A method and system for circulating cooling of a single-phase immersion liquid-cooled server

Abstract

This invention relates to the field of immersion liquid cooling data center cooling technology, and in particular provides a method and system for circulating cooling of a single-phase immersion liquid-cooled server. The method includes comparing the real-time temperature of the coolant with a preset temperature threshold, and the control system issuing a first program command to initiate a circulation mode within the immersion liquid-cooled cabinet; or issuing a second program command to initiate a heat exchange circulation mode. After the heat exchange circulation mode is activated, the main outlet pipe is opened; the coolant flows back to the outlet of the immersion liquid-cooled cabinet through return branch pipes, and the real-time temperature of the coolant is continuously compared with the preset temperature threshold. Based on the comparison result, the system switches between the immersion liquid-cooled cabinet circulation mode and the heat exchange circulation mode. The system includes an immersion liquid-cooled cabinet, branch pipe valves, a main pipe valve, a main outlet pipe, a pump, a heat exchanger, and branch pipes. This invention achieves topology reconfiguration capability for the coolant flow, dynamically switching between local and global circulation based on the heat load.

Description

Technical Field

[0001] This invention relates to the field of immersion liquid cooling data center technology, and particularly to a single-phase immersion liquid cooling server circulating cooling method and system. Background Technology

[0002] Cloud computing provides scalable computing, storage, and network resources through virtualization technology, enabling on-demand allocation and elastic scaling. AI, on the other hand, uses algorithms and massive amounts of data to train models, giving systems the ability to perceive, understand, and make decisions. With the widespread application and continuous evolution of AI and cloud computing technologies, the computing power density of chips involved in AI and cloud computing is growing exponentially. The power consumption of core components such as CPUs and GPUs has jumped from the hundreds of watts to the kilowatts. Traditional air cooling can no longer meet the heat dissipation requirements, and single-phase immersion liquid cooling has become the mainstream solution due to its efficient heat exchange capabilities. However, in high-power scenarios, the energy consumption problem of liquid cooling systems themselves is becoming increasingly prominent, especially the continuous operation of cooling equipment, which causes a large amount of energy waste. Traditional liquid cooling systems adopt a continuous cooling mode. Regardless of the chip load, the coolant that has absorbed heat will flow through the heat exchanger for forced cooling. However, server load fluctuates significantly. At night, when the load is low, the heat generated by the chip drops sharply, and the temperature of the coolant at the outlet may still be within a safe range. At this time, the cooling process is a complete waste of energy.

[0003] Prior art 1, Chinese Patent Application No. 202410928007.3, discloses a liquid-cooled server coolant circulation system and equipment. The system includes: a server cooling side, a circulating heat exchange side, and a remote control side. The server cooling side and the circulating heat exchange side exchange heat through a heat exchanger. The remote control side obtains the operating parameters of the server cooling side, the circulating heat exchange side, and the heat exchanger through communication. The server cooling side includes a cooling booster pump, a liquid dispensing unit, an immersion tank, a recovery unit, several server cooling plates, and several immersion server positions, with the immersion server positions located inside the immersion tank. The circulating heat exchange side includes a liquid storage unit, a natural cooler, a first refrigerant pump, a second refrigerant pump, and a compressor. Although intelligent control and precise adjustment have enabled the liquid cooling system to operate efficiently, safely, and reliably, effectively reducing server temperature and extending equipment lifespan, which is of great significance to the green environmental protection and sustainable development of data centers, there are still some drawbacks. These include the loss of secondary heat transfer efficiency due to indirect heat exchange through heat exchangers, control delays caused by the complex three-sided (server / circulation heat exchange / remote control) system architecture, and excessive energy consumption due to reliance on multiple refrigerant pumps and compressors.

[0004] Prior art two, Chinese patent application number 202310707957.9, discloses a single-phase liquid-cooled server and its cooling circulation system. The single-phase liquid-cooled server includes a main casing, a jet structure, an inlet, and an outlet. The jet structure is mounted on the main casing and contains a cooling medium. The inlet is mounted on the main casing and connected to the jet structure. The inlet drives the cooling medium to be ejected from the jet structure to provide jet liquid cooling to the interior of the main casing. The outlet is mounted on the main casing, through which the cooling medium inside the main casing is discharged. Although the jet structure dissipates heat from the entire internal cavity, and through an indirect cooling structure, the internal cavity exchanges heat with the external cavity, thus achieving indirect liquid cooling of the second heat-generating element, providing simple and efficient liquid cooling for the single-phase liquid-cooled server; however, the jet structure increases flow resistance, and the indirect cooling method results in uneven temperature gradients, making the fixed flow rate mode unsuitable for dynamic heat loads.

[0005] Prior art three, Chinese patent application number 201210253663.5, discloses a forced cooling circulation structure for a fanless server, including a server rack housing and a server mounting rack installed inside the server rack housing. The key feature is that the server mounting rack divides the space inside the server rack housing into a hot-end area and a cold-end area; an air duct is formed at the top of the server rack housing by a sealing plate; a fan unit is installed on the side of the air duct located in the hot-end area, and a surface cooler unit is installed on the side of the air duct located in the cold-end area; a fanless server is installed on the server mounting rack, the fanless server including a server housing, with an air inlet on one side of the server housing and an air outlet on the other side, the air inlet facing the cold-end area and the air outlet facing the hot-end area. Although using a fanless server eliminates the need for the server's own exhaust fan for heat dissipation, increasing server capacity within the same rack, it suffers from the inherent air heat exchange efficiency bottleneck of air-cooled architecture, reduced space utilization due to hot-end and cold-end partitioning, and the risk of condensation from the surface cooler unit.

[0006] Current technologies 1, 2, and 3 suffer from energy waste in traditional single-mode operation, response lag in fixed threshold control, imbalance in branch / main pipeline flow distribution, and cumulative thermal resistance effects in single-phase immersion direct cooling architectures with indirect heat transfer. Therefore, this invention provides a single-phase immersion liquid-cooled server circulating cooling method and system. Summary of the Invention

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] In one aspect, the present invention provides a method for circulating cooling of a single-phase immersion liquid-cooled server, comprising the following steps:

[0009] The coolant flows through the chip in the immersion liquid cooling cabinet. The real-time temperature of the coolant is compared with a preset temperature threshold. When the real-time temperature does not exceed the preset threshold, the control system issues the first program command to start the circulation mode in the immersion liquid cooling cabinet. When the real-time temperature reaches and exceeds the preset threshold, the control system issues the second program command to start the heat exchange circulation mode.

[0010] After the heat exchange cycle mode is started, the main outlet pipe is opened; at the same time, the preset threshold is dynamically corrected by combining the chip temperature trend, load rate and return port temperature of the immersion liquid cooling cabinet; the opening degree of the branch pipe valve and the main pipe valve is controlled by the PWM signal to realize the transition state of fully open branch, fully open main cycle or proportional adjustment.

[0011] The coolant flows back to the outlet of the immersion liquid-cooled cabinet through the return branch pipe. The real-time temperature of the coolant is then compared with the preset temperature threshold. Based on the comparison results, the internal circulation mode and heat exchange circulation mode of the immersion liquid-cooled cabinet are switched.

[0012] In one optional implementation, the process of achieving a transitional state of fully open branch loops, fully open main loop, or proportional adjustment includes the following steps:

[0013] The system receives feedback of temperature contact point coordinates, real-time chip load rate and return port temperature, generates a time axis extension vector based on temperature contact point coordinates, converts real-time chip load rate into intensity modulation coefficient, constructs a spatial compensation scalar based on return port temperature, and outputs dynamic deformation threshold boundary.

