Liquid leakage monitoring method and system for cold plate liquid cooling server based on multiple physical characteristics

Abstract

The application discloses a cold plate type liquid cooling server leakage monitoring method and system based on multiple physical characteristics, and relates to the field of fluid leakage detection; by synchronously collecting real-time impedance characteristics of a leakage sensing rope and real-time temperature characteristics of a corresponding area, a double-physical-quantity coupling data set under the same time reference is established; based on the two-dimensional feature recognition, the correlation characteristics of impedance change with temperature are identified, the liquid species distinction between condensate water and coolant is completed; based on the baseline feature offset after the impedance drop recovery event of the sensing rope impedance, the historical tracing judgment of the dried trace leakage is completed; the two judgment results are fused, the alarm and early warning signals of the leakage monitoring are generated in stages, the impedance drop caused by condensate water and coolant is effectively distinguished under the working condition with temperature fluctuation, the dried leakage that has triggered the impedance drop event is historically traced, the false alarm and missed alarm probability of single-dimensional impedance detection is reduced, and the accuracy and long-term operation reliability of the cold plate type liquid cooling server leakage monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of fluid leakage detection technology, specifically to a method and system for monitoring leakage in cold-plate liquid-cooled servers based on multiple physical characteristics. Background Technology

[0002] With the large-scale application of high-power-density computing servers, the operating power of a single rack has increased significantly. Cold-plate liquid cooling, with its core advantages of high heat exchange efficiency, strong adaptability to heat density, and low operating noise, has become the mainstream cooling solution for high-power-density servers. Leakage monitoring, as the last line of defense for the safety of liquid cooling systems, directly determines the security of server hardware and the continuous operational reliability of computing clusters. It is a core technical issue that must be addressed for the large-scale deployment of cold-plate liquid cooling systems.

[0003] Currently, single-dimensional impedance amplitude detection based on leakage sensing ropes is a common, large-scale application solution in the field of fluid leakage detection for cold-plate liquid-cooled servers. The core principle of this solution is as follows: a dual-core parallel leakage sensing rope is used as the detection element. Two parallel conductive cores are wrapped with a microporous liquid-absorbing insulating layer. In a dry state, the two cores are in a high-resistance isolation state. When the sensing rope comes into contact with liquid, the insulating layer absorbs the liquid through the micropores, creating a conductive path between the two cores, significantly reducing the equivalent impedance. Existing technology collects the real-time impedance amplitude of the sensing rope and presets a fixed leakage threshold. When the impedance value remains below the threshold, a leakage alarm is triggered. This solution, due to its simple hardware structure, flexible deployment, and controllable cost, can be adapted to point-based monitoring needs in confined spaces such as cold-plate joints, pipe connections, and liquid collection tanks at the bottom of server chassis, making it a mainstream technology in the industry.

[0004] To reduce the probability of false alarms during field operation, existing technologies have undergone several optimizations and improvements around the single-dimensional impedance detection framework. These include adding transient electromagnetic interference filtering algorithms, setting multi-gradient hierarchical alarm thresholds, and employing multi-node sensor rope linkage judgment logic. While these optimizations effectively reduce occasional false alarms caused by external environmental interference within the existing detection principle framework, they still have some inherent limitations in identification that are difficult to completely resolve due to the inherent limitations of the single-dimensional impedance amplitude detection principle.

[0005] The aforementioned single-dimensional impedance detection scheme relies solely on a single electrical physical quantity for leakage detection. It cannot accurately distinguish between coolant leaks posing a hardware short-circuit damage risk and condensation events without hardware hazard through coupled correlation analysis of physical characteristics across different dimensions. It also cannot identify trace historical leaks that have dried after impedance recovery, leading to a certain probability of false alarms and missed alarms. Since the impedance amplitudes of condensation and coolant overlap within specific temperature ranges, relying solely on a single impedance amplitude is insufficient to accurately determine the true cause of impedance drop. This is because single-dimensional impedance detection cannot acquire the crucial physical characteristic of the liquid's conductivity temperature coefficient, which is the essential physical basis for distinguishing between condensation and coolant. Furthermore, when a trace leak occurs at a cold plate joint or pipe connection, the leaked coolant will naturally evaporate under the heat generated by server operation, causing the sensor rope impedance to return to the normal range. Current technology only monitors the real-time impedance state and cannot detect the irreversible shift in the static reference value of the sensor rope caused by the residual non-volatile components after coolant evaporation, thus failing to promptly identify early micro-leakage hazards. Furthermore, existing technologies do not fully consider the long-term impact of residue accumulation on the sensor rope surface on detection sensitivity. As residue accumulates, the detection sensitivity of the sensor rope will gradually change, which will have a certain impact on the long-term operational reliability of the system.

