load trend dataset. The specific steps for collecting temperature characteristic data and fluid flow velocity and pressure data during the operation of each layer of the server are as follows: Collect temperature characteristic data during the operation of each server layer. The temperature characteristic data includes: real-time surface temperature of each layer's two-phase cold plate (4), inlet coolant temperature, outlet coolant temperature and hot spot temperature peak. At the same time, calculate and record the average temperature of each layer's two-phase cold plate (4), the instantaneous temperature difference between inlet and outlet and the temperature standard deviation. The temperature difference between the inlet coolant temperature and the outlet coolant temperature at each moment is calculated to form a temperature difference sequence. The difference between the maximum and minimum temperature values ​​is extracted from the temperature difference sequence and recorded as the temperature fluctuation intensity. Collect fluid flow rate and pressure data during the operation of each server layer. The fluid flow rate and pressure data include: the return flow rate value of the condenser return port, the inlet coolant flow rate value of each layer's two-phase cold plate (4), the outlet coolant flow rate value, the inlet coolant pressure, and the outlet coolant pressure. At the same time, calculate and record the coolant pressure difference between the inlet and outlet. S2, based on a standardized heat load trend dataset, combines coolant flow rate to assess the coolant supply status of each layer, and dynamically adjusts the coolant supply rhythm of the corresponding layer based on the assessment results. S3, based on a standardized heat load trend dataset, performs demand analysis on the cooling demand intensity of each floor, and drives the corresponding floor branch pump operating frequency to change to match the liquid supply output based on the demand analysis results. S4 uses the supply status assessment results and demand analysis results as inputs to assess the coordination status of the liquid supply, heat dissipation and liquid return processes of each layer, and dynamically adjusts the liquid cooling control strategy based on the assessment results.

2. The layered microchannel liquid cooling method based on a magnetic interface according to claim 1, characterized in that: The specific steps for preprocessing the collected temperature characteristic data and fluid flow velocity and pressure data to construct a standardized heat load trend dataset are as follows: For temperature characteristic data, the measurement deviations of different sensors are corrected using the initial calibration coefficient; for fluid flow velocity and pressure data, extreme values ​​are denoised to eliminate abnormal data points caused by instantaneous fluctuations and acquisition jitter, and linear interpolation is uniformly used to fill in the timestamp misalignment data caused by temporary packet loss. The temperature characteristic data and fluid velocity and pressure data that have been synchronously corrected are uniformly normalized to construct a standardized heat load trend dataset.

3. The layered microchannel liquid cooling method based on a magnetic interface according to claim 1, characterized in that: The specific steps for assessing the coolant supply status of each layer based on the standardized heat load trend dataset and coolant flow rate are as follows: Divide the inlet coolant flow rate of the current layer two-phase cold plate (4) by the coolant pressure difference between the inlet and outlet to obtain the flow adaptation value; Take the reciprocal of the average temperature of the two-phase cold plate (4) in the current layer to obtain the reciprocal of the average temperature; Multiply the temperature fluctuation intensity by the corresponding temperature change modulation coefficient to obtain the equivalent thermal fluctuation intensity value; add one to the equivalent thermal fluctuation intensity value and then take the reciprocal to obtain the thermal disturbance attenuation value. Multiply the flow fit value, the reciprocal of the mean temperature, and the thermal disturbance attenuation value, then add one and take the logarithm to obtain the liquid supply adequacy assessment value.

4. The layered microchannel liquid cooling method based on a magnetic interface according to claim 3, characterized in that: The specific steps for dynamically adjusting the liquid supply rhythm of the corresponding layer based on the supply status assessment results are as follows: Real-time comparison of the current liquid supply adequacy assessment value and adequacy assessment threshold of the two-phase cold plate (4): When the liquid supply adequacy assessment value is less than or equal to the adequacy assessment threshold, the branch pump at the corresponding crossbeam is automatically driven to increase the extraction flow rate, increase the coolant flow rate distributed to the current layer microchannel through the main channel of the column (3), and simultaneously adjust the contact pressure of the current layer magnetic interface to maintain the stable contact of the liquid metal contact, while triggering the top condenser (1) to accelerate the switching of the return path. When the liquid supply adequacy assessment value is greater than the adequacy assessment threshold, reduce the operating frequency of the current layer branch pump, reduce the inflow of coolant, and switch the return liquid channel to the buffer branch of the liquid storage tank (2), while keeping the steam pipe (8) in a conductive state to maintain the overall balance of circulation in each layer.