[0014] Phase lag compensation is applied to the dynamic deformation threshold boundary. When the boundary contraction rate exceeds the critical value, a full-open trigger pulse is generated. When the boundary fluctuation frequency is in the transition zone, a proportional adjustment waveform is output to form a PWM fundamental sequence. The historical flow ratio of the total outflow pipeline is extracted from the PWM fundamental sequence as an inertial weight. The transient compensation amount is calculated based on the temperature difference between the current deformation threshold boundary and the return port, and the valve opening combination weight is synthesized.

[0015] By combining the valve opening weights, the fully open trigger pulse drives the branch valves to form a zero-resistance channel, and the proportional adjustment waveform matches the main pipeline valve to generate a gradual throttling surface, thereby achieving three-state output: fully open branch, fully open main circulation, or proportional transition.

[0016] In one optional implementation, the process of synthesizing the combined weights of valve opening degrees includes the following steps:

[0017] The periodic performance consumption envelope is extracted from the historical flow data of the total liquid outlet pipeline. The inertial strong and weak oscillation sequence is generated by the flow difference operation between adjacent time periods. The inertial strong and weak oscillation sequence is input into the decay memory filter and the inertial weighted basis is output.

[0018] The real-time difference between the dynamic deformation threshold boundary and the return port temperature is received, processed by the spatial thermo-pressure converter, and a transient compensation vector field is generated with the deformation threshold boundary curvature as the field strength gradient reference and the absolute value of the temperature difference as the action intensity scalar.

[0019] An inertial weighted base and a transient compensation vector field are projected along the pipe axis to form a steady-state skeleton. Transient compensation vector field components are injected at the nodes of the steady-state skeleton to form a weighted grid with compensation nodes.

[0020] The weighted mesh with compensation nodes is reconstructed by the valve response adapter, the free flow domain of the segment corresponding to the fully open trigger pulse is extracted, the mesh density distribution is adjusted according to the waveform slope, and finally the valve opening combination weight is output.

[0021] In one alternative implementation, the process of forming a weighted grid with compensated nodes includes the following steps:

[0022] Using the centerline of the main outlet pipeline as the reference axis, the inertial weight base is axially stretched and transformed according to the direction of fluid motion to form a tubular weight frame, and equidistant skeleton control nodes are set on the surface of the tubular weight frame.

[0023] The three-dimensional transient compensation vector field is compressed radially along the pipeline, the projection intensity spectrum of the vector field in the axial direction is extracted, the vortex component magnitude of the vector field in the circumferential direction is captured, and a two-dimensional compensation action surface is generated.

[0024] The two-dimensional compensation action surface is spatially phase matched with the skeleton control nodes. The compensation intensity base of each node is determined based on the projected intensity spectrum. The coupling action between adjacent nodes is allocated according to the vortex component magnitude, forming an enhanced node set with energy markers.

[0025] An energy-labeled enhanced node set is used as the input field constraint generator. A tubular weighted frame is used as the initial mesh topology. Mesh density expansion is generated at the nodes according to the compensation intensity base. Non-uniform connection ribs are generated between the nodes according to the coupling action, and finally a weighted mesh with compensated nodes is constructed.

[0026] In one alternative implementation, the process of forming an energy-tagged augmented node set includes the following steps:

[0027] The amplitude distribution of the projected intensity spectrum is spatially bound to the axial position of the skeleton control node. The zero point of the reference axis is used as the origin of the spectrum-axis alignment. The pipe length is divided equally according to the intensity spectrum wavelength to generate a spectrum node mapping table.

[0028] The spectral node mapping table is input into the intensity converter. The logarithmic attenuation of the spectral amplitude corresponding to each node is taken as the base matrix. The curvature change rate of the deformation threshold boundary at the node is superimposed, and the dynamic compensation base matrix is ​​output.

[0029] The captured vortex component magnitude is processed by the circulation decomposer, and the phase difference angle between the peak values ​​of the magnitude is extracted to generate the interaction coefficient of adjacent nodes. The magnitude envelope area and the filling ratio of the pipe section are obtained to form a set of potential transfer factors.

[0030] The core of the input field is the fusion of the dynamic compensation base matrix and the energy potential transfer factor set. The elements of the compensation base matrix are used as node energy cores. The energy cores are extended to adjacent nodes according to the action coefficient to form radiation energy bands. The width of the radiation energy bands is adjusted according to the filling ratio, and finally, an enhanced node set with energy labels is generated.

[0031] In one optional implementation, the process of obtaining the modulus envelope area and the filling ratio of the pipe cross-section includes the following steps:

[0032] The captured vortex component magnitude waveform is input into the extreme value topology analyzer to identify the effective peak group set whose magnitude exceeds the critical amplitude, record the azimuth coordinates of each peak in the pipe circumferential coordinate system, and generate a peak group azimuth distribution map.

[0033] The peak group azimuth distribution map is processed by the relative motion solver to calculate the tangential offset of the azimuth angle of adjacent peaks, convert the offset into the vector angle with the center of the pipe as the origin, and output the node action angle matrix.

[0034] The modulus envelope area is input into the geometric constraint converter. The cross-sectional area of ​​the pipe is used as the reference constraint box. The dynamic overlap rate between the envelope area and the reference constraint box is calculated, and the vortex core occupancy factor is generated.

[0035] The node action angle matrix and the vortex core occupancy factor are input into the fluid interference synthesizer. The tangent of the vector angle is used as the action intensity coefficient. The effective action radius of the action intensity coefficient is modulated by the vortex core occupancy factor, and finally the energy potential transfer factor set is formed.

[0036] In one optional implementation, the process of switching between the internal circulation mode and the heat exchange circulation mode of the immersion liquid-cooled cabinet based on the comparison results includes the following steps:

[0037] The coolant temperature data obtained from the return branch pipe is processed by noise suppression and time domain alignment to form a pure return waveform. The pure return waveform is then input into the phase analyzer to output the return phase marker.

[0038] The return phase marker and the current operating cooling mode state input migration criterion synthesizer form a dual-mode migration trigger field;

[0039] When in heat exchange mode: the amplitude of the return waveform remains below the dynamic deformation threshold boundary, generating an internal circulation enabling factor; when in internal circulation mode: the peak of the return waveform contacts the deformation threshold boundary, triggering a heat exchange activation factor.

[0040] The dual-mode migration trigger field input physical field mapping module drives the closure of the main liquid outlet pipe valve to form a closed loop topology. The heat exchange activation factor triggers the branch valve zero-resistance channel to establish an open flow channel configuration, completing the cooling mode switching.

[0041] In one optional implementation, the process of switching cooling modes includes the following steps:

[0042] The inner loop enables the input constraint field generator, extracts the factor strength value as the closure strength coefficient, associates the free flow domain closure sequence of the weighted grid, and generates the main pipeline valve lockout instruction set;

[0043] The main pipeline valve lockout command set, under the action of the fluid path constraint field, forms a closed loop topology with the submerged cavity wall as the fixed boundary and the return branch pipeline as the unidirectional conduction axis;

[0044] The heat exchange activation factor triggers the zero-resistance channel builder, activates the spatiotemporal coordinates of the fully open trigger pulse, calls the energy distribution of the compensated nodes of the weighted grid, and generates the expansion wavefront of the branch valve.

[0045] The closed-loop topology and open-channel configuration are output to the thermal inertia balance monitor to detect the degree of agreement between the temperature gradient and the heat flow tendency vector in the cavity, verify that the return pure waveform conforms to the phase characteristic spectrum of the target mode, and confirm that the cooling mode switching is completed.

[0046] In another aspect, the present invention provides a single-phase immersion liquid-cooled server circulating cooling system for implementing the single-phase immersion liquid-cooled server circulating cooling method, comprising: an immersion liquid-cooled cabinet, branch pipe valves, a main pipe valve, a main liquid outlet pipe, a pump, a heat exchanger, and branch pipes.