[0006] Therefore, there is an urgent need in this field for a leakage monitoring method that can distinguish between condensate and coolant from a physical perspective based on multi-physical feature coupling analysis, and can trace the history of dried-up trace leaks. This would further solve the fundamental problems of false alarms and missed alarms in existing single-dimensional impedance detection, and better meet the core technical requirements of cold plate liquid-cooled servers for high accuracy and long-term reliable operation of leakage monitoring. Summary of the Invention

[0007] This invention aims to solve the technical problems of false alarms, missed alarms, and inability to trace dried traces of leakage in existing single-dimensional impedance detection. By simultaneously collecting impedance and temperature dual physical characteristics for coupled analysis, identifying baseline offsets to achieve historical tracing, and fusing judgment results to dynamically compensate for detection sensitivity, this invention effectively improves the accuracy and long-term operational reliability of leakage monitoring for cold plate liquid-cooled servers.

[0008] To overcome the shortcomings of existing technologies and achieve the above objectives, this invention proposes a method for monitoring leakage in cold-plate liquid-cooled servers based on multiple physical characteristics, comprising the following steps: S1: Synchronously collect multi-dimensional physical features to obtain the real-time impedance characteristics of the leakage sensing rope and the real-time temperature characteristics of the monitoring area corresponding to the leakage sensing rope. S2: Based on the impedance and temperature characteristics, identify the correlation characteristics of impedance changes with temperature, perform liquid type differentiation judgment, and generate a liquid type differentiation result that distinguishes between condensate and coolant. S3: Based on the impedance characteristics, monitor and identify the baseline characteristic offset of the sensing rope after the impedance drop recovery event, perform leakage history tracing judgment, and generate historical tracing results indicating that the dried-up trace leakage event has been generated. S4: Integrate the liquid type determination results with the historical traceability results to generate the final alarm and early warning signals for leakage monitoring.

[0009] Preferably, in step S2, the specific steps for identifying the correlation characteristics of impedance changes with temperature and performing liquid type differentiation include: S21: When the absolute value of the temperature change within the preset time window is greater than or equal to the preset minimum temperature change threshold, calculate the correlation parameter characterizing the degree of correlation between the impedance change and the temperature change. The correlation parameter is the ratio of the relative impedance change to the temperature change. The relative impedance change is the ratio of the absolute value of the impedance change within the preset time window to the impedance value at the beginning of the time window. S22: Match the associated parameters with the preset first feature interval and second feature interval, wherein the first feature interval corresponds to the high temperature coefficient characteristic of conductivity, and the second feature interval corresponds to the low temperature coefficient characteristic of conductivity. S23: When the associated parameter falls into the first characteristic range, it is determined that the impedance drop is caused by condensate; when the associated parameter falls into the second characteristic range, it is determined that the impedance drop is caused by coolant.

[0010] Preferably, in step S22, the determination of the first feature interval and the second feature interval includes: Based on the physical differences between condensate and coolant in terms of conductivity temperature coefficient, the first characteristic interval corresponds to the range of values ​​of the associated parameter where the conductivity of condensate increases exponentially with increasing temperature, and the second characteristic interval corresponds to the range of values ​​of the associated parameter where the rate of increase of coolant conductivity with increasing temperature is significantly lower than the rate of increase of condensate conductivity.

[0011] Preferably, in step S23, after determining that the impedance drop is caused by the coolant, the following steps are further included: S24: Compare the real-time impedance value with the preset leakage impedance threshold, and perform joint verification by combining the impedance amplitude condition and the duration condition; when the liquid type is determined to be coolant and the impedance value is continuously lower than the leakage impedance threshold for more than the preset time threshold, mark the determination result as a valid coolant leakage determination result and include it in the valid output item of the liquid type determination result.

[0012] Preferably, in step S3, the specific steps for monitoring and identifying the baseline characteristic shift of the sensing rope after the impedance drop recovery event and performing leakage history tracing determination include: S31: When the sensing rope is in an initial clean state, measure and store the initial static reference value of the sensing rope. S32: Monitor the occurrence and end of impedance drop recovery events. The impedance drop recovery event refers to the complete process in which the impedance value drops from the normal range to below the preset leakage impedance threshold and continues to exceed the preset effective drop time threshold, and then rises back to the normal range. The normal range refers to the state range in which the impedance value is higher than the leakage impedance threshold. S33: After the impedance drop recovery event ends and the sensor rope surface is dry and stable, measure the current static reference value of the sensor rope. S34: Calculate the decrease of the current static reference value relative to the initial static reference value. When the decrease exceeds the preset offset threshold, determine that there are irreversible conductive residues on the surface of the sensing rope due to the evaporation of coolant, and generate a historical traceability warning for the dried coolant leakage.

