A method and system for detecting leakage in server liquid cooling systems based on multi-sensor fusion

Abstract

This invention discloses a method and system for detecting leaks in a server liquid cooling system based on multi-sensor fusion. The method includes: deploying leak detection sensing terminals at potential leak points and the bottom tray of the liquid cooling system to collect environmental signals in real time; preprocessing the collected environmental signals and extracting their time-domain features; comprehensively analyzing the extracted time-domain features based on hierarchical judgment logic to determine a leak event; and triggering an alarm and taking appropriate action for confirmed leak events. This invention, through a hierarchical judgment logic of instantaneous triggering and continuous confirmation, combined with multi-sensor data fusion, effectively avoids false alarms caused by dust, condensation, or transient interference from a single sensor, reducing the false alarm rate and achieving high reliability.

Description

Technical Field

[0001] This invention relates to the field of server liquid cooling system leakage detection technology, and in particular to a method and system for server liquid cooling system leakage detection based on multi-sensor fusion. Background Technology

[0002] With the explosive growth in computing power demand in data centers, the power consumption and heat dissipation pressure of servers continue to rise. Traditional air cooling methods can no longer meet the heat dissipation needs of high-density computing clusters. Liquid cooling technology, with its advantages of high heat dissipation efficiency, low energy consumption and low noise, has gradually become the mainstream solution for heat dissipation of data center servers and is widely used in various high-computing-power server deployment scenarios.

[0003] Liquid cooling systems rapidly dissipate heat generated by core server components through the circulation of coolant in a closed loop, significantly improving heat dissipation efficiency. However, this technology also introduces a new risk – leakage. Server motherboards, CPUs, graphics cards, and other core electronic components are densely packed and highly sensitive to liquids. If a coolant leak occurs in the liquid cooling system, the liquid coming into contact with these electronic components can directly cause short circuits, burnouts, and other malfunctions. This can not only lead to the downtime of one or more servers, resulting in data loss and business interruption, but also trigger a chain reaction, damaging other equipment within the rack and causing incalculable economic losses to the data center. Therefore, leak detection and early warning systems for liquid cooling systems are crucial for the successful application of liquid cooling technology.

[0004] To mitigate the system risks posed by liquid leakage, the industry has developed and applied various liquid leakage detection technologies for server liquid cooling systems. The mainstream technologies mainly include three types: point-type leakage sensor detection, line-type (cable-type) leakage sensor detection, and flow / pressure sensor monitoring. However, all of these technologies have significant technical shortcomings in practical applications, making it difficult to simultaneously meet the requirements of timeliness, accuracy, reliability, and cost control in detection.

[0005] 1. Point-type leak detection: This technology uses detection probes based on resistance or capacitance principles placed at key locations such as the bottom of the cabinet and below pipe joints. When leaking liquid comes into contact with the probe, an alarm is triggered. Its core drawback is the limited detection range, which can only monitor a very small area around the probe. The leaking liquid needs to spread to the probe location to be detected, resulting in a large response delay. At the same time, environmental factors such as dust, condensation, and moisture in the computer room can easily cause false triggering of the probe, leading to a high rate of missed and false alarms.

[0006] 2. Linear (Cable) Leakage Sensor Detection: Sensor cables are laid along liquid cooling pipes, cabinet edges, etc., and an alarm can be triggered at any point where liquid comes into contact with the cable. Compared to point detection, this expands the monitoring range, but several problems remain: First, sensor cables are expensive and require strict installation techniques, needing to be tightly fitted to the area being detected; improper installation can easily lead to detection failure. Second, in complex systems, the sensor cable routing is cumbersome, occupying a significant amount of system space and affecting data center cabinet layout and equipment expansion. Third, for small liquid cooling products, it is impossible to accurately locate the leak, requiring manual secondary inspection by maintenance personnel. Fourth, changes in environmental humidity and cable surface contamination can still easily trigger false alarms.

[0007] 3. Flow / Pressure Sensor Monitoring: By detecting the flow rate and pressure difference of the coolant at the inlet and outlet of the liquid cooling circuit, the presence of leaks in the circuit can be determined. This method has a relatively fast response speed, but it requires extremely high sensor accuracy, resulting in high hardware procurement and maintenance costs. At the same time, this technology can only determine whether there is a leak in the liquid cooling circuit, but cannot locate the specific leak point, which brings great inconvenience to subsequent maintenance. In addition, normal pressure and flow fluctuations during the operation of the liquid cooling system can also be easily misinterpreted as leak signals, leading to false alarms in the system.