[0047] The immersion liquid-cooled cabinet is equipped with at least one liquid outlet and two liquid inlets. The liquid outlet is installed at the upper end of one side of the immersion liquid-cooled cabinet, and the two liquid inlets are located at the middle and lower ends of one side of the immersion liquid-cooled cabinet. The liquid inlet at the middle position is connected to the main liquid outlet pipe through a branch pipe and a branch pipe valve. A main pipe valve is installed on the main liquid outlet pipe. The outlet end of the main liquid outlet pipe is connected to the inlet end of the pump. The outlet end of the pump is connected to the inlet end of the heat exchanger. The outlet end of the heat exchanger is connected to the liquid inlet at the lower end through a return branch pipe.

[0048] This invention achieves efficient thermal management of an immersion liquid cooling system through a multimodal control strategy. A dual-mode adaptive switching mechanism, based on a binary judgment logic of temperature thresholds, constructs the basic control framework. The first program instruction maintains micro-circulation within the cabinet, reducing pump power loss; the second program instruction triggers forced heat exchange, ensuring heat dissipation under transient thermal shocks. A dynamic parameter correction system establishes a multivariate feedback model of chip temperature change rate, load rate, and return liquid temperature. An online threshold adjustment algorithm avoids control lag, and PWM modulation enables stepless adjustment of valve opening, balancing the flow distribution between branch and main pipelines. A closed-loop control architecture is constructed by setting a secondary temperature detection node in the return branch, forming a hysteresis comparison circuit for mode switching, and building a dual feedback control loop for the coolant temperature field. Attached Figure Description

[0049] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0050] Figure 1 This is a flowchart of the single-phase immersion liquid-cooled server circulating cooling method provided in Embodiment 1 of the present invention;

[0051] Figure 2 This is a schematic diagram of the single-phase immersion liquid cooling server circulating cooling method provided in Embodiment 1 of the present invention;

[0052] Figure 3 This is a process diagram of comparing the real-time temperature of the coolant with a preset temperature threshold provided in Embodiment 2 of the present invention;

[0053] Figure 4 This is a process diagram illustrating the transition states of fully open branches, fully open main circulation, or proportional adjustment provided in Embodiment 5 of the present invention.

[0054] Figure 5 This is a diagram illustrating the process of switching between the internal circulation mode and the heat exchange circulation mode of the immersion liquid-cooled cabinet based on the comparison results, as provided in Embodiment 10 of the present invention.

[0055] Figure 6 This is a block diagram of the single-phase immersion liquid-cooled server circulating cooling system provided in Embodiment 13 of the present invention;

[0056] Figure 7 A block diagram of the electronic device provided by the present invention;

[0057] Figure 8 A block diagram of a computer-readable storage medium provided for this invention;

[0058] Reference numerals: 1. Immersion liquid cooling cabinet; 2. Branch pipe valve; 3. Main pipe valve; 4. Main outlet pipe; 5. Pump; 6. Heat exchanger; 7. Branch pipe; 8. Central processing unit / microprocessor / main control chip; 9. Storage medium; 10. Data bus 1; 11. Input / output bus / external bus / device bus; 12. Display; 13. Input / output device; 14. Computer-readable instructions; 15. Non-transitory computer-readable storage medium. Detailed Implementation

[0059] The technical solutions of the present invention will now be described with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0060] Hereinafter, the terms "first," "second," etc., are used for descriptive convenience only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0061] In this invention, unless otherwise explicitly specified and limited, the term "connection" should be interpreted broadly. For example, "connection" can be a fixed mechanical connection, a detachable mechanical connection, or an integral part; or, "connection" can be a direct connection or an indirect connection through an intermediate medium. Furthermore, unless otherwise explicitly specified and limited, the term "coupling" should be interpreted broadly. For example, "coupling" can be a direct electrical connection, such as physical contact and electrical conduction between two components; it can also be understood as an electrical connection between different components in a circuit structure through physical lines capable of transmitting electrical signals, such as copper foil or wires on a printed circuit board (PCB), to transmit electrical signals; or, "coupling" can be an indirect electrical connection between two components through an intermediate medium; or, "coupling" can be an electrical connection between two components in a non-contact manner, such as an electrical connection between two components using capacitive coupling to transmit electrical signals.

[0062] In this embodiment of the invention, directional terms such as "up," "down," "left," and "right" may be defined relative to the orientation of the components shown in the accompanying drawings. It should be understood that these directional terms can be relative concepts, used for relative description and clarification, and can change accordingly depending on the orientation of the components in the accompanying drawings.

[0063] The inventive concept of this invention:

[0064] Example 1: As Figure 1 As shown, this embodiment of the invention provides a single-phase immersion liquid-cooled server circulating cooling method, comprising the following steps:

[0065] Step S100: The coolant in the immersion liquid cooling cabinet flows through the chip. The real-time temperature of the coolant is compared with the preset temperature threshold. When the real-time temperature does not exceed the preset threshold, the control system issues the first program command to start the circulation mode in the immersion liquid cooling cabinet. When the real-time temperature reaches and exceeds the preset threshold, the control system issues the second program command to start the heat exchange circulation mode.

[0066] Step S200: After the heat exchange cycle mode is started, the main outlet pipe is opened; at the same time, the preset threshold is dynamically corrected by combining the chip temperature trend, load rate and return port temperature of the immersion liquid cooling cabinet; the opening degree of the branch pipe valve and the main pipe valve is controlled by the PWM signal to realize the transition state of fully open branch, fully open main cycle or proportional adjustment.

[0067] Step S300: The coolant flows back to the outlet of the immersion liquid-cooled cabinet through the return branch pipe. The real-time temperature of the coolant is then compared with the preset temperature threshold. Based on the comparison result, the internal circulation mode and heat exchange circulation mode of the immersion liquid-cooled cabinet are switched.

[0068] In the above embodiments, the specific principle is as follows: Figure 2 As shown, this embodiment achieves efficient thermal management of the immersion liquid cooling system through a multi-modal control strategy. The dual-mode adaptive switching mechanism (step S100) constructs a basic control framework based on a binary judgment logic of temperature thresholds. The first program instruction maintains micro-circulation within the cabinet, reducing pump power loss, while the second program instruction triggers forced heat exchange to ensure heat dissipation under transient thermal shock. The dynamic parameter correction system (step S200) establishes a multi-factor feedback model of chip temperature change rate, load rate, and return liquid temperature. An online threshold adjustment algorithm avoids control lag, and PWM modulation enables stepless adjustment of valve opening, balancing the flow distribution between the branch and main pipeline. The closed-loop control architecture (step S300) sets a secondary temperature detection node in the return branch, forming a hysteresis comparison circuit for mode switching, thus constructing a dual feedback control loop for the coolant temperature field.

[0069] In summary, this embodiment achieves the ability to reconstruct the topology of the coolant flow, dynamically switches between local and global circulation based on thermal load, eliminates system hydraulic oscillations through continuous adjustment of valve opening, improves the response speed to sudden loads through dynamic thresholds, and ensures the timing accuracy of mode switching by monitoring the return flow temperature; ultimately achieving Pareto optimality by minimizing pumping power consumption and maximizing heat dissipation efficiency.

[0070] Example 2: As Figure 3As shown, based on Example 1, the process of comparing the real-time temperature of the coolant with a preset temperature threshold in step S100 of this embodiment of the invention includes the following steps:

[0071] Step S101: Obtain the original coolant temperature sequence, and form a pure temperature waveform after noise suppression and time alignment processing; input the preset temperature threshold into the load-temperature coupling program, receive the chip load rate and historical temperature rise curve, and output the deformation threshold boundary. The upper boundary value will generate elastic expansion and contraction with load fluctuation.

