A high-performance liquid-cooled server system and device

Abstract

The application discloses a kind of high-performance liquid cooling server systems and equipment, comprising: sealed module, inside filling cooling fluid containing self-repairing microcapsule;Heat exchange module, adhere to heat chip, inside is equipped with adjustable flow channel structure and surface has porous composite coating;Fluid drive module;Sensing module detects temperature, pressure, dielectric constant and flow rate;Control module, according to sensing data predicts thermal field distribution, adjusts pump speed and flow channel geometry when predicting will overheat, simultaneously according to sensing data judges leakage and controls microcapsule breakage crosslinking into gel sealing.The above module is integrated in independent capsule type package, only retains power supply and data interface.The application significantly improves thermal management response speed, system reliability and severe environment adaptability by predictive active regulation, dynamic flow channel adaptation and self-repairing leakage sealing.

Description

Technical Field

[0001] This invention belongs to the field of liquid-cooled server heat dissipation technology, and particularly relates to a high-efficiency liquid-cooled server system and equipment. Background Technology

[0002] Liquid-cooled server systems are the mainstream solution for heat dissipation of high-power-density chips. Existing liquid-cooled server systems typically employ a split architecture, including an external cooling distribution unit, connecting pipes, and an internal cold plate. The cold plate often uses a fixed structure with parallel microchannels or spiral flow channels, with a pump driving the coolant to flow across the cold plate to remove heat from the chips. Some systems add nanoparticles to the coolant to enhance thermal conductivity or fabricate microstructures on the cold plate surface to promote boiling heat transfer. Regarding sensors, existing systems typically include temperature sensors and flow switches to monitor coolant temperature and pump operating status. Control strategies are mostly proportional-integral-derivative (PID) control, adjusting pump speed based on outlet water temperature feedback. Furthermore, some high-end liquid-cooling systems employ dual-pump redundancy designs to improve reliability and seal pipe joints to reduce leakage risks. These solutions are widely used in data center applications and can meet the heat dissipation requirements of conventional environments.

[0003] However, existing liquid-cooled server systems still have the following shortcomings: First, thermal management uses hysteresis feedback control. When the chip power increases dramatically, the pump speed is adjusted only after the temperature rises, and the response delay causes the chip junction temperature to briefly exceed the limit, triggering frequency reduction protection. Second, the cold plate flow channel is fixed and cannot adapt to dynamic hot spots caused by changes in chip power distribution or aging. Local hot spots are difficult to eliminate, and the temperature difference on the chip surface is large. Third, the system lacks active leak defense capabilities. The large number of pipes and joints means that leaks can only be detected manually. Coolant leakage may cause short circuits, and the system cannot recover automatically after a leak. Fourth, the split structure results in exposed pipes, low overall protection level, and inability to deploy in harsh environments such as vehicles, fields, and industrial sites. Fifth, a single pump failure can cause the entire system to shut down. Although there are two redundant pumps, there is a lack of intelligent rotation and health management. Sixth, changes in coolant volume with temperature cause internal pressure fluctuations, affecting pump suction efficiency and boiling stability. To address these issues, there is an urgent need for an integrated, intelligent, and self-healing high-efficiency liquid-cooled server system. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a high-efficiency liquid-cooled server system and equipment, which solves the problems of delayed thermal response, fixed flow channels that cannot adapt to dynamic hot spots, lack of leakage self-repair capability, low protection level and insufficient redundancy in the prior art liquid-cooled server system.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A high-performance liquid-cooled server system and device, comprising:

[0007] A sealed module, internally filled with coolant, wherein self-healing microcapsules are dispersed in the coolant;

[0008] A heat exchange module is disposed within the sealing module for attaching the heating chip. The heat exchange module has an adjustable flow channel structure inside and a porous composite coating on its surface.

[0009] A fluid drive module, disposed within the sealing module, is used to drive the coolant to flow through the heat exchange module;

[0010] A sensing module, disposed within the sealed module, is used to detect temperature, pressure, dielectric constant, and flow rate;

[0011] A control module is disposed within or adjacent to the sealing module, and the control module is electrically or communicatively connected to the adjustable flow channel structure, the fluid drive module, and the sensing module, respectively.

[0012] The control module receives the detection data from the sensing module and predicts the thermal field distribution of the heating chip based on the detection data. When the predicted thermal field distribution indicates that the temperature will exceed a preset threshold within a preset future time period, the control module sends a speed adjustment command to the fluid drive module and a geometry adjustment command to the adjustable flow channel structure to suppress the temperature from exceeding the preset threshold. The control module also determines whether a leak has occurred based on the detection data from the sensing module, and when a leak is detected, controls the self-healing microcapsules to rupture and cross-link to form a gel to seal the leak.