[0013] Preferably, in step S34, the specific steps for determining whether there are irreversible conductive residues on the surface of the sensing rope due to coolant evaporation include: S341: Within a preset time period after the impedance drop recovery event ends, collect the static reference value of the sensing rope under at least two different ambient humidity conditions, and identify the type of deviation of the current static reference value relative to the initial static reference value: If the current static reference value is continuously lower than the initial static reference value, and its value decreases as the ambient humidity increases and increases as the ambient humidity decreases, it is an irreversible deviation caused by hygroscopic conductive residues formed after the evaporation of non-volatile components in the coolant; if the current static reference value gradually rises back to near the initial static reference value as the ambient humidity decreases, it is a reversible deviation caused by the evaporation of condensate. S342: When an irreversible offset is identified and the reduction exceeds a preset offset threshold, it is determined that there are irreversible conductive residues on the surface of the sensing rope, and a historical traceability warning for dried coolant leakage is generated.

[0014] Preferably, in step S34, the offset threshold is pre-calibrated based on the sensor rope type and coolant component concentration, and the ratio of the offset threshold to the initial static reference value is a preset proportional coefficient. The range of the preset proportional coefficient is determined based on the content of non-volatile components in the coolant.

[0015] Preferably, in step S4, the specific steps for determining the fusion liquid type and comparing the historical traceability results include: S41: Establish a fusion judgment logic hierarchy, wherein the liquid type judgment result is used to distinguish between condensate events and coolant events in real time, and the historical traceability result is used to identify the state of residues already existing on the surface of the sensing rope. S42: When the historical tracing result is positive, the leakage impedance threshold is adjusted downward to compensate for the reduction in the baseline impedance of the sensing rope caused by residue and to maintain the leakage detection sensitivity; the adjustment is provided with a preset lower limit to prevent over-adjustment.

[0016] Preferably, in step S42, the downward adjustment range of the leakage impedance threshold is dynamically determined based on the graded and quantified residue accumulation level identified by the historical traceability results. The higher the residue accumulation level, the greater the adjustment range. The adjusted leakage impedance threshold is not lower than the preset lower limit. When the leakage impedance threshold has been adjusted to the preset lower limit but still needs further compensation, a sensor rope contamination replacement warning is triggered.

[0017] In addition, to achieve the above objectives, the present invention also proposes a liquid leakage monitoring system for cold plate liquid-cooled servers based on multiple physical characteristics, including: an impedance acquisition module, a temperature acquisition module, a memory, a processor, and a liquid leakage monitoring program for cold plate liquid-cooled servers based on multiple physical characteristics stored in the memory and executable on the processor. The impedance acquisition module is used to acquire the real-time impedance characteristics of the leakage sensing rope. The temperature acquisition module is used to acquire the real-time temperature characteristics of the monitoring area corresponding to the leakage sensing rope. The processor is connected to the impedance acquisition module, the temperature acquisition module, and the memory respectively. The liquid leakage monitoring program for the cold plate liquid-cooled server based on multiple physical characteristics is configured to implement the steps of the liquid leakage monitoring method for the cold plate liquid-cooled server based on multiple physical characteristics.

[0018] The beneficial effects of this invention are: 1. By synchronously collecting the real-time impedance characteristics of the leakage sensing rope and the real-time temperature characteristics of the corresponding monitoring area, a dual physical quantity coupled dataset under the same time reference is established; based on the inherent physical differences in the temperature coefficients of different liquid conductivity, by identifying the correlation characteristics of the relative rate of change of impedance with temperature, condensate and coolant can be effectively distinguished, reducing the probability of misjudgment caused by the overlap of impedance amplitudes of different liquids under single-dimensional impedance detection, improving the accuracy of qualitative analysis of leakage events, and ensuring the reliability of alarm signals.

[0019] 2. By monitoring the full-cycle changes in impedance characteristics, the baseline characteristic shift of the sensing rope after an impedance drop recovery event is accurately captured. Based on the irreversible shift of the static reference value of the sensing rope caused by the residue of non-volatile components after coolant evaporation, the identification and tracing of dry-out leakage events that have triggered impedance drops are completed. This expands the monitoring dimensions beyond the traditional method of only monitoring real-time leakage status, achieving full-process coverage of leakage risk from occurrence to dry-out. It can assist in identifying potential micro-leakage hazards at cold plate joints and pipe connections, providing data support for preventive maintenance of equipment.

[0020] 3. By integrating liquid type differentiation results with historical tracing results, a closed-loop monitoring logic driven by multiple physical features is constructed, outputting final alarm and early warning signals in a tiered manner. The qualitative judgment of real-time leakage events is deeply linked with the risk identification of the historical status of the sensor rope, forming a full-scenario monitoring system covering real-time leakage handling and historical hazard early warning. Balancing high-priority response to real-time leakage with early warning of potential hazards, this effectively reduces the risk of false alarms and missed alarms during long-term system operation, improving the long-term operational stability and reliability of the cold plate liquid-cooled server leakage monitoring system. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the steps of the liquid leakage monitoring method for cold plate liquid-cooled servers based on multiple physical characteristics as described in an embodiment of the present invention. Figure 2 This is a schematic diagram of the liquid type differentiation and determination process based on impedance temperature correlation parameters as described in an embodiment of the present invention; Figure 3 This is a schematic diagram of the process for determining leakage history based on impedance baseline offset according to an embodiment of the present invention; Figure 4 This is a structural diagram of the liquid leakage monitoring system for cold plate type liquid-cooled servers based on multiple physical characteristics, as described in an embodiment of the present invention.