[0008] In addition to the above-mentioned problems, some existing leakage detection solutions also suffer from low modularity, high deployment difficulty, and poor adaptability, making it difficult to meet the flexible deployment needs of data centers of different sizes (single node, rack, cluster) and also unable to adapt to the transformation scenarios of old data centers that have been converted from air cooling to liquid cooling.

[0009] For example, invention application No. 202511364285.1 discloses a method, device, and system for detecting leakage in a server's liquid cooling system. This method generates an early warning message based on the air quality index of a first air sensor being greater than or equal to a preset normal threshold. The early warning message indicates that there is a coolant leak in the server's liquid cooling system. This method aims to achieve timely detection of liquid cooling system leaks and improve the real-time performance of leak detection. However, this solution also has the following drawbacks: the detection mechanism is simplistic and easily affected by environmental interference; it cannot directly confirm the presence of liquid, and the response may be delayed.

[0010] Therefore, existing liquid cooling system leakage detection technologies for servers suffer from technical pain points such as high detection latency, high false alarm and false alarm rates, low positioning accuracy, high cost, and complex installation and deployment. The industry urgently needs a new leakage detection method and system that can achieve early warning, accurate judgment, reliable alarm, low cost, and easy deployment and expansion to address the shortcomings of existing technologies and provide core support for the large-scale and reliable application of liquid cooling technology. Summary of the Invention

[0011] To address the aforementioned problems, the present invention aims to provide a method and system for detecting leakage in server liquid cooling systems based on multi-sensor fusion, thereby solving the problems of high detection latency, high false alarm and false negative rates, low positioning accuracy, and high cost in existing server liquid cooling system leakage detection technologies, and achieving early warning, accurate judgment, and reliable alarm for server liquid cooling system leakage phenomena.

[0012] The objective of this invention can be achieved through the following technical solutions:

[0013] A method for detecting leakage in a server liquid cooling system based on multi-sensor fusion, comprising:

[0014] S1: Deploy leak detection and sensing terminals at potential leak points and bottom trays of the liquid cooling system to collect environmental signals in real time;

[0015] S2: After preprocessing the collected environmental signals, extract the time-domain features of the signals;

[0016] S3: Based on hierarchical judgment logic, the extracted signal time-domain features are comprehensively analyzed to determine leakage events;

[0017] S4: For confirmed leaks, trigger alarms and take appropriate action.

[0018] As a further embodiment of the present invention, the leakage detection sensing terminal includes a group of:

[0019] Humidity sensor: Used to detect changes in ambient air humidity;

[0020] Temperature sensor: used to monitor the temperature near the leak point;

[0021] Optical sensor: Used to detect the presence of droplets or changes in reflectivity.

[0022] As a further embodiment of the present invention, the environmental signal preprocessing in S2 includes signal filtering, amplification, and analog-to-digital conversion; the extraction of signal time-domain features includes rate of change, absolute value, and duration.

[0023] As a further embodiment of the present invention, the layering judgment logic includes:

[0024] The first level of judgment is triggered instantaneously, and the second level of judgment is formed through continuous confirmation.

[0025] As a further aspect of the present invention, the instantaneously triggered first-level judgment is formed based on the sudden increase in humidity value of any humidity sensor within a very short period of time.

[0026] As a further aspect of the present invention, the second-level judgment and confirmation condition formed by the continuous confirmation includes, within a certain period of time:

[0027] a. The reading of the humidity sensor used for the first-level judgment remains above the threshold;

[0028] b. A sudden change in humidity value occurs between adjacent humidity sensors or multiple humidity sensors used for the first-level judgment;

[0029] c. The optical sensor detects the presence of droplets or exhibits persistently abnormal reflectivity;

[0030] d. The corresponding temperature sensor detected an abnormally low temperature.

[0031] As a further embodiment of the present invention, S3 further includes a false alarm prevention logic judgment, including:

[0032] If only a single humidity sensor experiences a momentary jump, and its reading quickly returns to normal levels during the continuous monitoring period, it is considered interference.