[0013] Preferably, the adjustable flow channel structure includes multiple flexible fins driven by actuators, each of which can deflect independently; the heat exchange module is also provided with a flow channel of non-uniform width, the flow channel density in the preset high heat zone is greater than that in the low heat zone, and the high heat zone is provided with jet holes or vortex cavities.

[0014] Preferably, the porous composite coating comprises a micron-sized copper pillar layer, a nanoporous alumina layer, and a fluorinated organic molecular layer stacked sequentially, and the contact angle between the coating and the coolant is less than 10 degrees.

[0015] Preferably, the base liquid of the coolant is a fluorinated liquid, and the coolant also contains heat transfer enhancing particles, which are graphene nanosheets with a hydrophilic polymer attached to their surface; the capsule wall of the self-healing microcapsule is polyurea or polyurethane, and the capsule contains isocyanate prepolymer and polyamine. When the contents of the microcapsule rupture and come into contact with air or moisture, cross-linking polymerization occurs.

[0016] Preferably, the sensing module includes: a temperature sensing unit, a pressure sensing unit, a dielectric constant sensing unit, a flow velocity sensing unit, and an acceleration sensing unit.

[0017] Preferably, the control module is further configured to: when the leakage probability calculated based on the pressure pulsation spectrum characteristics and dielectric constant fluctuation characteristics exceeds a first preset value, send a deceleration command to the fluid drive module and a start command to the mechanical clamping unit inside the sealing module housing; when the leakage probability exceeds a second preset value and the detected pressure drop rate exceeds a threshold, send a stop command to the fluid drive module and control the heating unit disposed in the sealing module to generate pulse heat, causing the self-healing microcapsule to rupture; wherein the leakage probability is obtained by comparing the pressure pulsation spectrum characteristics and the dielectric constant fluctuation characteristics with a preset leakage determination rule.

[0018] Preferably, the sealing module further includes a compensation chamber, which houses a flexible pressure regulating bladder filled with inert gas to maintain positive pressure within the sealing module.

[0019] Preferably, the fluid drive module includes at least two drive units connected in parallel, and the control module controls the drive units to operate in turn at a power value less than or equal to the rated power, and switches to the other drive unit when an abnormal current or flow rate is detected in any of the drive units.

[0020] Preferably, in a high-efficiency liquid-cooled server device, the sealing module, the heat exchange module, the fluid drive module, the sensing module, and the control module are all integrated into an independent capsule-shaped package. The capsule-shaped package only has a power supply interface and a data interface externally, with no exposed liquid pipelines.

[0021] Preferably, the capsule-type package is composed of a metal inner liner and a fiber-reinforced composite material outer shell. The overall protection level of the capsule-type package is IP68, the operating temperature range is -40℃ to 105℃, and it can withstand 20g of vibration. The capsule-type package is also equipped with a self-sealing supply valve.

[0022] The technical effects and advantages of this invention's high-performance liquid-cooled server system and equipment are as follows:

[0023] 1. This invention utilizes the collaborative operation of a heat exchange module, a fluid drive module, a sensing module, and a control module within a sealed module. The control module can predict the thermal field distribution of the heating chip based on real-time temperature, pressure, dielectric constant, and flow rate data detected by the sensing module. When it predicts that the temperature will exceed a preset threshold, it proactively adjusts the rotational speed of the fluid drive module and the geometry of the adjustable flow channel structure within the heat exchange module. Compared to traditional hysteresis control relying on temperature feedback, this predictive proactive adjustment method can complete the distribution of cooling capacity before thermal shock occurs, effectively suppressing chip temperature exceeding limits and avoiding frequency throttling or overheating protection due to response delays. This significantly improves the system's thermal management response speed and temperature control accuracy.

[0024] 2. This invention disperses self-healing microcapsules in the coolant. The control module determines the leakage probability based on the pressure pulsation spectrum and dielectric constant fluctuation characteristics detected by the sensor module. When a leak is detected, the microcapsules rupture, causing the encapsulated material inside to cross-link and form a gel to seal the leak. This mechanism achieves a fully autonomous protection chain from leak prediction and graded defense to active coagulation. It can seal tiny cracks within milliseconds to seconds without manual intervention, fundamentally solving the long-standing leakage safety hazards of liquid-cooled servers and greatly improving the system's operational reliability and maintenance safety.

[0025] 3. The heat exchange module in this invention features an adjustable flow channel structure, including multiple flexible fins driven by actuators, each capable of independent deflection. The control module can dynamically adjust the local flow channel cross-section and flow resistance by driving the fins in different areas to change their deflection angles based on real-time detected thermal field distribution. Compared to traditional fixed-channel cold plates, the flow channel geometry of this invention can be reconstructed online, precisely matching dynamic hot spots generated by changes in chip power distribution or aging, effectively eliminating localized hot spots, reducing chip surface temperature differences, and extending chip lifespan.