[0072] Step S102: The pure temperature waveform and the deformation threshold boundary are synchronously input into the phase analyzer. When the amplitude of the pure temperature waveform is continuously lower than the boundary, a steady-state feature vector is generated. When the peak of the pure temperature waveform touches the boundary, a transition feature marker is triggered.

[0073] Step S103: The steady-state feature vector activates the circulating command encoder and outputs the first program command; the transition feature marker triggers the heat exchange enabler and outputs the second program command while feeding back the current temperature contact point coordinates to the preset threshold for dynamic correction; in the reflow temperature detection, the phase analyzer performs boundary penetration detection on the corrected deformation threshold boundary, and triggers the command switching condition when the reflow temperature waveform separates from the boundary.

[0074] In the above embodiments, this embodiment achieves nonlinear deformation of a preset threshold through a load-temperature coupling program, enabling the temperature boundary to have adaptive elastic expansion and contraction characteristics based on the chip load rate and historical temperature rise curve, thus solving the problem of hysteresis response to transient thermal loads with a fixed threshold. Based on the phase analysis of the noise-suppressed pure temperature waveform and the deformation threshold, a steady-state / transition dual-mode feature detection system is established. Continuous amplitude detection ensures the stability of mode switching, while peak contact detection provides critical state response capability. The steady-state feature vector and transition feature marker respectively drive the circulating current command encoder and the heat exchange enabler, forming a feedback closed loop of command output and threshold correction. The boundary penetration detection mechanism achieves precise triggering of mode switching conditions by separating the return flow temperature waveform from the correction boundary. Time alignment processing ensures the temporal consistency of the temperature sequence and load data, and the synchronous input of the phase analyzer ensures the real-time performance of threshold comparison; temperature contact point coordinate feedback enables the dynamic correction of the preset threshold to have spatial positioning capability, improving the control accuracy of the transition process.

[0075] In summary, this embodiment constructs an adaptive liquid cooling control system with load following characteristics. Through the synergistic effect of noise suppression, elastic boundary, dual-mode detection and closed-loop correction, it achieves a smooth transition and energy consumption optimization during the cooling mode switching process.

[0076] Example 3: Based on Example 2, the process of forming a clean temperature waveform through noise suppression and time alignment in step S101 of this embodiment of the invention includes the following steps:

[0077] Step S1011: The sensor array distributed in the three-dimensional coordinate system of the immersion cavity of the immersion liquid cooling cabinet generates the original temperature field intensity distribution, and extracts the heat flow tendency vector through the differential operation of temperature difference between adjacent nodes.

[0078] Step S1012: Input the heat flow tendency vector into the transient response screen, and divide the time axis according to the chip power pulse frequency: use the average field strength convergence during the power steady period, and enable gradient change rate capture during the power jump period to output a time-scale aligned field strength sequence.

[0079] Step S1013: The time-scale aligned field strength sequence enters the eddy current interference canceller, identifies the periodic disturbance ripples formed by the coolant flow vortex, reverses the phase of the periodic disturbance ripples and superimposes them onto the original coolant temperature sequence to generate a decoherent field strength matrix.

[0080] Step S1014: The decoherent field strength matrix is ​​reconstructed by boundary constraints. The cavity wall temperature is used as the static anchor point, and the heat flow tendency vector is used as the dynamic correction reference. Finally, the pure temperature waveform is output.

[0081] In the above embodiments, this embodiment replaces the traditional point temperature acquisition with three-dimensional field strength gradient analysis, and combines power pulse time domain segmentation and eddy current coherence elimination technology to ensure that the output pure temperature waveform completely retains the true thermal state of the coolant, providing a basic waveform with physical field spatiotemporal characteristics for the subsequent phase analyzer.

[0082] Example 4: Based on Example 3, the process of extracting the heat flux tendency vector through the differential operation of temperature difference between adjacent nodes in step S1011 of this embodiment of the invention includes the following steps:

[0083] Step S10111: The three-dimensional coordinates of the sensor array form a node space topology. A bidirectional differential scan is performed on the real-time temperature of adjacent nodes. The X-axis node pairs generate a transverse thermal potential difference set, the Y-axis node pairs generate a longitudinal thermal potential difference set, and the Z-axis node pairs generate a vertical thermal potential difference set. The three sets of thermal potential difference sets are combined into a spatial thermal potential gradient map.

[0084] Step S10112: Analyze the spatial thermal potential gradient map, identify effective transition paths where the absolute value of the gradient exceeds the critical value, integrate the thermal potential difference along the effective transition path to generate the kinematic potential energy distribution, and output the vectorized potential energy field.

[0085] Step S10113: The vectorized potential energy field is focused by the vortex core, and the dominant trend component is synthesized by weighting the local potential energy extreme point as the convergence core, and finally the heat flow trend vector is formed.

[0086] In the above embodiments, this embodiment establishes spatial heat transfer correlation through bidirectional differential scanning, uses the path integral of kinetic potential energy to replace traditional vector calculation, and combines the vortex core focusing and aggregation mechanism to enable the heat flow tendency vector to accurately characterize the dynamic heat transport characteristics of coolant in three-dimensional space, providing a physical field kinematic basis for subsequent time-domain compression and waveform reconstruction.

[0087] Example 5: Figure 4 As shown, based on Embodiment 1, the process of achieving the transition state of fully open branch, fully open main loop, or proportional adjustment in step S200 of the present invention includes the following steps:

[0088] Step S201: Receive the feedback temperature contact point coordinates, real-time chip load rate and return port temperature, generate a time axis extension vector based on the temperature contact point coordinates, convert the real-time chip load rate into intensity modulation coefficients, construct a spatial compensation scalar based on the return port temperature, and output the dynamic deformation threshold boundary.

[0089] Step S202: Perform phase lag compensation on the dynamic deformation threshold boundary. When the boundary contraction rate exceeds the critical value, generate a fully open trigger pulse. When the boundary fluctuation frequency is in the transition zone, output a proportional adjustment waveform to form a PWM fundamental wave sequence. Extract the historical flow ratio of the total liquid outlet pipeline as the inertial weight from the PWM fundamental wave sequence. Calculate the transient compensation amount based on the temperature difference between the current deformation threshold boundary and the return port, and synthesize the valve opening combination weight.

[0090] Step S203: Combining the valve opening weights, the fully open trigger pulse drives the branch valve to form a zero-resistance channel, and the proportional adjustment waveform matches the main pipeline valve to generate a gradual throttling surface, realizing three-state output: branch fully open, main circulation fully open, or proportional transition.

[0091] In the above embodiments, this embodiment integrates multi-dimensional correction factors through three-dimensional deformation, uses a phase lag compensator to convert boundary dynamic characteristics into PWM waveforms, and combines a flow field equalization solver to achieve accurate mapping between valve state and thermodynamic deformation, forming a closed-loop adaptive control system.

[0092] Example 6: Based on Example 5, the process of synthesizing the valve opening combination weights in step S202 of this embodiment of the invention includes the following steps:

[0093] Step S2021: Extract the periodic performance consumption envelope from the historical flow data of the total liquid outlet pipeline, generate an inertial strong and weak oscillation sequence through the flow difference operation between adjacent time periods, input the inertial strong and weak oscillation sequence into the decay memory filter, and output the inertial weighted basis.

[0094] Step S2022: Receive the real-time difference between the dynamic deformation threshold boundary and the return port temperature, process it through the spatial thermo-pressure converter, use the deformation threshold boundary curvature as the field strength gradient reference, and use the absolute value of the temperature difference as the action intensity scalar to generate a transient compensation vector field.

[0095] Step S2023: Inertial weight base and transient compensation vector field. Project the inertial weight base along the pipeline axis to form a steady-state skeleton. Inject transient compensation vector field components at the nodes of the steady-state skeleton to form a weighted grid with compensation nodes.