[0033] If the humidity of the entire computer room increases, causing multiple humidity sensor readings to change abruptly, this is considered interference.

[0034] As a further aspect of the present invention, the alarm and handling actions performed in step S4, which determine a genuine leakage event, include:

[0035] Send a leak alarm message to the monitoring center, and simultaneously shut down the liquid supply solenoid valve of the liquid cooling system and activate the audible and visual alarm in the computer room.

[0036] A server liquid cooling system leakage detection system based on multi-sensor fusion, the system comprising:

[0037] Distributed sensor network: consists of multiple sets of low-cost sensors deployed at key nodes and on the bottom tray;

[0038] Area data collector: responsible for collecting data from all sensors within a designated area and performing preliminary preprocessing;

[0039] Central processing unit: Receives data from all area collectors and performs multi-layered logical judgments and false alarm prevention logic judgments;

[0040] Actuator unit: Receives instructions from the central processing unit and executes the actions of the liquid supply solenoid valve and alarm;

[0041] Power supply and communication bus: Provides power and data transmission channels for the entire system.

[0042] As a further embodiment of the present invention, the distributed sensor network is based on a modular leakage detection sensing terminal, which has a unified interface for power supply and communication, enabling plug-and-play rapid deployment.

[0043] The beneficial effects of this invention are:

[0044] 1. This invention effectively avoids false alarms caused by dust, condensation, instantaneous interference, etc., caused by a single sensor through a multi-level judgment mechanism of instantaneous triggering and continuous confirmation, combined with multi-sensor data fusion, thereby reducing the false alarm rate and achieving high reliability.

[0045] 2. The device of this invention can use mature commercial off-the-shelf sensors (such as DHT11 / DHT22 temperature and humidity sensors, phototransistors), and the central processing unit can use a low-cost MCU (such as STM32 series). The overall cost is far lower than that of high-precision flow / pressure difference detection solutions or long-distance linear sensors.

[0046] 3. The temperature and humidity sensor of this invention has an extremely fast response speed to liquids (millisecond level), which can detect leaks at a very early stage after they occur. At the same time, it can work in conjunction with an optical sensor to accurately identify liquid films or droplets that have already formed. Through collaborative work, it achieves full-stage monitoring coverage from trace leakage to obvious leakage, further improving the comprehensiveness and timeliness of detection.

[0047] 4. The distributed sensor network of this invention can accurately locate the cabinet or even specific component where a leak has occurred, greatly facilitating maintenance personnel in carrying out repairs.

[0048] 5. This invention adopts a modular design and is connected via a bus. New detection points can be added simply by connecting them in parallel to the bus, making it very suitable for the renovation of existing data centers and the construction of new data centers, and easy to deploy and expand. Attached Figure Description

[0049] Figure 1 This is a schematic flowchart of the method of the present invention;

[0050] Figure 2 This is a schematic diagram of the overall architecture of the system of the present invention;

[0051] Figure 3 This is a schematic diagram showing the deployment location of the sensor in this invention;

[0052] Figure 4 This is a schematic diagram of the leakage detection sensing terminal structure of the present invention;

[0053] Figure 5 This is a diagram of the software implementation logic code for the central processing unit of this invention. Detailed Implementation

[0054] Embodiments of the present invention are described in detail below. Examples of these embodiments are illustrated in the accompanying drawings, wherein the same or similar symbols denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0055] Example 1:

[0056] like Figure 1 As shown, this invention discloses a method for detecting leakage in a server liquid cooling system based on multi-sensor fusion, comprising the following steps:

[0057] First, leak detection terminals are deployed at potential leak points and the bottom tray of the liquid cooling system to collect environmental signals in real time.

[0058] Liquid cooling systems include those used in single compute nodes, cabinets, or clusters of multiple cabinets. Leak detection sensors are deployed at potential leak points in the liquid cooling system (e.g., under quick connectors, water pumps, and cold plates) and on the bottom trays of individual nodes / cabinets. Because these leak detection sensors are small and lightweight (requiring only multiple sensors), they can be adhesively fixed to the desired locations and communicate with the motherboard via connector cables, making installation convenient and simple.