[0026] 4. This invention integrates the sealing module, heat exchange module, fluid drive module, sensing module, and control module into a single, independent capsule-shaped package. The capsule only has power and data interfaces externally, with no exposed liquid piping. The capsule-shaped package's outer shell is constructed from a metal inner liner and a fiber-reinforced composite material outer shell, achieving an overall protection rating of IP68, a wide operating temperature range, and resistance to strong vibrations. This fully sealed, piping-free design enables the system to be deployed in harsh environments beyond traditional data centers, including vehicle-mounted, field, industrial sites, and deep-sea scenarios, significantly expanding the application boundaries of liquid-cooled servers.

[0027] 5. In this invention, the fluid drive module includes at least two parallel drive units. The control module controls each drive unit to operate in turn at a power value less than or equal to its rated power, and automatically switches to another drive unit when an abnormal current or flow rate is detected in any drive unit. This redundant design and intelligent switching strategy balances the wear of each drive unit, avoids system downtime caused by single-point failure, and ensures continuous cooling capacity in the event of drive unit failure or performance degradation, significantly improving system availability and lifespan.

[0028] 6. This invention incorporates a compensation chamber within the sealing module, housing a flexible pressure regulating bladder filled with inert gas. This chamber absorbs the volume expansion or contraction of the coolant caused by temperature changes, maintaining a stable micro-positive pressure within the sealing module. This design prevents abnormal precipitation of dissolved gases and bubble formation in the coolant under low-pressure conditions, ensuring the intake efficiency of the fluid drive module and the boiling stability of the heat exchange module, further optimizing the long-term operational consistency of the system.

[0029] 7. The sensing module in this invention includes multiple sensing units such as temperature, pressure, dielectric constant, flow rate, and acceleration. The control module can calculate the leakage probability by comprehensively considering the pressure pulsation spectrum characteristics and dielectric constant fluctuation characteristics, and execute graded responses (such as deceleration, mechanical clamping, stopping the drive, and triggering coagulation) when the leakage probability exceeds different preset values. This multimodal predictive maintenance mechanism can issue early warnings several hours before a leak actually occurs and take progressive defensive measures, upgrading leakage accidents from "post-event remediation" to "pre-event immunity," greatly reducing the probability of catastrophic leakage events. Attached Figure Description

[0030] Figure 1 This is a system composition block diagram of a high-performance liquid-cooled server system and equipment proposed in this invention;

[0031] Figure 2 This is a control flowchart of a high-performance liquid-cooled server system and equipment proposed in this invention;

[0032] Figure 3 This is a schematic diagram of the adjustable flow channel structure of a high-efficiency liquid-cooled server system and equipment proposed in this invention. Detailed Implementation

[0033] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0034] 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 entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include," "contain," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "includes..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0035] refer to Figure 1-3 This invention discloses a high-efficiency liquid-cooled server system and device, aiming to solve the problems of slow thermal response, inability to adapt to dynamic hotspots, lack of self-healing leakage capability, and poor environmental adaptability of existing liquid cooling technologies. The technical solution includes: a sealing module filled with coolant containing self-healing microcapsules; a heat exchange module housed within the sealing module and fitted with a heating chip, featuring an adjustable flow channel structure and a porous composite coating on its surface; a fluid drive module within the sealing module that drives the coolant to flow through the heat exchange module; a sensing module that detects temperature, pressure, dielectric constant, and flow rate; and a control module that predicts the thermal field distribution based on the sensing data, adjusts the pump speed and flow channel geometry when overheating is predicted, and simultaneously determines leakage based on the sensing data and controls the microcapsules to rupture and cross-link into a gel for sealing. The sealing module, heat exchange module, fluid drive module, sensing module, and control module are integrated into an independent capsule-type package, retaining only the power supply and data interfaces. This invention significantly improves thermal management response speed, system reliability, and adaptability to harsh environments through predictive active control, dynamic flow channel adaptation, and self-healing leak sealing.

[0036] Example 1

[0037] Basic type of biomimetic self-healing liquid-cooled capsule.

[0038] Purpose of implementation:

[0039] This embodiment aims to provide a liquid-cooled server device that integrates predictive thermal management and leak self-healing functions to solve the problems of slow response, high leakage risk, and poor environmental adaptability of traditional split liquid cooling systems.

[0040] Implementation System:

[0041] The system in this embodiment includes a sealing module, a heat exchange module, a fluid drive module, a sensing module, and a control module, all integrated into a single capsule-shaped package. The capsule-shaped package is rectangular, 120mm long, 80mm wide, and 15mm high. The outer shell is composed of a metal inner liner (6061 aluminum alloy, 0.6mm thick) and a fiber-reinforced composite outer shell (carbon fiber reinforced PEEK, 1.2mm thick). The overall protection rating is IP68, the operating temperature range is -40℃ to 105℃, and it can withstand 20g vibration. The capsule only has a power supply interface (48VDC) and a data interface (…). There are no exposed liquid pipelines. A self-sealing refill valve is located on the side of the capsule.