[0096] Step S2024: The weighted mesh with compensation nodes is reconstructed by the valve response adapter, the free flow domain of the corresponding section of the fully open trigger pulse is extracted, the mesh density distribution is adjusted according to the waveform slope, and finally the valve opening combination weight is output.

[0097] In the above embodiments, this embodiment uses periodic performance consumption envelope analytical quantification of historical flow inertia, utilizes spatial thermo-pressure conversion to transform temperature difference into physical action field, and combines fluid-structure interferometry mesh synthesis technology to achieve spatiotemporal coupling of valve control elements, providing the execution terminal with precise control parameters that are adaptable to fluid dynamics.

[0098] Example 7: Based on Example 6, the process of forming a weighted mesh with compensated nodes in step S2023 of this embodiment of the invention includes the following steps:

[0099] Step S20231: Using the center line of the main outlet pipe as the reference axis, the inertial weight base is axially stretched and transformed according to the direction of fluid movement to form a tubular weight frame, and equidistant skeleton control nodes are set on the surface of the tubular weight frame.

[0100] Step S20232: Compress the three-dimensional transient compensation vector field radially along the pipeline, extract the projection intensity spectrum of the vector field in the axial direction, capture the vortex component magnitude of the vector field in the circumferential direction, and generate a two-dimensional compensation action surface.

[0101] Step S20233: Spatial phase matching is performed between the two-dimensional compensation action surface and the skeleton control nodes. The compensation intensity base of each node is determined based on the projected intensity spectrum. The coupling action between adjacent nodes is allocated according to the vortex component magnitude to form an enhanced node set with energy markers.

[0102] Step S20234: The energy-labeled enhanced node set is input to the field constraint generator. The tubular weighted frame is used as the initial mesh topology. Mesh density expansion is generated at the nodes according to the compensation intensity base. Non-uniform connection ribs are generated between the nodes according to the coupling action. Finally, a weighted mesh with compensated nodes is constructed.

[0103] In the above embodiments, this embodiment solidifies the historical flow characteristics into a spatial framework through axial stretching transformation, realizes the three-dimensional to two-dimensional transformation of transient compensation by using field strength dimensionality reduction mapping, and combines node energy injection and topology adaptive growth mechanism to enable the weighted grid to have dual controllable properties of historical inertia characteristics and real-time thermodynamic response.

[0104] Example 8: Based on Example 7, the process of forming an enhanced node set with energy labels in step S20233 of this embodiment of the invention includes the following steps:

[0105] Step S202331: Spatial coordinate binding of the amplitude distribution of the projected intensity spectrum with the axial position of the skeleton control node, with the reference axis zero point as the spectrum-axis alignment origin, and the pipe length equally divided according to the intensity spectrum wavelength to generate a spectrum node mapping table;

[0106] Step S202332: Input the spectral node mapping table into the intensity converter, take the logarithmic attenuation of the spectral amplitude corresponding to each node as the base, superimpose the curvature change rate of the deformation threshold boundary at the node, and output the dynamic compensation base matrix.

[0107] Step S202333: The captured vortex component magnitude is processed by the circulation decomposer, and the phase difference angle between the peak values ​​of the magnitude is extracted to generate the interaction coefficient of adjacent nodes. The magnitude envelope area and the filling ratio of the pipe section are obtained to form a set of potential transfer factors.

[0108] Step S202334: The core of the dynamic compensation base matrix and the energy potential transfer factor set input field is fused. The elements of the compensation base matrix are used as node energy cores. The energy cores are extended to adjacent nodes according to the action coefficient to form radiation energy bands. The width of the radiation energy bands is adjusted according to the filling ratio, and finally, an enhanced node set with energy labels is generated.

[0109] In the above embodiments, this embodiment achieves precise axial positioning of energy distribution through spectrum-axis spatial binding, utilizes dual dynamic calibration compensation bases of logarithmic decay and curvature differential, and combines the combined effect of vortex phase difference and cross-section filling to model the enhanced nodes as a comprehensive energy carrier that carries historical inertia, real-time thermal pressure and fluid motion characteristics.

[0110] Example 9: Based on Example 8, the process of obtaining the modulus envelope area and the filling ratio of the pipe cross-section in step S202333 of this embodiment of the invention includes the following steps:

[0111] Step S2023331: Input the captured vortex component magnitude waveform into the extreme value topology analyzer, identify the effective peak group set whose magnitude exceeds the critical amplitude, record the azimuth coordinates of each peak in the pipe circumferential coordinate system, and generate a peak group azimuth distribution map.

[0112] Step S2023332: The peak group azimuth distribution map is processed by the relative motion solver to calculate the tangential offset of the azimuth angle of adjacent peaks. The offset is converted into the vector angle with the center of the pipe as the origin, and the node action angle matrix is ​​output.

[0113] Step S2023333: Input the modulus envelope area into the geometric constraint converter, use the pipe cross-sectional area as the reference constraint box, calculate the dynamic overlap rate between the envelope area and the reference constraint box, and generate the vortex core occupancy factor;

[0114] Step S2023334: Input the node action angle matrix and vortex core occupancy factor into the fluid interference synthesizer, use the tangent of the vector angle as the action intensity coefficient, and use the vortex core occupancy factor to modulate the effective action radius of the action intensity coefficient, and finally form the energy potential transfer factor set.

[0115] In the above embodiments, this embodiment locks the vortex energy core through extreme topology analysis, establishes a circumferential kinematic model by using tangential offset-vector angle conversion, and combines the envelope-section dynamic coupling mechanism to enable the energy potential transfer factor set to accurately characterize the energy transfer characteristics of the vortex field in the pipe space.

[0116] Example 10: As Figure 5 As shown, based on Example 1, the process of switching between the internal circulation mode and the heat exchange circulation mode of the immersion liquid-cooled cabinet according to the comparison result in step S300 of this embodiment of the invention includes the following steps:

[0117] Step S301: The coolant temperature data obtained in the return branch pipe is processed by noise suppression and time domain alignment to form a return pure waveform. The return pure waveform is input into the phase analyzer to output the return phase mark.

[0118] Step S302: The return phase marker and the current operating cooling mode state input migration criterion synthesizer form a dual-mode migration trigger field;

[0119] When in heat exchange mode: the amplitude of the return waveform remains below the dynamic deformation threshold boundary, generating an internal circulation enabling factor; when in internal circulation mode: the peak of the return waveform contacts the deformation threshold boundary, triggering a heat exchange activation factor.

[0120] Step S303: The dual-mode migration trigger field input physical field mapping module, the internal circulation enable factor drives the closing of the main liquid outlet pipe valve to form a closed loop topology, the heat exchange activation factor triggers the branch valve zero resistance channel to establish an open flow channel configuration, and completes the cooling mode switching.

[0121] In the above embodiments, this embodiment uses the core analysis module of regeneration and reuse of reflux phase characteristics to integrate boundary penetration events and operating status through dual-mode migration trigger field, and combines physical field mapping inversion to achieve uninterrupted switching of cooling modes, forming a closed-loop migration system from temperature detection to fluid path control.

[0122] Example 11: Based on Example 10, the cooling mode switching process in step S303 of this embodiment of the invention includes the following steps:

[0123] Step S3031: The inner loop enables the input constraint field generator, extracts the factor strength value as the closure strength coefficient, associates the free flow domain closure sequence of the weighted grid, and generates the main pipeline valve lockout instruction set;

[0124] Step S3032: The main pipeline valve lockout command set, under the action of the fluid path constraint field, forms a closed loop topology with the submerged cavity wall as the fixed boundary and the return branch pipeline as the unidirectional conduction axis;

[0125] Step S3033: The heat exchange activation factor triggers the zero-resistance channel builder, activates the spatiotemporal coordinates of the fully open trigger pulse, calls the energy distribution of the compensated nodes of the weighted grid, and generates the expansion wavefront of the branch valve.