[0059] Specifically, the deployment of the leakage detection sensing terminals is shown in Table 1:

[0060] Table 1 Deployment Table of Liquid Cooling System Leakage Detection Sensing Terminals

[0061]

[0062] The leakage detection sensing terminal includes a low-cost humidity sensor, a humidity sensor and / or an optical sensor. The humidity sensor and humidity sensor can be integrated into a temperature and humidity sensor, further reducing equipment costs and saving installation space.

[0063] Among them, the humidity sensor is used to detect changes in ambient air humidity. When a small leak occurs in the liquid cooling system, the evaporation of the leaked liquid will cause the ambient air humidity to rise. The humidity sensor can keenly capture this humidity change and convert it into an electrical signal output, providing a preliminary basis for the ambient humidity for leak detection.

[0064] Temperature sensors are used to monitor the temperature near the leak point and can be used in conjunction with the inlet and outlet temperatures of the cold plate for analysis. When a leak occurs in the liquid cooling system, the leaking cryogenic coolant absorbs heat from the surrounding environment, causing an abnormal drop in temperature near the leak point. The temperature sensor can capture this temperature change in real time and convert it into an electrical signal, feeding it back to the system. By analyzing this temperature data in conjunction with the inlet and outlet temperatures of the cold plate, if the leak point temperature is significantly lower than the cold plate outlet temperature and shows an abnormal deviation from the inlet temperature, it can further assist in determining whether a leak has occurred and its severity. This provides crucial temperature-related information for leak detection, improving the accuracy and reliability of the detection.

[0065] Optical sensors: used to detect the presence of droplets or changes in reflectivity. Optical sensors are optional and can be flexibly selected and deployed according to the leakage detection requirements of the actual application scenario.

[0066] Optical sensors illuminate the monitored area by emitting light of a specific wavelength (such as infrared or visible light). When leaking droplets are present in that area, the droplets scatter, refract, or absorb the light, causing changes in the intensity of the light signal received by the sensor. If the leaking liquid forms a film on the surface, it will also alter the reflectivity of the monitored surface. Optical sensors can accurately capture these changes in light signal and convert them into electrical signals, which are then transmitted to the system. By analyzing the changes in light signal intensity or abnormal fluctuations in reflectivity, it is possible to directly determine whether droplets have formed or liquid residue remains, providing intuitive optical evidence for leak detection. This, combined with humidity and temperature sensors, forms a multi-dimensional detection synergy, further enhancing the comprehensiveness and sensitivity of leak detection.

[0067] Real-time acquisition of environmental signals in the configured area based on low-cost humidity sensors, humidity sensors and / or optical sensors.

[0068] Then, the collected environmental signals are preprocessed, and their time-domain features are extracted.

[0069] The acquired raw environmental signals are preprocessed by filtering, amplification, and analog-to-digital conversion to remove high-frequency noise, interference signals, baseline drift, and other interference factors, ensuring signal stability and accuracy.

[0070] The filtering stage employs low-pass filters or adaptive filtering algorithms to effectively remove high-frequency noise generated by electromagnetic interference and equipment vibration in the environment. The amplification stage, based on the weak characteristics of the sensor output signal, uses a high-precision operational amplifier to adjust the signal amplitude to a suitable range for subsequent analog-to-digital conversion, avoiding signal distortion. The analog-to-digital conversion stage utilizes a high-performance A / D converter to convert the processed analog signal into a digital signal, providing reliable digital input for subsequent feature extraction and analysis. The preprocessed signal more realistically reflects the actual environmental conditions of the server liquid cooling system, laying a solid data foundation for subsequent time-domain feature extraction.

[0071] Extracting time-domain features of a signal, such as rate of change, absolute value, and duration, allows us to characterize the dynamic changes of the signal from different dimensions. The rate of change reflects the degree of change of the signal per unit time, obtained by calculating the ratio of the difference between signal values ​​at adjacent sampling points to the time interval. This effectively captures abrupt changes in the signal, such as the rapid rise or fall of humidity or optical signals during a leak. The absolute value is a direct measure of signal strength, intuitively reflecting the overall signal level. When a leak causes the sensor's detection value to exceed the normal range, its absolute value will significantly deviate from the baseline value. The duration feature focuses on the duration of the abnormal signal state. By setting a reasonable threshold, we can determine whether the signal abnormality is a continuous event, avoiding misjudgments caused by transient interference or fluctuations.