[0042] The sealing module is a closed cavity inside the capsule, with a volume of approximately 80 mL, filled with coolant. The coolant base is perfluorotriethylamine, which contains dispersed heat transfer enhancing particles (graphene nanosheets grafted with polyethylene glycol, lateral dimension 300 nm, thickness 3 nm, concentration 0.1 wt%) and self-healing microcapsules (capsule walls are made of polyurea, wall thickness 2 μm, average diameter 20 μm, encapsulated with isocyanate prepolymer and polyamine, mass fraction 1%).

[0043] The heat exchange module is an oxygen-free copper-based cold plate, measuring 70mm × 60mm × 4mm, and is attached to the back of the CPU chip. The heat exchange module internally features non-uniformly wide flow channels: the main flow channel is 2.0mm wide, and the branch flow channels are 0.5mm wide. In the chip's pre-designed high-heat zone, the flow channel density is three times that of the low-heat zone. The high-heat zone also features jet orifices (0.2mm in diameter) and vortex cavities. Eight flexible fins (monocrystalline silicon, 0.1mm thick) are integrated within the flow channels. Each fin is driven by an independent electrothermal actuator and can deflect from 0° to 30°. The surface of the heat exchange module is coated with a porous composite layer: a micron-level copper pillar layer (10μm high, 20μm spacing), a nanoporous alumina layer (100nm pore size), and a fluorinated organic molecular layer. The contact angle between the coating and the coolant is 5°.

[0044] The fluid drive module includes two miniature piezoelectric pumps connected in parallel (model number...). (Dimensions: 15×15×5mm). The inlets of the two pumps are immersed in the coolant, and the outlets are connected to the inlet of the heat exchange module via a three-way valve; the outlet of the heat exchange module leads to the internal space of the sealed module.

[0045] The sensing module includes: six thin-film thermopile temperature sensors (accuracy ±0.1℃, sampling rate 100Hz), two MEMS silicon piezoresistive pressure sensors (range 0~100kPa), one interdigital electrode dielectric constant sensor (excitation frequency 1MHz), one ultrasonic flow velocity sensor (2MHz), and one triaxial MEMS accelerometer (range ±50g).

[0046] The control module is a microcontroller (STM32U5) with a main frequency of 160MHz and built-in memory. The control module is electrically connected to the flexible fin actuator, piezoelectric pump, and all sensors. The control module's memory contains a pre-stored thermal characteristic model of the chip, a linear recursive prediction algorithm (temperature field in the next second), and leakage detection rules (including pressure drop rate threshold and dielectric constant change rate threshold).

[0047] The sealing module also has a compensation chamber (5 mL volume) with a built-in stainless steel flexible corrugated bladder filled with nitrogen (10 kPa pressure) to maintain positive pressure inside the sealing module.

[0048] Implementation steps:

[0049] After the system is powered on, the control module reads all sensor data at a 200ms cycle and predicts the chip junction temperature distribution within the next second based on the chip thermal characteristic model.

[0050] If it is predicted that the chip junction temperature will exceed the preset threshold of 85°C within the next second, the control module sends a speed adjustment command (PWM step +5%) to the fluid drive module, and at the same time sends a deflection command (increase 10°) to the flexible fin actuator in the corresponding hot spot area to increase the local turbulence intensity.

[0051] The control module calculates the leakage probability every second. The leakage probability is obtained by comparing the pressure pulsation spectrum characteristics and dielectric constant fluctuation characteristics with preset leakage judgment rules. If the leakage probability exceeds the first preset value of 0.7, a deceleration command is sent to the fluid drive module (reduced to 50% of the rated speed), and the shape memory alloy ring inside the capsule shell is activated for mechanical clamping. If the leakage probability exceeds the second preset value of 0.95 and the pressure drop rate is detected to exceed 10 kPa / 10 ms, a stop command is sent to the fluid drive module, and the heating unit (thin-film resistor) set in the sealing module is controlled to generate pulse heat (50V, 10 ms), causing the self-healing microcapsule to rupture. The released contents cross-link upon contact with air to form a gel, sealing the leak.

[0052] The two piezoelectric pumps of the fluid drive module are controlled by the control module to operate alternately at 80% of their rated power, with each pump working continuously for 30 minutes before switching. If the current of either pump deviates from the rated value by ±25% or the flow rate is lower than 70% of the rated flow rate of a single pump, the control module immediately switches to the other pump.

[0053] Implementation results:

[0054] Thermal shock testing was conducted on the capsule of this embodiment: the CPU power jumped from 100W to 300W, and the control module predicted overheating within 180ms and adjusted accordingly. The actual highest junction temperature was 82.5℃, and the temperature protection was not triggered. Micropores with a diameter of 0.1mm were created using a needle-punching method. After leakage, the pressure dropped by 12kPa within 15ms, the coagulation mode was activated, and the gel sealed the crack within 200ms. No leakage occurred for the subsequent 48 hours. The capsule operated continuously for 72 hours in an environment ranging from -40℃ to 105℃, demonstrating stable performance.