[0022] Figure Labels 10. Impedance acquisition module; 20. Temperature acquisition module; 30. Memory; 40. Processor. Detailed Implementation

[0023] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described content is only a part of the embodiments of the present invention, and not all of the 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.

[0024] The features and effects of the present invention will be further described in detail below with reference to embodiments.

[0025] This invention provides a method for detecting liquid leakage in cold-plate liquid-cooled servers based on multiple physical characteristics, such as... Figure 1 As shown, it includes the following steps: Step S1: Simultaneously collect multi-dimensional physical features to obtain the real-time impedance characteristics of the leakage sensing rope and the real-time temperature characteristics of the monitoring area corresponding to the leakage sensing rope. Step S2: Based on the impedance and temperature characteristics, identify the correlation between impedance and temperature changes, perform liquid type differentiation judgment, and generate a liquid type determination result that distinguishes between condensate and coolant. Step S3: Based on the impedance characteristics, monitor and identify the baseline characteristic shift of the sensing rope after the impedance drop recovery event, perform leakage history tracing judgment, and generate historical tracing results indicating dried-up traceable leakage events. Step S4: Combine the liquid type determination results with the historical traceability results to generate the final alarm and early warning signals for leakage monitoring.

[0026] In this embodiment, multi-dimensional physical characteristics refer to characteristic parameters that characterize leakage events and environmental states from different physical quantity dimensions, including at least two dimensions: the impedance characteristics of the leakage sensing rope and the temperature characteristics of the monitoring area. Impedance characteristics can be characterized by the equivalent resistance or corresponding voltage division value caused by the change in the internal conductive medium of the sensing rope after contact with the liquid; temperature characteristics can be collected by a thermistor placed near the path of the sensing rope. The monitoring area corresponding to the leakage sensing rope refers to the physical space range in which the sensing rope is actually laid and performs the leakage detection function, exemplarily including locations such as below the cold plate joint, pipe connections, and the liquid collection tank at the bottom of the chassis. The location for collecting temperature characteristics should maintain a spatial correspondence with the path of the sensing rope to ensure that the correlation analysis between impedance changes and temperature changes has physical correspondence. The correlation characteristic of impedance with temperature changes refers to the difference in the rate or trend of impedance change caused by the sensing rope contacting liquids of different properties under different temperature conditions. This difference stems from the inherent difference in the temperature coefficient of conductivity of different liquids. Baseline characteristic offset refers to the irreversible change in the static reference value of the sensing rope before and after experiencing an impedance drop recovery event. This change is used to reflect whether there are conductive substances remaining on the surface of the sensing rope.

[0027] Synchronous acquisition of impedance and temperature characteristics can be implemented as follows: The main control unit simultaneously triggers analog-to-digital conversion of both the impedance and temperature acquisition channels at a fixed sampling period, with an exemplary sampling period of 100ms. The impedance acquisition channel adopts a voltage divider circuit structure, connecting the leakage sensing rope in series with a voltage divider resistor of known resistance to a reference voltage. An exemplary value for the voltage divider resistor is 10kΩ, and an exemplary value for the reference voltage is 3.3V. The connection point between the sensing rope and the voltage divider resistor is connected to the ADC input of the main control unit. The equivalent impedance of the sensing rope is calculated by acquiring the voltage value at this point. After each impedance acquisition, outlier filtering is performed simultaneously. If the deviation between the current acquisition value and the previous three valid acquisition values ​​exceeds ±30%, it is considered transient electromagnetic interference, and the outlier value is discarded, while the previous valid impedance data is used. The temperature acquisition channel uses a digital temperature sensor, communicating with the main control unit via an I²C bus. The temperature sensor is positioned near the sensing rope's laying path, within a spatial distance of no more than 5cm, ensuring that the acquired temperature value accurately reflects the temperature changes of the microenvironment in which the sensing rope is located. For sensing ropes less than or equal to 1 meter in length, temperature sensors are typically positioned at the geometric midpoint of the rope, as this location most accurately reflects the average temperature of the entire short rope. If the sensing rope is arranged along a significant temperature gradient (e.g., near the inlet and outlet of the cold plate), an additional sensor can be added at the point of maximum temperature gradient. For sensing ropes longer than 1 meter, to more accurately reflect the average temperature of the entire rope, a temperature sensor is placed every 50 cm. If the sensing rope is arranged along a significant temperature gradient (e.g., near the inlet and outlet of the cold plate), an additional temperature sensor can be added at the point of maximum temperature gradient. The data from these additional sensors are included in the calculation of the overall average, and when impedance drop occurs in the temperature gradient region, the correlation parameter calculation prioritizes the data collected by these additional sensors. When the sensing rope length exceeds 1 meter and the impedance change exhibits significant local characteristics, the correlation parameter calculation prioritizes the average temperature of the one or several temperature sensors spatially corresponding to the impedance drop region. If the spatial correspondence cannot be clearly defined, the arithmetic mean of all temperature sensors along the entire rope is used. The impedance and temperature data acquired synchronously are stamped with the same timestamp and stored in the data buffer for use in subsequent steps. The data buffer retains the most recent 30 seconds of valid synchronous data by default to meet the time window requirements for subsequent correlation parameter calculations.