[0126] Step S3034: The closed-loop topology and open-channel configuration are output to the thermal inertia balance monitor to detect the degree of agreement between the temperature gradient and the heat flow tendency vector in the cavity, verify that the return pure waveform conforms to the phase characteristic spectrum of the target mode, and confirm that the cooling mode switching is completed.

[0127] In the above embodiments, this embodiment transforms logical factors into physical constraints through a constraint field generator, uses the fluid path constraint field to reconstruct the geometric space, and combines the wavefront propagation mechanism of the zero-resistance channel builder to enable mode switching to have both instantaneous responsiveness and thermodynamic stability, thus completing the precise mapping from control commands to fluid space topology.

[0128] Example 12: Based on Example 11, the process of verifying the phase characteristic spectrum of the return pure waveform conforming to the target mode in step S3034 of this embodiment of the invention includes the following steps:

[0129] Step S30341: Call the historical phase marker library according to the switching instruction type, the steady-state feature vector set corresponding to the inner loop mode, the transition feature marker sequence associated with the hot exchange mode, and synthesize the target mode reference spectrum;

[0130] Step S30342: The reflux pure waveform is decomposed along the projection axis using the heat flow tendency vector to generate spatial projection feature lines; the spatial projection feature lines and the target mode reference spectrum are input into the curvature matching degree analyzer to obtain the geometric similarity ratio between the curvature radius of the feature lines and the wavelength of the reference spectrum, detect the spatiotemporal alignment between the inflection point of the feature lines and the phase jump of the reference spectrum, and output the dynamic matching coefficient matrix.

[0131] Step S30343: The dynamic matching coefficient matrix is ​​determined by the threshold penetration detector. When the geometric similarity ratio is continuously greater than the critical value and the spatiotemporal alignment reaches 90%, the steady-state authentication signal of the mode is triggered to complete the verification.

[0132] In the above embodiments, this embodiment establishes a mode authentication benchmark by reconstructing historical phase markers, transforms the waveform into quantifiable geometric features by using heat flow vector space projection, and achieves accurate determination of the steady state of the cooling mode by combining curvature-phase dual matching analysis, thus forming a complete verification chain from fluid thermodynamic features to control logic authentication.

[0133] Example 13: As Figure 6 As shown, based on Embodiments 1-12, the single-phase immersion liquid-cooled server circulating cooling system provided in this embodiment of the invention includes: an immersion liquid-cooled cabinet 1, branch pipe valves 2, a main pipe valve 3, a main liquid outlet pipe 4, a pump 5, a heat exchanger 6, and branch pipes 7.

[0134] The immersion liquid-cooled cabinet 1 is provided with at least one liquid outlet and two liquid inlets. The liquid outlet is installed at the upper end of one side of the immersion liquid-cooled cabinet 1, and the two liquid inlets are located at the middle and lower ends of one side of the immersion liquid-cooled cabinet 1. The liquid inlet at the middle position is connected to the main liquid outlet pipe 4 through a branch pipe 7 and a branch pipe valve 2. A main pipe valve 3 is installed on the main liquid outlet pipe 4. The outlet end of the main liquid outlet pipe 4 is connected to the inlet end of the pump 5. The outlet end of the pump 5 is connected to the inlet end of the heat exchanger 6. The outlet end of the heat exchanger 6 is connected to the liquid inlet at the lower end through a return branch pipe.

[0135] In the above embodiments, in this embodiment, the temperature sensor is directly attached to the chip surface to monitor the core temperature of the chip in real time; the outlet temperature sensor is installed at the outlet of the main outlet pipe 4 of the immersion liquid cooling cabinet 1 to monitor the temperature of the coolant after absorbing the heat generated by the chip during operation; the chip temperature and the outlet coolant temperature are monitored in real time; when the chip temperature is lower than the set threshold, the branch pipe valve 2 is automatically opened by the control system to allow the coolant to flow directly back to the immersion liquid cooling cabinet 1; when the chip temperature rises, the branch pipe valve 2 is closed to allow the coolant to enter the heat exchanger 6 for cooling. A return branch pipe is added at the lower inlet of the immersion liquid cooling cabinet 1 to achieve on-demand distribution by dynamically adjusting the circulation path of the coolant.

[0136] In practical implementation, it also includes components such as electrically controlled valves (branch pipe valves 2 and main pipe valve 3), flow regulation devices, etc.; high-precision thermocouples or infrared temperature sensors are attached to the surface of core chips such as CPUs or GPUs to collect chip temperature in real time, reflecting the most direct indicator of heat dissipation requirements. A platinum resistance temperature sensor is embedded in the main outlet pipe 4 to collect the temperature of the coolant after absorbing heat; at the same time, auxiliary sensors are deployed at the middle and lower inlets to monitor the actual temperature of the return coolant for verifying the cooling effect. A flow control device is also installed to precisely control the flow rate of the return coolant based on the heat generated by the chip and the temperature of the coolant, ensuring effective chip cooling without affecting the cooling effect and system stability due to excessive or insufficient flow.

[0137] In this embodiment, during the operation of the single-phase immersion liquid-cooled server circulating cooling system, the chips within the AI ​​and cloud computing servers are the source of heat generation. Coolant flows through the chips within the immersion liquid-cooled cabinet 1. When the coolant absorbs a small amount of heat from the chips within the immersion liquid-cooled cabinet 1, the temperature rises from its initial value but does not exceed a threshold. The control system determines that the coolant is still within the effective cooling temperature range and does not require cooling. The control system then issues a command to close the main pipeline valve 3 and open the branch pipeline valve 2. Path switching conditions are defined using preset thresholds (chip temperature, coolant temperature, and temperature fluctuation rate). When both the chip temperature and the outlet coolant temperature are below the set low value and an effective temperature difference exists, it is determined that the return branch pipeline can be started. However, when both the chip temperature and the outlet coolant temperature exceed the set low value, the return branch pipeline can be started. When the temperature fluctuates too rapidly or the set value is too high, the main circulation cooling of the main outlet pipe 4 is activated. Simultaneously, dynamic corrections are made based on multi-dimensional data such as chip temperature trends, load rate, and return port temperature. For example, the path is switched in advance when a temperature rise is predicted, and the threshold is tightened under high load. At the execution level, a solenoid three-way valve responds quickly, and the opening of branch pipe valve 2 and main pipe valve 3 is controlled by a PWM signal to achieve transitional states of fully open branch pipes, fully open main circulation, or proportional adjustment. An electric throttle valve and differential pressure sensor ensure flow balance. Furthermore, a status check is performed every 500ms, comparing the actual temperature of the return port (lower inlet) with the expected value to correct the action. If a sensor malfunction or valve abnormality occurs, a safety mode is immediately triggered and an alarm is issued to ensure stable system operation. The coolant flows back directly through the return branch pipe: from the outlet of the immersion liquid cooling cabinet 1, branch pipe 7, return port, and re-entering the immersion liquid cooling cabinet 1 to contact the chip again. The returning coolant has a temperature difference with the chip, continuing to absorb heat (while the chip is still slowly generating heat), completing secondary cooling and preventing heat waste due to idle coolant. When the coolant absorbs a large amount of heat, the temperature rises from the initial value to exceed the threshold. The temperature sensor transmits the chip and outlet temperature signals to the control unit. The control system determines that the coolant temperature is too high, and direct recirculation will lead to insufficient chip heat dissipation (too small a temperature difference, resulting in reduced heat absorption efficiency). The control system issues a command: close branch pipe valve 2 and open main pipe valve 3, allowing the coolant to enter the traditional cooling process: from the outlet, heat exchanger, return port, and immersion liquid cooling cabinet 1, to reabsorb heat from the chip, ensuring effective cooling.