[0055] Example 2

[0056] An adaptive cold plate system with variable flow channel topology.

[0057] Purpose of implementation:

[0058] This embodiment aims to achieve more precise hot spot elimination and reduce the temperature difference on the chip surface by dynamically reconstructing the adjustable flow channel structure for AI chips (such as GPUs) with multiple non-uniform heat sources.

[0059] Implementation System:

[0060] This embodiment is basically the same as Embodiment 1, except that the heat exchange module does not use a fixed non-uniform width flow channel, but instead uses a parallel flow channel with a uniform width (1.0 mm). Multiple flexible fins (24 in total) are installed in each flow channel, and each fin is driven by an independent actuator with an initial angle of 0°. The control module embeds a lightweight reinforcement learning model (DQN), with a state space of a 16×16 temperature grid and an action space of a discrete set of 24 fin angles. The remaining modules (sealing module, coolant, fluid drive module, sensing module, etc.) are the same as in Embodiment 1.

[0061] Implementation steps:

[0062] The sensing module collects the temperature distribution of various areas on the surface of the heat exchange module and forms a real-time heat map.

[0063] The control module inputs the heat map into the reinforcement learning model, and the model outputs the fin angle combination that minimizes the maximum hot spot temperature and increases pump power consumption by no more than 20%.

[0064] The control module sends angle commands to each flexible fin actuator. In some areas, the fins deflect to a large angle (e.g., 25°) to reduce the local flow channel cross-sectional area and increase the flow velocity and turbulence. In other areas, the fins maintain a small angle or zero angle.

[0065] The control module repeats the above steps at a 100ms cycle to dynamically adapt to changes in the chip's power distribution.

[0066] Implementation results:

[0067] A GPU chip with 8 independent cores (total power consumption 400W) was tested. A traditional uniform cooling plate resulted in a 15°C temperature difference between hot and cold spots. After adopting this embodiment, the temperature difference between hot and cold spots was reduced to 6°C, the pump power consumption increased by only 12%, and the temperature of all cores was below 82°C.

[0068] Example 3

[0069] Pump-free immersion self-healing edge computing system.

[0070] Purpose of implementation:

[0071] This embodiment aims to provide an edge computing device suitable for harsh environments such as the field and vehicles, requiring no moving parts, extremely high reliability, and self-healing capabilities.

[0072] Implementation System:

[0073] This embodiment is essentially the same as Embodiment 1, except that: no separate heat exchange module (cold plate) and fluid drive module are installed within the sealing module. The heat-generating chips (CPU and GPU) are directly immersed in the coolant, and the chip surface is covered with the same porous composite coating as in Embodiment 1. A condenser (aluminum fins, integrally formed with the capsule shell) is located at the bottom of the sealing module, and the upper part is a vapor space. The composition of the coolant is the same as in Embodiment 1, but the concentration of the self-healing microcapsules is increased to 2 wt%. The control module is only responsible for data acquisition from the sensing module and leak detection, and does not perform active flow regulation.

[0074] Implementation steps:

[0075] The chip heats up, causing the nearby coolant to boil and produce steam. The steam rises to the condenser and liquefies. The liquefied coolant then flows back to the bottom under gravity, forming a natural circulation.

[0076] The sensing module continuously monitors pressure, dielectric constant, and temperature. The control module calculates the leakage probability every 500ms.

[0077] When the leakage probability exceeds 0.95 and the detected pressure drop rate exceeds 5 kPa / 10 ms, the control module sends pulse heat to the heating unit in the sealing module, causing the nearby self-healing microcapsules to rupture and the cross-linked gel to seal the cracks.

[0078] The system outputs maintenance signals but continues to operate in degraded mode.

[0079] Implementation results:

[0080] This embodiment has no moving parts such as pumps or fans. At an ambient temperature of 55°C and a CPU power consumption of 150W, the junction temperature is controlled below 95°C. A drop test (1.5m height) generated microcracks in the casing; the self-healing mechanism sealed these cracks within 300ms, with no coolant leakage, and the system continued to operate stably for 72 hours. This equipment has passed vehicle vibration testing (ISO16750-3).

[0081] Example 4

[0082] Predictive leakage prevention and multi-level response system.

[0083] Purpose of implementation:

[0084] This embodiment aims to achieve early warning and tiered defense before leaks occur, upgrading leak incidents from "post-event remediation" to "pre-event immunity".