[0072] Then, based on the hierarchical judgment logic, the extracted signal time-domain features are comprehensively analyzed to determine the leakage event.

[0073] Establish a hierarchical judgment logic to comprehensively analyze the extracted signal time-domain features and determine whether it is a leakage event.

[0074] The logic for determining the layer includes:

[0075] Level 1 Judgment (Instantaneous Trigger): If the reading of any humidity sensor increases abruptly within a very short time (e.g., within 1 second) (e.g., the humidity value suddenly rises from 50%RH to above 95%RH), a Level 1 Leakage Suspicion Signal is generated.

[0076] Level 2 Judgment (Continuous Confirmation): After generating a Level 1 suspected leak signal, a continuous monitoring timer is started (e.g., 10 seconds). Within this time window, if one of the following conditions is met, a Level 2 leak confirmation signal is generated:

[0077] a. The humidity sensor reading used for the first-level judgment consistently remains above the threshold (e.g., 90% RH).

[0078] b. A sudden change in humidity value occurs between adjacent humidity sensors or multiple humidity sensors used for the first-level judgment, indicating that the liquid is diffusing.

[0079] c. The optical sensor detects the presence of droplets or the reflectivity remains abnormal.

[0080] d. The corresponding temperature sensor detected an abnormally low temperature.

[0081] If any one of the four judgment conditions mentioned above is met at the second level, it can be judged as a leakage event and the alarm mechanism can be triggered.

[0082] This layered, multi-condition judgment design effectively avoids false alarms or missed alarms from a single sensor, significantly improving the accuracy and reliability of leak detection. When a humidity sensor experiences a momentary high humidity due to localized fluctuations in ambient moisture, if adjacent sensors do not show a sudden change in humidity and the optical and temperature sensors are functioning normally, it will not be misjudged as a leak. Conversely, if the humidity sensor maintains high humidity, or if adjacent sensors simultaneously show humidity changes, or if the optical sensor detects droplets, or if the temperature sensor detects localized low temperatures caused by a leak, a leak can be quickly confirmed, thus providing timely assurance for the safe and stable operation of the server liquid cooling system.

[0083] Finally, for confirmed leaks, an alarm should be triggered and appropriate action taken.

[0084] Through layered judgment logic, a genuine leak event is identified, triggering an alarm and related actions, including:

[0085] Send the highest level of leak alarm information to the monitoring center, including the leak location (located to the sensor ID), the timestamp of the leak, real-time data and trends of key sensors such as humidity, optics, and temperature, so that personnel can quickly grasp the details of the leak.

[0086] At the same time, the system will automatically trigger preset emergency response procedures, such as immediately shutting off the liquid supply solenoid valve of the liquid cooling system in the corresponding area to prevent the leakage range from expanding; at the same time, it can start the liquid pumping device in the area to quickly collect the leaked coolant; and link the air conditioning system to reduce the ambient humidity to avoid secondary damage to the server equipment caused by water vapor.

[0087] In addition, the computer room's audible and visual alarms will be activated, using high-decibel sounds and high-brightness flashing lights to create a strong warning signal in the computer room, reminding on-site maintenance personnel to go to the leak area for emergency handling as soon as possible.

[0088] These multi-dimensional and multi-layered alarm and response mechanisms work together to form the last solid line of defense against leakage in server liquid cooling systems, minimizing the risk of equipment damage and business interruption that may result from leaks.

[0089] Example 2:

[0090] Based on Example 1, this embodiment further discloses the logic judgment for preventing false alarms and introduces an exclusion mechanism to prevent false alarms. The system will perform multi-dimensional cross-verification on the data collected by each sensor to further improve the accuracy and reliability of leakage detection and ensure the authenticity and effectiveness of alarm information.

[0091] Furthermore, the false alarm prevention logic includes:

[0092] If only a single humidity sensor experiences a momentary jump, and its reading quickly returns to normal levels during the continuous monitoring period, it is considered interference and will not trigger a leakage alarm. For example, temporary fluctuations in humidity caused by personnel briefly entering the computer room and carrying moisture, or by local airflow disturbances at the air conditioning vents, usually do not pose an actual risk of leakage.