[0138] In the above embodiments, this embodiment adds a branch to the traditional coolant outlet pipe to directly return the coolant to the server, where it exchanges heat with the chip again, removing the heat generated by the chip. The chip temperature and the coolant temperature at the outlet are monitored in real time. The control system automatically opens the branch valve, reducing the operation of the heat exchanger and related cooling equipment, lowering energy consumption, improving the system's energy efficiency, and effectively reducing operating costs in data center and other application scenarios. By dynamically judging the coolant temperature and switching the circulation path, the start-up frequency of the heat exchanger can be reduced during low-load periods, or even completely shut down. In scenarios with large server load fluctuations, the energy consumption of the heat exchange system can be reduced by 30%-60%, corresponding to savings of hundreds of thousands to millions of kilowatt-hours of electricity annually in large data centers, significantly reducing the PUE value and helping data centers achieve energy conservation and carbon reduction goals. The entire circulation system requires three key components to work together. First, precise monitoring is achieved by deploying high-precision temperature sensors on the chip surface and at the liquid outlet to capture temperature changes in real time, providing data for path switching. Second, intelligent control is implemented. After receiving temperature signals from the sensors, the system determines whether to open a branch circuit according to preset logic, typically using a PLC or microcontroller to achieve millisecond-level valve control response. Finally, a reasonable pipeline design is crucial. Branch circuits must form a parallel structure with the main circulation path, achieving seamless switching through three-way valves. Simultaneously, the pipeline diameter, length, and flow distribution must be calculated to avoid flow field disturbances caused by resistance changes, ensuring that the coolant evenly covers the chip surface regardless of whether the main circulation or branch circuit is running, preventing localized overheating and avoiding wasted cooling capacity. Through the logic of "temperature triggering - path switching - dynamic circulation," the system reduces ineffective operation of the heat exchanger while ensuring heat dissipation reliability, thereby lowering system energy consumption.

[0139] The return flow branch design in this embodiment reduces the operating time of the heat exchanger and related pumps and valves, lowers the probability of wear on mechanical components, thereby extending equipment maintenance cycles and reducing losses from downtime maintenance. Simultaneously, the direct return flow of the cryogenic coolant avoids the over-cooling problem that may occur in traditional full-time cooling, reducing thermal stress on chips caused by sudden temperature rises and falls, and indirectly improving the lifespan of server hardware. Flexibility and adaptability: It can adapt to different load fluctuation scenarios. Whether it's full-load cooling during the day or energy-saving cycling during the night, the system can automatically switch through intelligent control without manual intervention. The branch design has strong compatibility with the modification of existing liquid cooling systems. Upgrades can be achieved simply by adding pipes, sensors, and control units, lowering the threshold and cost of technology implementation. Environmental protection: By reducing energy consumption, it indirectly reduces the data center's dependence on grid power, especially in areas where thermal power is the main source of electricity, reducing emissions of greenhouse gases such as carbon dioxide.

[0140] Figure 7 A block diagram of an exemplary electronic device suitable for implementing embodiments of the present invention is shown.

[0141] The electronic device may include a central processing unit / microprocessor / main control chip, etc. 8; and a storage medium 9, coupled to the central processing unit / microprocessor / main control chip, etc. 8, and storing computer-executable instructions therein for performing the steps of various methods of embodiments of the present invention when executed by the processor.

[0142] The central processing unit / microprocessor / main control chip, etc., may include, but are not limited to, one or more processors or microprocessors.

[0143] Storage medium 9 may include, but is not limited to, random access memory (RAM), read-only memory (ROM), flash memory, EPROM memory, EEPROM memory, registers, computer storage media (e.g., hard disk, floppy disk, solid-state drive, removable disk, CD-ROM, DVD-ROM, Blu-ray disc, etc.).

[0144] In addition, the electronic device may also include (but is not limited to) a data bus 10, an input / output bus / external bus / device bus 11, a display 12, and input / output devices 13 (e.g., keyboard, mouse, speaker, etc.).

[0145] The central processing unit / microprocessor / main control chip, etc., 8 can communicate with external devices (12, 13, etc.) via I / O bus 11 through wired or wireless network (not shown).

[0146] The storage medium 9 may also store at least one computer-executable instruction for performing the steps of various functions and / or methods in the embodiments described herein when the central processing unit / microprocessor / main control chip, etc., 8 is running.

[0147] In one embodiment, the at least one computer-executable instruction may also be compiled into or comprise a software product, wherein one or more computer-executable instructions are executed by a processor to perform the steps of the various functions and / or methods in the embodiments described herein.

[0148] Figure 8 A schematic diagram of a computer-readable storage medium according to an embodiment of the present invention is shown.

[0149] like Figure 8As shown, instructions, such as computer-readable instructions 14, are stored on the non-transitory computer-readable storage medium 15. When the computer-readable instructions 14 are executed by a processor, the various methods described above can be performed. The non-transitory computer-readable storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-transitory non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. For example, the non-transitory computer-readable storage medium 15 can be connected to a computing device such as a computer, and then, when the computing device executes the computer-readable instructions 14 stored on the computer-readable storage medium 15, the various methods described above can be performed.

[0150] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0151] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0152] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0153] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the methods of the various embodiments of this invention through a computer device (which may be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0154] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for circulating cooling of a single-phase immersion liquid-cooled server, characterized in that, Includes the following steps: The coolant flows through the chip in the immersion liquid cooling cabinet. The real-time temperature of the coolant is compared with a preset temperature threshold. When the real-time temperature does not exceed the preset threshold, the control system issues the first program command to start the circulation mode in the immersion liquid cooling cabinet. When the real-time temperature reaches and exceeds the preset threshold, the control system issues the second program command to start the heat exchange circulation mode. After the heat exchange cycle mode is started, the main outlet pipe is opened; at the same time, the preset threshold is dynamically corrected by combining the chip temperature trend, load rate and return port temperature of the immersion liquid cooling cabinet; the opening degree of the branch pipe valve and the main pipe valve is controlled by the PWM signal to realize the transition state of fully open branch, fully open main cycle or proportional adjustment. The coolant flows back to the outlet of the immersion liquid-cooled cabinet through the return branch pipe. The real-time temperature of the coolant is then compared with the preset temperature threshold. Based on the comparison results, the internal circulation mode and heat exchange circulation mode of the immersion liquid-cooled cabinet are switched.

2. The single-phase immersion liquid-cooled server circulating cooling method as described in claim 1, characterized in that, The process of achieving a transition state of fully open branch circuits, fully open main circulation, or proportional adjustment includes the following steps: The system receives feedback of temperature contact point coordinates, real-time chip load rate and return port temperature, generates a time axis extension vector based on temperature contact point coordinates, converts real-time chip load rate into intensity modulation coefficient, constructs a spatial compensation scalar based on return port temperature, and outputs dynamic deformation threshold boundary. Phase lag compensation is applied to the dynamic deformation threshold boundary. When the boundary contraction rate exceeds the critical value, a fully open trigger pulse is generated. When the boundary fluctuation frequency is in the transition zone, a proportional adjustment waveform is output to form the PWM fundamental wave sequence. The historical flow ratio of the total outflow pipeline is extracted from the PWM fundamental sequence as the inertial weight. The transient compensation is calculated based on the current deformation threshold boundary and the temperature difference at the return port, and the valve opening combination weight is synthesized. By combining the valve opening weights, the fully open trigger pulse drives the branch valves to form a zero-resistance channel, and the proportional adjustment waveform matches the main pipeline valve to generate a gradual throttling surface, thereby achieving three-state output: fully open branch, fully open main circulation, or proportional transition.