[0085] Implementation System:

[0086] This embodiment is basically the same as Embodiment 1, except that the control module pre-stores a leakage prediction model based on a long short-term memory network. The model input consists of the pressure pulsation spectrum characteristics (peak frequency and amplitude of 10-500Hz extracted by FFT) and dielectric constant fluctuation characteristics (standard deviation, skewness) over the past 300 time steps (10ms per step), and the output is the leakage probability (0-1). The control module is configured with a three-level response strategy and adds an accelerometer-triggered protection mode. The remaining modules are the same as in Embodiment 1.

[0087] Implementation steps:

[0088] The control module acquires pressure, dielectric constant, and acceleration data at 10ms intervals.

[0089] The LSTM model is run every 100ms to calculate the leakage probability.

[0090] If the leakage probability is ≥0.5 (first preset value), the control module sends an "imminent leakage" warning signal to the outside world through the data interface and sends a speed reduction command to the fluid drive module (reduced to 30% of rated power).

[0091] If the leakage probability is ≥0.7 (second preset value), in addition to the above operations, the shape memory alloy ring inside the capsule shell is activated to apply a radial clamping force of 50N to the shell.

[0092] If the leakage probability is ≥0.95 and the detected pressure drop rate is >5kPa / 10ms, then the coagulation mode is executed: the fluid drive module is stopped, and the heating unit pulse triggers the microcapsule rupture and gel sealing.

[0093] If the accelerometer detects an instantaneous impact greater than 10g, the control module automatically enters protection mode: suspends the fluid drive module, shuts down all actuators, maintains sensor monitoring, and automatically resumes operation after the impact disappears.

[0094] Implementation results:

[0095] Accelerated aging tests were conducted on the capsules (temperature cycling from -40℃ to 105℃, 1000 cycles). The LSTM model provided an average early warning time of 2.3 hours before crack propagation and penetration, achieving an early warning success rate of 96%. Compared to systems without early warning, the actual leakage accident rate was reduced by 80%. Mechanical pre-tightening reduced the crack propagation rate by approximately 40%.

[0096] Example 5

[0097] A swarm intelligence collaborative system in a high-density cluster.

[0098] Purpose of implementation:

[0099] This embodiment aims to achieve collaborative cooling among multiple liquid-cooled server devices (capsules), thereby improving the energy efficiency and thermal stability of the entire cluster through shared heat dissipation resources and dynamic load balancing.

[0100] Implementation System:

[0101] This embodiment includes multiple liquid-cooled server devices (capsules) as described in Embodiment 1, and a shared backplane. The control modules of each capsule communicate with each other via a backplane bus (I2C). The internal structure of the capsules is the same as in Embodiment 1. The backplane also has a shared heat dissipation backplane (containing a heat pipe array), and the bottom of each capsule contacts the heat dissipation backplane via thermal pads.

[0102] Implementation steps:

[0103] Each capsule's control module periodically (every 500ms) broadcasts its own information via the bus: current power load, power prediction for the next 2 seconds, current pump speed, and current hot spot temperature.

[0104] A master control module is determined in the cluster through a distributed election algorithm.

[0105] The main control module runs a consensus algorithm, which measures the total power and cooling capacity of the computer cabinet, and identifies the capsule that is at risk of overheating (the temperature is predicted to exceed 85°C within 1 second and its pump is already close to full load).

[0106] The main control module sends coordination commands to adjacent lightly loaded capsules, requesting them to temporarily reduce their pump speed (e.g., to 60%) so that more heat pipe resources on the shared heat sink backplate can be allocated to the overloaded capsules.

[0107] The overload capsule simultaneously increases its own pump speed to 100% and adjusts the angle of the flexible fins to enhance local heat dissipation.

[0108] Once the overload is relieved, the main control module sends a recovery command, and each capsule returns to independent control mode.

[0109] Implementation results:

[0110] Running an unbalanced AI training task in a cluster of 48 capsules, the maximum temperature difference of the entire cluster was reduced from 12℃ to 4℃ after adopting a collaborative strategy. The junction temperature of all capsules was below 80℃, and the total power consumption of the cluster pump was reduced by 22%. Compared with the system without collaboration, the probability of overload capsules triggering temperature protection was reduced from 15% to 0%.

[0111] Comparative Example 1

[0112] Traditional split-type liquid-cooled server system.

[0113] Purpose of implementation:

[0114] This comparative example is used to compare with the embodiments of the present invention in order to highlight the beneficial effects of the present invention.

[0115] Implementation System:

[0116] The comparative system employs a conventional indirect liquid cooling system, including: a rack-mounted cooling distribution unit (CDU) connected to a cold plate inside the server via external hoses. The cold plate is a conventional copper microchannel cold plate (50 parallel microchannels, 0.5mm wide and 2mm deep), with no adjustable flow channel structure. The pump is a single, constant-speed centrifugal pump (without redundancy). The coolant is pure deionized water, free of nanoparticles and self-healing microcapsules. The only sensing device is a water temperature probe (installed at the cold plate outlet). The controller is a simple PID controller that adjusts the pump speed solely based on the outlet water temperature feedback, lacking thermal field prediction and leak detection functions. The system has no sealed cavity; all piping and the cold plate are exposed inside the chassis.