[0028] To further enhance the reliability of the monitoring system, this embodiment also provides a self-test function for line breakage and a dynamic baseline calibration function. The self-test function for line breakage can be implemented by connecting a pull-down resistor in parallel at the input of the voltage follower. An exemplary value for the pull-down resistor is 1MΩ. When the sensor rope is normally connected, the voltage at the input of the ADC is determined by the voltage division between the sensor rope and the voltage divider resistor. When the sensor rope breaks, the input of the ADC is grounded through the pull-down resistor. When the voltage value is continuously lower than the line breakage threshold, a line breakage alarm is triggered. An exemplary value for the line breakage threshold is 50mV. The duration of the continuous judgment is exemplaryly set to 5 consecutive sampling cycles to avoid false alarms triggered by a single interference. The dynamic baseline calibration function collects the initial static baseline value of the sensor rope in a dry and clean state during system power-on initialization and stores it in non-volatile memory. During operation, every preset period, when the environment is confirmed to be stable and there is no leakage event, the static baseline value is re-collected, and the leakage impedance threshold is adjusted proportionally to compensate for the baseline drift caused by long-term aging of the sensor rope. An exemplary preset period is 24 hours. The quantitative judgment rules for environmental stability are as follows: the fluctuation range of the current ambient temperature and relative humidity does not exceed ±1℃ and ±5%RH, respectively, and this stable state lasts for at least 30 minutes. The quantitative judgment rules for no leakage events are as follows: within 1 hour prior to the current moment, the impedance value is always higher than the leakage impedance threshold, and no impedance drop recovery event occurs. If the above conditions are not met simultaneously, the calibration execution time is postponed until the conditions are met. Dynamic baseline calibration is only used to adjust the impedance threshold of the real-time leakage alarm. Historical tracing judgment always uses the initial static reference value of the first power-on or manual calibration as the sole reference. The two are completely isolated to avoid failure of the historical tracing function. When the historical tracing result is positive, the leakage impedance threshold after residue compensation is executed first, and the adjustment of the leakage impedance threshold by dynamic baseline calibration is suspended to avoid judgment distortion caused by double correction. When the maintenance personnel reset the initial static reference value by manually triggering the calibration process and the historical tracing result turns negative, the dynamic baseline calibration function is automatically restored. The initial static reference value can only be updated by manually triggering the calibration process after the maintenance personnel manually confirm that the sensor rope has been cleaned / replaced. The steps of the manually triggered calibration process include: First, maintenance personnel input authorization commands through the management system interface or local debugging interface. After the system verifies the operation permissions, it enters calibration mode. Then, the system prompts the maintenance personnel to confirm that the sensor rope is clean and dry, and provides the current ambient temperature, humidity, and real-time impedance values ​​for reference. Next, after the maintenance personnel confirm, the system continuously collects data for 5 minutes according to the standard measurement conditions of the initial static reference value, calculates the average value as the new initial static reference value, and automatically overwrites the original stored value. Finally, the system records the timestamp of this update, the operator's identification, and the comparison information between the old and new reference values, generating a calibration log for traceability. After the initial static reference value is updated, the system automatically clears all historical traceability records and residue accumulation level classifications, and restores the leakage impedance threshold to the initial calibration value.The above process avoids the historical reference benchmark from being overwritten, which could lead to functional failure, while also ensuring the auditability of update operations.

[0029] In this embodiment, the standard measurement conditions for static reference values ​​are uniformly defined as follows: the fluctuation range of ambient temperature and humidity shall not exceed ±1℃ / ±5%RH, and the fluctuation range of the sensor rope impedance value shall not exceed ±5% for 10 consecutive sampling cycles; the initial static reference value shall be continuously collected for 5 minutes and the average value shall be stored in non-volatile memory when the system is powered on for the first time and the sensor rope is clean and dry, so as to ensure the stability and repeatability of the reference value.