[0093] In addition, the system will also perform trend analysis on the historical data of the humidity sensor. When the reading of a humidity sensor fluctuates frequently and irregularly over a period of time, but does not reach the alarm threshold, the sensor will be marked as suspicious and the maintenance personnel will be prompted to calibrate or replace the equipment to avoid false alarms or missed alarms caused by sensor failure.

[0094] If the humidity of the entire computer room increases, causing multiple humidity sensors to change abruptly, it may be a fault in the air conditioning system. In this case, it is determined to be interference, and the system will suppress local leakage alarms or suppress the issuance of system-level warnings.

[0095] By employing the above-mentioned false alarm prevention logic, false alarms caused by environmental interference, momentary sensor anomalies, or systemic environmental changes can be effectively filtered out, significantly improving the accuracy and reliability of leak detection. This allows the system to more accurately identify genuine leak risks, avoiding unnecessary maintenance responses and reducing the waste of human and material resources caused by false alarms, thus ensuring the stable operation of the server liquid cooling system and the security of the data center.

[0096] Example 3:

[0097] This embodiment discloses a server liquid cooling system leakage detection system, applied to the method described in Embodiment 1 or 2, such as... Figure 2 As shown, the system includes:

[0098] Distributed sensor network: It consists of multiple leakage detection and sensing terminals deployed at key nodes and the bottom tray.

[0099] The leakage detection and sensing terminal consists of multiple low-cost sensors, including humidity sensors, temperature sensors, temperature and humidity sensors, and / or optical sensors, which can collect multi-dimensional environmental parameters such as humidity, temperature, and liquid state in key parts of the liquid cooling system in real time.

[0100] Among them, the humidity sensor can keenly detect the abnormal increase in local humidity caused by liquid leakage, the temperature sensor can help determine whether there is a temperature change caused by liquid leakage, and the optical sensor can directly identify the presence of liquid by detecting the reflection or refraction characteristics of liquid to light. The three types of sensors work together to form a multi-dimensional cross-verification of the leakage state, which effectively improves the limitations of single sensor detection and provides rich and reliable raw data support for subsequent false alarm prevention logic judgment.

[0101] Furthermore, such as Figure 3 and Figure 4 As shown, a distributed sensor network can integrate multiple sensors (such as temperature and humidity sensors, optical sensors, and non-contact temperature sensors) onto a single detection board. By creating a highly integrated, multi-functional leak detection sensing terminal group (i.e., the "detection board"), it serves as the basic sensing unit deployed in various key locations within the server rack. Each unit possesses comprehensive environmental sensing capabilities and achieves rapid, plug-and-play deployment through a unified interface for power supply and communication, forming a unified leak detection sensing terminal for server liquid cooling systems. This fulfills the requirements of feasibility and cost control.

[0102] Area data collector: responsible for collecting data from all sensors within a designated area and performing preliminary preprocessing.

[0103] The area data collector is responsible for collecting data from sensors in a small area (such as a cabinet). After preprocessing, the area data collector uploads the processed area data to the central processing unit via the communication bus, realizing the key transition from decentralized sensing to centralized processing and providing an efficient and reliable data transmission channel for the overall leakage monitoring and decision-making of the system.

[0104] The preprocessing process includes signal filtering, amplification, and analog-to-digital conversion. The filtering stage employs low-pass filters or adaptive filtering algorithms to remove high-frequency noise generated by environmental electromagnetic interference and equipment vibration. The amplification stage uses a high-precision operational amplifier to adjust the signal amplitude to a suitable range for subsequent analog-to-digital conversion, based on the weak characteristics of the sensor output signal, thus avoiding distortion. The analog-to-digital conversion stage uses a high-performance A / D converter to convert the processed analog signal into a digital signal, providing reliable digital input for subsequent feature extraction and analysis. The preprocessed signal more accurately reflects the actual environmental conditions of the server's liquid cooling system, laying a data foundation for subsequent time-domain feature extraction.

[0105] Central Processing Unit: Receives data from all area collectors, performs multi-layered logical judgments and false alarm prevention logic judgments. The central processing unit is the core decision-making module of the system.