5. The layered microchannel liquid cooling method based on a magnetic interface according to claim 4, characterized in that: The specific steps for demand analysis of cooling demand intensity for each floor based on the standardized heat load trend dataset are as follows: Divide the collected coolant pressure difference by the outlet coolant flow rate to obtain the actual flow resistance value of the microchannel of the two-phase cold plate (4), and at the same time record and calculate the average flow resistance value of each layer of the two-phase cold plate (4). The difference between the peak temperature and the average temperature is divided by the sum of the temperature standard deviation plus one to obtain the thermal driving intensity value. The thermal driving intensity value is squared, then the logarithm is taken to obtain the thermal driving response value. The absolute value of the difference between the actual flow resistance value and the mean flow resistance value is calculated, multiplied by the flow resistance fluctuation adjustment factor, and then the square root is obtained to obtain the flow resistance offset term. The flow resistance fluctuation adjustment factor, which ranges from 0.5 to 1.2, is used to control the dynamic coupling effect of the channel micro-flow resistance change on the target frequency calculation value. It comes from the degree of deviation and the rate of change between the actual flow resistance of the current microchannel group and the historical sliding window mean. Divide the pressure fluctuation compensation factor by the value of the coolant pressure difference plus one to obtain the pressure difference compensation term. The pressure fluctuation compensation factor, which ranges from 0.8 to 1.5, is used to compensate for the pressure response fluctuation caused by the coolant flow disturbance caused by the alternation of hot and cold liquids in the channel. It is derived from the relationship between the peak difference and duration of the coolant pressure fluctuation amplitude between the inlet and outlet of the two phase plates. Subtract the flow resistance offset term from the thermally driven response value and then add it to the differential pressure compensation term to obtain the target frequency value of the branch pump.

6. The layered microchannel liquid cooling method based on a magnetic interface according to claim 5, characterized in that: The specific steps for driving the corresponding layer branch pump's operating frequency variation based on the demand analysis results to match the liquid supply output are as follows: Based on the calculation results of the target frequency value of the branch pump, the target frequency value of the branch pump is automatically sent to the diaphragm branch pump at the connection position of the beam and the column (3), and the diaphragm branch pump is directly driven to perform pulse liquid supply according to the branch pump; as the liquid supply frequency is continuously updated, the dynamic change of the flow rate in the coolant inlet pipe (6) is collected in real time and used to automatically adjust the pressure balance of the return path to ensure that the gas-liquid separation process in the steam pipe (8) remains stable. When the branch pump is running, when the target frequency value of the branch pump continues to rise, the layer where the current two-phase cold plate (4) is located is identified as a high heat load area. The condenser (1) is linked to open the auxiliary return liquid sub-channel to accelerate the condensation efficiency of the top steam and reduce the degree of heat accumulation. At the same time, the coolant outlet pipe (7) records the real-time temperature change and dynamically updates the basic liquid supply rate of the main circulation pump according to the real-time temperature change, and redistributes the liquid flow rate ratio sent from the liquid storage tank. When the target frequency value of the branch pump continues to decrease, shorten the single start duration of the branch pump, and automatically adjust the instantaneous flow distribution ratio of each layer based on the real-time fluid flow rate and pressure data of the two-phase cold plates (4) of each layer: adjust the actual liquid supply of each layer under the condition that the total flow rate of the main channel remains unchanged, reduce the liquid supply of the two-phase cold plate layer with an instantaneous temperature difference at the inlet and outlet less than the temperature difference threshold, increase the liquid supply of the two-phase cold plate layer with an instantaneous temperature difference at the inlet and outlet greater than or equal to the temperature difference threshold, and eliminate the difference in coolant flow rate between layers.