[0030] This embodiment constructs a synchronous acquisition architecture for impedance and temperature characteristics, acquiring the physical quantities of impedance and temperature in the sensor rope monitoring area under the same time reference. This provides a synchronous data foundation for distinguishing condensate and coolant based on the coupling relationship between impedance and temperature. The synchronous acquisition mechanism ensures the accuracy of the correlation parameter calculation and avoids distortion of correlation characteristics caused by inconsistent impedance and temperature sampling times. Limiting the temperature acquisition location to the corresponding monitoring area of ​​the sensor rope ensures the physical correspondence between temperature changes and impedance changes, providing a reliable data basis for subsequent liquid type differentiation based on differences in conductivity temperature coefficients. Simultaneously, by configuring self-checking for line breaks and dynamic baseline calibration functions, outlier filtering and baseline protection rules are added, improving the long-term reliability of the system and eliminating the impact of sensor rope malfunctions and aging drift on the monitoring results.

[0031] In one feasible implementation, such as Figure 2 As shown, in step S2, the specific steps for identifying the correlation characteristics of impedance changes with temperature and performing liquid type differentiation include: S21: When the absolute value of the temperature change within the preset time window is greater than or equal to the preset minimum temperature change threshold, calculate the correlation parameter characterizing the degree of correlation between the impedance change and the temperature change. The correlation parameter is the ratio of the relative impedance change to the temperature change. The relative impedance change is the ratio of the absolute value of the impedance change within the preset time window to the impedance value at the beginning of the time window. S22: Match the associated parameters with the preset first feature interval and second feature interval, wherein the first feature interval corresponds to the high temperature coefficient characteristic of conductivity, and the second feature interval corresponds to the low temperature coefficient characteristic of conductivity. S23: When the associated parameter falls into the first characteristic range, it is determined that the impedance drop is caused by condensate; when the associated parameter falls into the second characteristic range, it is determined that the impedance drop is caused by coolant.

[0032] In step S22, the determination of the first feature interval and the second feature interval includes: Based on the physical differences between condensate and coolant in terms of conductivity temperature coefficient, the first characteristic interval corresponds to the range of values ​​of the associated parameter where the conductivity of condensate increases exponentially with increasing temperature, and the second characteristic interval corresponds to the range of values ​​of the associated parameter where the rate of increase of coolant conductivity with increasing temperature is significantly lower than the rate of increase of condensate conductivity.

[0033] In step S23, after determining that the impedance drop is caused by the coolant, the following steps are also included: Step S24: Compare the real-time impedance value with the preset leakage impedance threshold, and perform joint verification by combining the impedance amplitude condition and the duration condition; when the liquid type is determined to be coolant and the impedance value is continuously lower than the leakage impedance threshold for more than the preset time threshold, mark the determination result as a valid coolant leakage determination result and include it in the valid output item of the liquid type determination result.

[0034] In this embodiment, the correlation parameter refers to an indicator used to quantify the degree of correlation between impedance change and temperature change. In one specific implementation, the correlation parameter is the ratio of the relative change in impedance to the change in temperature, and its calculation formula is K=(|ΔZ| / Z0) / |ΔTemp|, where ΔZ is the change in impedance value within a preset time window, Z0 is the real-time impedance value of the sensing rope at the beginning of the preset time window, and ΔTemp is the change in temperature value within the same time window. An exemplary time window is 5s; both ΔZ and ΔTemp are the numerical difference between the first and last valid sampling points within the time window. Using the relative change form can eliminate the interference of leakage contact length on the correlation parameter, so that the parameter only reflects the conductivity temperature coefficient characteristics of the liquid itself, ensuring that the characteristic range has universal applicability. First, outlier filtering is performed. If the deviation between the current sampled value and the previous three valid sampled values ​​exceeds ±30%, it is determined to be transient electromagnetic interference, and the current outlier value is removed. If there is still an obvious impedance / temperature change within the filtered window (i.e., the value of a single sampling point deviates from the previous sampling point by more than ±30%), the slope of the linear fitting of the data within the window is used as the basis for calculating the correlation parameters. For boundary scenarios with minimal temperature fluctuations, supplementary processing rules are implemented: When the absolute value of ΔTemp is less than the preset minimum temperature change threshold (exemplary value 0.2℃), the calculation of the correlation parameters for this period is skipped, and the valid judgment result of the previous period is used; if the temperature in the monitored area is stable for a long period of time and the temperature change is less than 0.2℃ within 30 seconds, the system will automatically expand the time window step by step (maximum not exceeding 30 seconds) to attempt to complete the calculation of the correlation parameters; if ΔTemp still does not reach the minimum temperature change threshold after expanding to the maximum window, the calculation of the correlation parameters for this period is abandoned, and a basic leakage warning is output based on the impedance amplitude and duration conditions. At the same time, the liquid type is marked as pending confirmation. The qualitative judgment of the liquid type will be completed after the subsequent temperature shows a valid fluctuation, avoiding calculation errors caused by the denominator approaching 0 and long-term stagnation of liquid type judgment in constant temperature scenarios.