[0106] First, it aggregates and synchronizes the digital signals uploaded from the data collectors in each region, ensuring consistency of data across multiple regions in terms of time and providing a unified time benchmark for subsequent fusion analysis. Next, it executes multi-layered logical judgments.

[0107] Level 1 Judgment (Instantaneous Trigger): If the reading of any humidity sensor increases abruptly within a very short time (e.g., within 1 second) (e.g., the humidity value suddenly rises from 50%RH to above 95%RH), a Level 1 Leakage Suspicion Signal is generated.

[0108] Level 2 Judgment (Continuous Confirmation): After generating a Level 1 suspected leak signal, a continuous monitoring timer is started (e.g., 10 seconds). Within this time window, if one of the following conditions is met, a Level 2 leak confirmation signal is generated:

[0109] a. The humidity sensor reading used for the first-level judgment consistently remains above the threshold (e.g., 90% RH).

[0110] b. A sudden change in humidity value occurs between adjacent humidity sensors or multiple humidity sensors used for the first-level judgment, indicating that the liquid is diffusing.

[0111] c. The optical sensor detects the presence of droplets or the reflectivity remains abnormal.

[0112] d. The corresponding temperature sensor detected an abnormally low temperature.

[0113] After the above-mentioned multi-layered logical judgment, the central processing unit generates corresponding control commands based on the judgment results, sends them to the actuator unit, and sends relevant data to the monitoring center.

[0114] Actuator unit: Receives instructions from the central processing unit and executes the actions of the liquid supply solenoid valve and alarm.

[0115] Upon receiving a leak confirmation signal, the actuator unit immediately drives the liquid supply solenoid valve to cut off the liquid supply circuit of the liquid cooling system to prevent the leak from expanding further. At the same time, the alarm is activated. The alarm can be set up using an audible and visual alarm, which will promptly alert maintenance personnel to the liquid leak fault by emitting a high-decibel alarm sound and flashing warning lights, so that they can respond quickly and take appropriate measures.

[0116] Power supply and communication bus: Provides power and data transmission channels for the entire system. Low-cost RS-485 or CAN bus can be used, which greatly simplifies wiring and reduces costs.

[0117] It also supports long-distance data transmission, ensuring the stability and reliability of communication between sensor nodes and the central processing unit. The power supply section adopts a wide voltage input design, adaptable to power requirements in different scenarios, ensuring continuous and stable operation of the system in complex power environments. The communication bus has excellent anti-interference capabilities, effectively reducing the impact of electromagnetic interference in industrial environments on data transmission, ensuring that leakage detection data is transmitted to the central processing unit in real time and accurately, providing a solid guarantee for the system's accurate judgment and rapid response.

[0118] Example 4:

[0119] This embodiment combines the methods and systems of the above embodiments to implement the method of the present invention in a standard 42U server rack.

[0120] 42U server rack cold plate liquid cooling system.

[0121] Hardware Deployment: Five temperature and humidity sensors (using DHT22, cost-effective) are installed at the four corners and center of the load-bearing tray at the bottom of the rack. A miniature optical droplet sensor (using a self-made low-cost infrared pair tube) is installed below the quick connector of each server cold plate. A normally open solenoid valve is installed at the main inlet of the rack's liquid cooling circuit.

[0122] A zone data acquisition unit (based on an STM32F103C8T6 minimum system board) is installed in the middle of the rack, collecting signals from all sensors within the rack via wires. All zone data acquisition units are connected to the central processing unit (an industrial Raspberry Pi or similar embedded computer) in the server room monitoring room via an RS-485 bus.

[0123] Software logic implementation:

[0124] The program within the central processing unit continuously polls the data from each sensor. Its core decision-making pseudocode is as follows: Figure 5 As shown.

[0125] Workflow:

[0126] When a micro-leak occurs at a cold plate joint, droplets fall. These droplets evaporate or diffuse on the bottom tray, causing the humidity reading of the nearest temperature and humidity sensor to spike instantly, triggering a level one judgment.

[0127] Over the next 10 seconds, the system continuously observed that the humidity at that point remained above 95% RH, and the humidity at an adjacent sensor also began to rise. Based on this, the system determined that there was a real leak, immediately shut off the main liquid cooling valve of the cabinet, and reported an alarm.