7. The layered microchannel liquid cooling method based on a magnetic interface according to claim 6, characterized in that: The specific steps for evaluating the coordination status of the liquid supply, heat dissipation, and liquid return processes at each level, using the supply status assessment results and demand analysis results as input, are as follows: Add one to the instantaneous temperature difference between the inlet and outlet of the two-phase cold plate (4), take the natural logarithm, and then multiply it by the square of the value after subtracting the target frequency value of the branch pump to obtain the coordinated drive strength value. Square the return flow rate value of the condenser (1) return port, divide it by one and add the value of the inlet coolant pressure at the inlet of the two-phase cold plate (4), add it to the liquid supply adequacy assessment value, and then add one and take the natural logarithm to obtain the coordinated absorption capacity value. Divide the coordinated driving strength value by the coordinated absorption capacity value to obtain the closed-loop coordination evaluation value.

8. The layered microchannel liquid cooling method based on a magnetic interface according to claim 7, characterized in that: The specific steps for dynamically adjusting the liquid cooling control strategy based on the evaluation results are as follows: Real-time comparison of the current closed-loop coordination assessment value with the coordination assessment threshold, which includes a first coordination threshold and a second coordination threshold: When the closed-loop coordination assessment value is less than or equal to the second coordination threshold, the high-frequency emergency liquid supply mode of the two-phase cold plate (4) branch pump is immediately triggered, and the main bypass branch of the condenser (1) return liquid channel is opened simultaneously to force the liquid return to enter the crossbeam microchannel first. At the same time, the cold plate number, temperature difference value, pressure difference value and return liquid speed are recorded in the abnormal data buffer area and marked as structural destability warning. When the closed-loop coordination assessment value is greater than the second coordination threshold and less than or equal to the first coordination threshold, the branch pump single start time is dynamically adjusted on the basis of maintaining the current branch pump target frequency, and the distribution ratio of the basic liquid supply of the main circulation pump is updated in conjunction with the main circulation pump, so that the actual flow rate is enhanced under the premise of maintaining stable operation. At the same time, the working status monitoring process of the two-phase cold plate (4) is activated, and the change in the operating rhythm is transmitted back to the main control module in real time. When the closed-loop coordination assessment value is greater than the first coordination threshold, the branch pump is switched to the low load regulation mode, the flow distribution ratio is reduced, and the effective cross-section of the condenser (1) return liquid channel is simultaneously reduced, so as to guide more steam heat flow to the adjacent two-phase cold plate (4), and at the same time mark the current two-phase cold plate (4) as the liquid supply oversaturation area, and record the core data of the process from liquid supply to temperature rise to liquid return.

9. A layered microchannel liquid cooling system based on a magnetic interface, employing the layered microchannel liquid cooling method based on a magnetic interface as described in any one of claims 1-8, characterized in that: include: The multi-layer heat load dynamic acquisition module is used to collect temperature characteristic data and fluid flow rate and pressure data during the operation of each layer of the server, and to preprocess the collected temperature characteristic data and fluid flow rate and pressure data to construct a standardized heat load trend dataset. The branch flow response assessment module is used to assess the coolant supply status of each layer based on a standardized heat load trend dataset and coolant flow rate, and to dynamically adjust the coolant supply rhythm of the corresponding layer based on the assessment results. The branch pump control and coordination module is used to perform demand analysis on the cooling demand intensity of each floor based on a standardized heat load trend dataset, and drive the corresponding floor branch pump to change its operating frequency to match the liquid supply output based on the demand analysis results. The closed-loop liquid supply regulation feedback module is used to evaluate the coordination status of the liquid supply, heat dissipation and liquid return processes of each layer, taking the supply status assessment results and demand analysis results as inputs, and dynamically adjust the liquid cooling control strategy based on the assessment results.

Images