[0035] The coolant conductivity growth rate being significantly lower than the condensate conductivity growth rate means that within the same temperature range, the coolant conductivity growth rate with temperature change does not exceed 50% of the condensate conductivity growth rate. Correspondingly, in terms of the associated parameter values, the maximum value of the associated parameter in the second characteristic interval does not exceed 55% of the minimum value of the associated parameter in the first characteristic interval, ensuring that the determination boundary for liquid type differentiation is clear and repeatable. This 5% difference is a reserved engineering margin to offset the influence of sensor measurement errors and environmental interference, ensuring that the determination boundary for liquid type differentiation is clear and repeatable.

[0036] The first and second characteristic intervals are pre-calibrated ranges of correlation parameters based on the physical differences in the temperature coefficients of conductivity between condensate and coolant. The concentration of conductive ions in condensate is extremely low, and its conductivity is primarily determined by the dissociation constant of water. As temperature increases, the degree of water dissociation increases exponentially, and conductivity increases exponentially with temperature. This corresponds to a faster rate of decrease in the equivalent impedance of the sensing rope, ultimately manifesting as a high change in the correlation parameters, corresponding to the first characteristic interval. Within the typical operating temperature range of 20℃ to 60℃ for the cold-plate liquid-cooled server adapted to this invention, the conductivity of condensate shows a significant exponential increase with temperature. The impedance data of the sensing rope can be directly collected through point-by-point temperature change testing, and the correlation parameters can be calculated to calibrate the range of values ​​for the first characteristic interval. The coolant (such as a mixture of ethylene glycol and deionized water) already contains a high concentration of conductive ions. The conductivity is mainly determined by the ion concentration. The increase in temperature only intensifies the thermal motion of the ions. The rate of increase in conductivity with increasing temperature is significantly lower than the rate of increase in conductivity of condensate. The corresponding decrease rate of the equivalent impedance of the sensing rope is even slower, and ultimately it shows a low change in the correlation parameters, corresponding to the second characteristic interval.

[0037] It should be noted that the temperature coefficient of conductivity is the essential physical reason for the differentiation between the two types of liquids, and the correlation parameter K is a quantitative judgment index calculated based on the measured impedance of the sensing rope. The two are causally related rather than numerically equivalent. The boundary values ​​of all characteristic intervals are directly calibrated through actual temperature change leakage experiments of the sensing rope, without the need for conversion using the theoretical temperature coefficient of liquid conductivity.

[0038] The leakage impedance threshold is the critical impedance value used to determine whether leakage has occurred. It can be calibrated experimentally based on the sensor rope specifications and coolant type. An exemplary value is 30% to 50% of the reference impedance value of the sensor rope in a dry state. This value is suitable for industry-standard dual-core parallel leakage sensor ropes and coolants with a 25% ethylene glycol volume ratio. For four-core sensor ropes, the threshold can be adjusted to 40% to 60% of the reference value. The upper limit boundary of the second characteristic interval is the critical value of the correlation parameter that distinguishes between the coolant and the transition interval.

[0039] The characteristic interval was calibrated using a measured method. Specifically, in a controlled temperature and humidity chamber, quantitative drip tests were performed on the same specification sensor rope using ethylene glycol coolant and deionized water of known concentrations. An exemplary coolant was a mixture of ethylene glycol and deionized water at a volume ratio of 25%:75%. During the test, impedance data as a function of temperature was recorded point-by-point in 5°C increments within the range of 20°C to 60°C. Data was collected after each temperature point was held for 10 minutes, and the correlation parameter K value at each temperature point was calculated. After at least 10 repeated experiments, outliers exceeding three standard deviations were removed using the Grubbs test. The distribution ranges of K values ​​corresponding to condensate and coolant were statistically analyzed, and the statistical boundaries of the 95% confidence intervals of their respective distribution ranges were taken as the boundary values ​​of the characteristic interval. For a dual-core parallel sensing rope (electrode spacing 1.5mm, insulation material is PVC), the exemplary calibration results of the correlation parameters are: the correlation parameter value for the first characteristic interval is 1.2% / ℃~2.3% / ℃, and the correlation parameter value for the second characteristic interval is 0.2% / ℃~0.7% / ℃. Supplementary processing rules for the transitional zone scenario of leakage due to a mixture of condensate and coolant: When the correlated parameter falls into the transitional zone between the first and second characteristic zones, the observation window is automatically extended to 20 seconds to continuously monitor the trend of the correlated parameter with temperature; if the correlated parameter generally increases with rising temperature and enters the first characteristic zone, it is determined that condensate is dominant. The overall increasing trend refers to the average growth rate of the correlated parameter within the observation window being greater than 0.05% / (°C). s); If the correlation parameter remains flat without exponential growth, it is determined that the coolant is dominant. "Remaining flat" means that the difference between the maximum and minimum values ​​of the correlation parameter within the observation window does not exceed ±10% of the average value. This is marked as a priority determination result to be reviewed and included in the valid output of the coolant type determination result for priority processing in subsequent fusion determination stages. The transition range of the correlation parameter is, for example, 0.7% / ℃ to 1.2% / ℃.