[0128] Furthermore, the system can detect droplets using optical sensors, which can also be used to determine if there is a real leak, immediately shut off the main liquid cooling valve of the cabinet, and report an alarm.

[0129] This invention discloses a commercially promising solution for detecting liquid cooling leaks in servers, balancing reliability, cost, and feasibility, and is particularly suitable for high-density server rack-mounted liquid cooling systems. This solution utilizes multiple sensors working in tandem to accurately capture initial humidity changes at the micro-leakage points of the cold plate joints. It also incorporates optical detection methods for dual verification, effectively avoiding system downtime risks caused by false alarms from a single sensor. Compared to traditional single-sensor detection solutions, this system significantly reduces overall deployment costs while ensuring detection accuracy. It requires no large-scale modifications to existing rack structures; simply adding temperature and humidity sensors and optical sensors at key locations is sufficient for deployment. This provides economical and reliable technical support for the safe and stable operation of data center liquid cooling systems.

[0130] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for detecting liquid leakage of a server liquid cooling system based on multi-sensor fusion, characterized in that, include: S1: Deploy leak detection and sensing terminals at potential leak points and bottom trays of the liquid cooling system to collect environmental signals in real time; S2: After preprocessing the collected environmental signals, extract the time-domain features of the signals; S3: Based on hierarchical judgment logic, the extracted signal time-domain features are comprehensively analyzed to determine leakage events; S4: For confirmed leaks, trigger alarms and take appropriate action.

2. The method according to claim 1, characterized in that, The leakage detection sensing terminal includes a group of configured components: Humidity sensor: Used to detect changes in ambient air humidity; Temperature sensor: used to monitor temperature changes near the leak point; Optical sensor: used to detect the presence of droplets or changes in reflectivity.

3. The method according to claim 1, characterized in that, The environmental signal preprocessing in S2 includes signal filtering, amplification, and analog-to-digital conversion; the extraction of signal time-domain features includes rate of change, absolute value, and duration.

4. The method according to claim 2, characterized in that, The layering judgment logic includes: The first level of judgment is triggered instantaneously, and the second level of judgment is formed through continuous confirmation.

5. The method according to claim 4, characterized in that, The instantaneously triggered first-level judgment is based on the fact that the humidity value of any humidity sensor increases abruptly within a very short period of time.

6. The method according to claim 5, characterized in that, The second-level judgment and confirmation conditions formed by the continuous confirmation include, within a certain period of time: a. The reading of the humidity sensor used for the first-level judgment remains above the threshold; b. A sudden change in humidity value occurs between adjacent humidity sensors or multiple humidity sensors used for the first-level judgment; c. The optical sensor detects the presence of droplets or exhibits persistently abnormal reflectivity; d. The corresponding temperature sensor detected an abnormally low temperature.

7. The method according to claim 4, characterized in that, The S3 also includes false alarm prevention logic, including: If only a single humidity sensor experiences a momentary jump, and its reading quickly returns to normal levels during the continuous monitoring period, it is considered interference. If the humidity of the entire computer room increases, causing multiple humidity sensor readings to change abruptly, this is considered interference.

8. The method according to claim 1, characterized in that, The alarms and actions taken in S4 that determine a genuine leak event include: Send a leak alarm message to the monitoring center, and simultaneously shut down the liquid supply solenoid valve of the liquid cooling system and activate the audible and visual alarm in the computer room.

9. A server liquid cooling system leakage detection system, characterized in that, The system, applicable to the method of any one of claims 1 to 8, comprises: Distributed sensor network: consists of multiple sets of low-cost sensors deployed at key nodes and on the bottom tray; Area data collector: responsible for collecting data from all sensors within a designated area and performing preliminary preprocessing; Central processing unit: Receives data from all area collectors and performs multi-layered logical judgments and false alarm prevention logic judgments; Actuator unit: Receives instructions from the central processing unit and executes the actions of the liquid supply solenoid valve and alarm; Power supply and communication bus: Provides power and data transmission channels for the entire system.

10. The system according to claim 9, characterized in that, The distributed sensor network is based on modular leakage detection sensing terminals, which have a unified interface for power supply and communication, enabling plug-and-play rapid deployment.

Images