3. The single-phase immersion liquid-cooled server circulating cooling method as described in claim 2, characterized in that, The process of synthesizing the combined weights of valve openings includes the following steps: The periodic performance consumption envelope is extracted from the historical flow data of the total liquid outlet pipeline. The inertial strong and weak oscillation sequence is generated by the flow difference operation between adjacent time periods. The inertial strong and weak oscillation sequence is input into the decay memory filter and the inertial weighted basis is output. The real-time difference between the dynamic deformation threshold boundary and the return port temperature is received, processed by the spatial thermo-pressure converter, and a transient compensation vector field is generated with the deformation threshold boundary curvature as the field strength gradient reference and the absolute value of the temperature difference as the action intensity scalar. An inertial weighted base and a transient compensation vector field are projected along the pipe axis to form a steady-state skeleton. Transient compensation vector field components are injected at the nodes of the steady-state skeleton to form a weighted grid with compensation nodes. The weighted mesh with compensation nodes is reconstructed by the valve response adapter, the free flow domain of the segment corresponding to the fully open trigger pulse is extracted, the mesh density distribution is adjusted according to the waveform slope, and finally the valve opening combination weight is output.

4. The single-phase immersion liquid-cooled server circulating cooling method as described in claim 3, characterized in that, The process of forming a weighted grid with compensated nodes includes the following steps: Using the centerline of the main outlet pipeline as the reference axis, the inertial weight base is axially stretched and transformed according to the direction of fluid motion to form a tubular weight frame, and equidistant skeleton control nodes are set on the surface of the tubular weight frame. The three-dimensional transient compensation vector field is compressed radially along the pipeline, the projection intensity spectrum of the vector field in the axial direction is extracted, the vortex component magnitude of the vector field in the circumferential direction is captured, and a two-dimensional compensation action surface is generated. The two-dimensional compensation action surface is spatially phase matched with the skeleton control nodes. The compensation intensity base of each node is determined based on the projected intensity spectrum. The coupling action between adjacent nodes is allocated according to the vortex component magnitude, forming an enhanced node set with energy markers. An energy-labeled enhanced node set is used as the input field constraint generator. A tubular weighted frame is used as the initial mesh topology. Mesh density expansion is generated at the nodes according to the compensation intensity base. Non-uniform connection ribs are generated between the nodes according to the coupling action, and finally a weighted mesh with compensated nodes is constructed.

5. The single-phase immersion liquid-cooled server circulating cooling method as described in claim 4, characterized in that, The process of forming an energy-tagged set of augmented nodes includes the following steps: The amplitude distribution of the projected intensity spectrum is spatially bound to the axial position of the skeleton control node. The zero point of the reference axis is used as the origin of the spectrum-axis alignment. The pipe length is divided equally according to the intensity spectrum wavelength to generate a spectrum node mapping table. The spectral node mapping table is input into the intensity converter. The logarithmic attenuation of the spectral amplitude corresponding to each node is taken as the base matrix. The curvature change rate of the deformation threshold boundary at the node is superimposed, and the dynamic compensation base matrix is ​​output. The captured vortex component magnitude is processed by the circulation decomposer, and the phase difference angle between the peak values ​​of the magnitude is extracted to generate the interaction coefficient of adjacent nodes. The magnitude envelope area and the filling ratio of the pipe section are obtained to form a set of potential transfer factors. The core of the input field is the fusion of the dynamic compensation base matrix and the energy potential transfer factor set. The elements of the compensation base matrix are used as node energy cores. The energy cores are extended to adjacent nodes according to the action coefficient to form radiation energy bands. The width of the radiation energy bands is adjusted according to the filling ratio, and finally, an enhanced node set with energy labels is generated.

6. The single-phase immersion liquid-cooled server circulating cooling method as described in claim 5, characterized in that, The process of obtaining the modulus envelope area and the filling ratio of the pipe cross-section includes the following steps: The captured vortex component magnitude waveform is input into the extreme value topology analyzer to identify the effective peak group set whose magnitude exceeds the critical amplitude, record the azimuth coordinates of each peak in the pipe circumferential coordinate system, and generate a peak group azimuth distribution map. The peak group azimuth distribution map is processed by the relative motion solver to calculate the tangential offset of the azimuth angle of adjacent peaks, convert the offset into the vector angle with the center of the pipe as the origin, and output the node action angle matrix. The modulus envelope area is input into the geometric constraint converter. The cross-sectional area of ​​the pipe is used as the reference constraint box. The dynamic overlap rate between the envelope area and the reference constraint box is calculated, and the vortex core occupancy factor is generated. The node action angle matrix and the vortex core occupancy factor are input into the fluid interference synthesizer. The tangent of the vector angle is used as the action intensity coefficient. The effective action radius of the action intensity coefficient is modulated by the vortex core occupancy factor, and finally the energy potential transfer factor set is formed.

7. The single-phase immersion liquid-cooled server circulating cooling method as described in claim 1, characterized in that, The process of switching between the internal circulation mode and the heat exchange circulation mode of the immersion liquid-cooled cabinet based on the comparison results includes the following steps: The coolant temperature data obtained from the return branch pipe is processed by noise suppression and time domain alignment to form a pure return waveform. The pure return waveform is then input into the phase analyzer to output the return phase marker. The return phase marker and the current operating cooling mode state input migration criterion synthesizer form a dual-mode migration trigger field; When in heat exchange mode: the amplitude of the return waveform remains below the dynamic deformation threshold boundary, generating an internal circulation enabling factor; when in internal circulation mode: the peak of the return waveform contacts the deformation threshold boundary, triggering a heat exchange activation factor. The dual-mode migration trigger field input physical field mapping module drives the closure of the main liquid outlet pipe valve to form a closed loop topology. The heat exchange activation factor triggers the branch valve zero-resistance channel to establish an open flow channel configuration, completing the cooling mode switching.

8. The single-phase immersion liquid-cooled server circulating cooling method as described in claim 7, characterized in that, The process of switching cooling modes includes the following steps: The inner loop enables the input constraint field generator, extracts the factor strength value as the closure strength coefficient, associates the free flow domain closure sequence of the weighted grid, and generates the main pipeline valve lockout instruction set; The main pipeline valve lockout command set, under the action of the fluid path constraint field, forms a closed loop topology with the submerged cavity wall as the fixed boundary and the return branch pipeline as the unidirectional conduction axis; The heat exchange activation factor triggers the zero-resistance channel builder, activates the spatiotemporal coordinates of the fully open trigger pulse, calls the energy distribution of the compensated nodes of the weighted grid, and generates the expansion wavefront of the branch valve. The closed-loop topology and open-channel configuration are output to the thermal inertia balance monitor to detect the degree of agreement between the temperature gradient and the heat flow tendency vector in the cavity, verify that the return pure waveform conforms to the phase characteristic spectrum of the target mode, and confirm that the cooling mode switching is completed.

9. A single-phase immersion liquid-cooled server circulating cooling system, used to implement the single-phase immersion liquid-cooled server circulating cooling method as described in any one of claims 1 to 8, characterized in that, Includes: immersion liquid cooling cabinet, branch pipeline valves, main pipeline valves, main liquid outlet pipeline, pump, heat exchanger and branch pipeline; The immersion liquid-cooled cabinet is equipped with at least one liquid outlet and two liquid inlets. The liquid outlet is installed at the upper end of one side of the immersion liquid-cooled cabinet, and the two liquid inlets are located at the middle and lower ends of one side of the immersion liquid-cooled cabinet. The liquid inlet at the middle position is connected to the main liquid outlet pipe through a branch pipe and a branch pipe valve. A main pipe valve is installed on the main liquid outlet pipe. The outlet end of the main liquid outlet pipe is connected to the inlet end of the pump. The outlet end of the pump is connected to the inlet end of the heat exchanger. The outlet end of the heat exchanger is connected to the liquid inlet at the lower end through a return branch pipe.

Images