[0117] Implementation steps:

[0118] After the system starts, the pump drives the deionized water to flow through the cold plate at a constant speed.

[0119] When the CPU load increases, the water temperature at the cold plate outlet rises. After the PID controller detects the temperature rise (with a lag of about 2-3 seconds), it gradually increases the pump speed.

[0120] The system lacks leak detection and self-repair capabilities. If a leak occurs at a pipe joint, coolant (deionized water) may drip onto the circuit board, causing a short circuit; this can only be detected through regular manual inspections.

[0121] Implementation results:

[0122] Under the same thermal shock test (CPU power jump from 100W to 300W), the PID controller only started increasing the pump speed after 2.5 seconds, during which the CPU junction temperature briefly reached 96℃, triggering frequency reduction protection. The system lacks leak self-healing capability; in one instance of a loose connector, the leak persisted for 20 minutes until it was manually discovered, causing a short circuit and damage to the motherboard. The system's operating environment is limited to a 10~35℃ computer room, making it unsuitable for vehicle or outdoor environments. A single pump failure will cause the entire system to shut down.

[0123] Embodiments 1 to 5 of the present invention demonstrate the thermal shock response speed ( In terms of leakage handling (automatic clotting vs. none), redundancy reliability (dual pumps + self-repair vs. none), and environmental adaptability (IP68 / -40~105℃ / 20g vs. computer room), it is significantly superior to traditional liquid cooling systems. The shortcomings of Comparative Example 1 precisely demonstrate the non-obviousness and practical advancement of the technical solution of this invention.

[0124] Compared to Examples 1-5 and Comparative Example 1, Examples 1-5 of the present invention are all based on the same integrated biomimetic self-healing liquid cooling architecture, while Comparative Example 1 adopts a traditional split-type liquid cooling solution. In terms of system structure, the examples integrate the sealing module, heat exchange module, fluid drive module, sensing module, and control module into a single independent capsule-shaped package, retaining only power supply and data interfaces externally, with no exposed liquid pipelines. The overall system achieves an IP68 protection rating and can operate stably in a wide temperature range of -40℃ to 105℃ and under 20g vibration. Comparative Example 1, on the other hand, uses an external cooling distribution unit connected to the internal cold plate of the server via flexible hoses. Both the pipelines and the cold plate are exposed outside the chassis, resulting in a lower protection level and suitability only for constant temperature server rooms. Regarding the cooling medium, the examples use a fluorinated liquid containing self-healing microcapsules and graphene nanosheets, possessing enhanced two-phase heat transfer and leak self-healing capabilities. Comparative Example 1 uses pure deionized water, which has no self-healing function, and leaks may directly lead to short circuits. In terms of control strategy, the embodiment is based on multimodal sensor fusion and preset judgment rules, which can predict the chip thermal field distribution and actively adjust the pump speed and flow channel geometry before overheating. At the same time, it can monitor the leakage probability in real time and trigger graded defense and coagulation sealing. In contrast, the first embodiment only relies on the feedback of the outlet water temperature for hysteretic PID adjustment, without thermal field prediction, leakage prediction, or active protection.

[0125] The specific performance differences are particularly significant. Regarding thermal shock response, in Example 1, when the CPU power jumped from 100W to 300W, the control module predicted overheating within 180ms and adjusted the pump speed and fin angle in advance, resulting in a maximum junction temperature of only 82.5℃ and no frequency reduction was triggered. In Comparison 1, under the same conditions, the PID delayed by 2.5 seconds before increasing the pump speed, and the junction temperature briefly exceeded 96℃, leading to frequency reduction protection. In terms of hot spot suppression, Example 2 reduced the temperature difference between hot and cold spots from 15℃ in traditional cold plates to 6℃ for multi-core GPUs, with only a 12% increase in pump power consumption. Comparison 1 lacked any flow channel adjustment capability, and the temperature difference could not be improved. Regarding leakage handling, Example 1 created 0.1mm micropores using a needle-punching method. The system detected a sudden pressure drop within 15ms and completed gel sealing within 200ms, with no leakage after 48 hours of continuous operation. Comparison 1 lacked any leakage detection and self-repair mechanism; a single loose connector caused continuous leakage for 20 minutes, ultimately resulting in a short circuit and damage to the motherboard. Regarding environmental adaptability, the pumpless immersion edge computing device in Example 3 operates stably under high temperature of 55℃ and vehicle vibration, and microcracks caused by drops can self-repair within 300ms; in contrast, it cannot be deployed in the field or in vehicle environments. In cluster collaboration, Example 5 reduces the maximum temperature difference of the 48-node cluster from 12℃ to 4℃ through inter-capsule communication, and reduces the total pump power consumption by 22%; in contrast, each node in Comparison 1 operates independently, and can only passively reduce its frequency when there is a local overload. In summary, the embodiments of this invention significantly outperform traditional liquid cooling solutions in terms of response speed, self-healing capability, environmental tolerance, and system energy efficiency, fully demonstrating its non-obvious technological advancements.

[0126] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of protection of the claims.

[0127] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A high-performance liquid-cooled server system, characterized in that, include: A sealed module, internally filled with coolant, wherein self-healing microcapsules are dispersed in the coolant; A heat exchange module is disposed within the sealing module for attaching the heating chip. The heat exchange module has an adjustable flow channel structure inside and a porous composite coating on its surface. A fluid drive module, disposed within the sealing module, is used to drive the coolant to flow through the heat exchange module; A sensing module, disposed within the sealed module, is used to detect temperature, pressure, dielectric constant, and flow rate; A control module is disposed within or adjacent to the sealing module, and the control module is electrically or communicatively connected to the adjustable flow channel structure, the fluid drive module, and the sensing module, respectively. The control module receives the detection data from the sensing module and predicts the thermal field distribution of the heating chip based on the detection data. When the predicted thermal field distribution indicates that the temperature will exceed a preset threshold within a preset future time period, the control module sends a speed adjustment command to the fluid drive module and a geometry adjustment command to the adjustable flow channel structure to suppress the temperature from exceeding the preset threshold. The control module also determines whether a leak has occurred based on the detection data from the sensing module, and when a leak is detected, controls the self-healing microcapsules to rupture and cross-link to form a gel to seal the leak.

2. The high-performance liquid-cooled server system as described in claim 1, characterized in that, The adjustable flow channel structure includes multiple flexible fins driven by actuators, each of which can deflect independently; the heat exchange module also has a flow channel of non-uniform width, with a flow channel density greater than that in the preset high-heat zone than in the low-heat zone, and the high-heat zone is provided with jet holes or vortex cavities.

3. The high-performance liquid-cooled server system as described in claim 1, characterized in that, The porous composite coating comprises a micron-sized copper pillar layer, a nanoporous alumina layer, and a fluorinated organic molecule layer stacked sequentially, and the contact angle between the coating and the coolant is less than 10 degrees.

4. The high-performance liquid-cooled server system as described in claim 1, characterized in that, The base liquid of the coolant is a fluorinated liquid, and the coolant also contains heat transfer enhancing particles, which are graphene nanosheets with a hydrophilic polymer attached to their surface; the capsule wall of the self-healing microcapsule is polyurea or polyurethane, and the capsule contains isocyanate prepolymer and polyamine. When the contents of the microcapsule rupture and come into contact with air or moisture, cross-linking polymerization occurs.

5. The high-performance liquid-cooled server system as described in claim 1, characterized in that, The sensing module includes: a temperature sensing unit, a pressure sensing unit, a dielectric constant sensing unit, a flow velocity sensing unit, and an acceleration sensing unit.

6. The high-performance liquid-cooled server system as described in claim 1, characterized in that, The control module is further configured to: when the leakage probability calculated based on the pressure pulsation spectrum characteristics and dielectric constant fluctuation characteristics exceeds a first preset value, send a deceleration command to the fluid drive module and a start command to the mechanical clamping unit inside the sealing module housing; when the leakage probability exceeds a second preset value and the detected pressure drop rate exceeds a threshold, send a stop command to the fluid drive module and control the heating unit located inside the sealing module to generate pulse heat, causing the self-healing microcapsule to rupture; wherein the leakage probability is obtained by comparing the pressure pulsation spectrum characteristics and the dielectric constant fluctuation characteristics with preset leakage determination rules.

7. The high-performance liquid-cooled server system as described in claim 1, characterized in that, The sealing module also includes a compensation chamber, which contains a flexible pressure regulating bladder filled with inert gas to maintain positive pressure within the sealing module.

8. The high-performance liquid-cooled server system as described in claim 1, characterized in that, The fluid drive module includes at least two drive units connected in parallel. The control module controls the drive units to operate in turn at a power value less than or equal to the rated power, and switches to the other drive unit when an abnormal current or flow rate is detected in any of the drive units.

9. A high-performance liquid-cooled server device, applied to the high-performance liquid-cooled server system according to any one of claims 1-8, characterized in that, The sealing module, the heat exchange module, the fluid drive module, the sensing module, and the control module are all integrated into an independent capsule-shaped package. The capsule-shaped package only has a power supply interface and a data interface externally, with no exposed liquid pipelines.

10. A high-performance liquid-cooled server device as described in claim 9, characterized in that, The capsule-type package is composed of a metal inner liner and a fiber-reinforced composite material outer shell. The overall protection level of the capsule-type package is IP68, the operating temperature range is -40℃ to 105℃, and it can withstand 20g of vibration. The capsule-type package is also equipped with a self-sealing supply valve.

Images