[0040] An exemplary implementation of the validity verification method is as follows: After outputting the coolant leakage judgment result, further check whether the real-time impedance value is lower than a preset leakage impedance threshold. If it is lower than the threshold, start a timer. When the duration for which the impedance value remains lower than the threshold reaches a preset time threshold, mark the judgment result as a valid coolant leakage judgment result; if the impedance value rises back above the threshold before the timer completes, cancel the timer and determine it as transient interference, marking the judgment result as invalid. An exemplary value for the preset time threshold is 3 seconds.

[0041] This embodiment transforms the fundamental physical difference between condensate and coolant in terms of conductivity temperature coefficient into quantifiable judgment logic by introducing a calculation mechanism for impedance and temperature-related parameters and a feature interval matching mechanism. Utilizing the physical characteristics that condensate conductivity increases exponentially with temperature while coolant conductivity increases gradually, accurate differentiation between the two liquids is achieved within a temperature range where impedance amplitudes overlap, effectively reducing the probability of fundamental misjudgments caused by existing technologies relying solely on impedance amplitude judgment. Furthermore, processing rules for boundary scenarios and mixed operating conditions are supplemented, and the liquid type differentiation results are jointly validated with impedance amplitude and duration conditions. A three-dimensional judgment system of qualitative differentiation, quantitative confirmation, and timed confirmation is constructed, further reducing the probability of invalid judgments caused by transient interference and condensate events, providing a highly reliable input basis for subsequent fusion judgments.

[0042] In one feasible implementation, such as Figure 3 As shown, in step S3, the specific steps for monitoring and identifying the baseline characteristic shift of the sensing rope after the impedance drop recovery event and performing leakage history tracing determination include: Step S31: When the sensing rope is in an initial clean state, measure and store the initial static reference value of the sensing rope. Step S32: Monitor the occurrence and end of the impedance drop recovery event. The impedance drop recovery event refers to the complete process in which the impedance value drops from the normal range to below the preset leakage impedance threshold and continues to exceed the preset effective drop time threshold, and then rises back to the normal range. The normal range refers to the state range in which the impedance value is higher than the leakage impedance threshold. Step S33: After the impedance drop recovery event ends and the sensor rope surface is dry and stable, measure the current static reference value of the sensor rope. Step S34: Calculate the decrease of the current static reference value relative to the initial static reference value. When the decrease exceeds the preset offset threshold, it is determined that there are irreversible conductive residues on the surface of the sensing rope due to the evaporation of coolant, and a historical traceability warning of dried coolant leakage is generated.

[0043] In step S34, the specific steps for determining whether there are irreversible conductive residues on the surface of the sensing rope due to coolant evaporation include: S341: Within a preset time period after the impedance drop recovery event ends, collect the static reference value of the sensing rope under at least two different ambient humidity conditions, and identify the type of deviation of the current static reference value relative to the initial static reference value: If the current static reference value is continuously lower than the initial static reference value, and its value decreases as the ambient humidity increases and increases as the ambient humidity decreases, it is an irreversible deviation caused by hygroscopic conductive residues formed after the evaporation of non-volatile components in the coolant; if the current static reference value gradually rises back to near the initial static reference value as the ambient humidity decreases, it is a reversible deviation caused by the evaporation of condensate. S342: When an irreversible offset is identified and the reduction exceeds a preset offset threshold, it is determined that there are irreversible conductive residues on the surface of the sensing rope, and a historical traceability warning for dried coolant leakage is generated.

[0044] In step S34, the offset threshold is pre-calibrated based on the sensor rope type and coolant component concentration. The ratio of the offset threshold to the initial static reference value is a preset proportional coefficient. The range of the preset proportional coefficient is determined based on the content of non-volatile components in the coolant.

[0045] In this embodiment, the initial static reference value refers to the static impedance value or corresponding leakage current value measured according to the aforementioned specified standard measurement conditions, under the condition that the sensing rope is clean and has no history of leakage. This value is obtained by measuring after the system is first powered on or after the maintenance personnel confirm that the sensing rope is clean, and is stored in non-volatile memory as a reference benchmark for historical traceability. To improve the accuracy of the reduction calculation, the current static reference value is measured under conditions of similar ambient humidity to that at the time of the initial static reference value measurement. If the ambient humidity at the time of measurement differs significantly from the reference measurement conditions, the measured value can be normalized using a pre-calibrated humidity-impedance correction coefficient before calculating the reduction. The humidity-impedance correction coefficient can be obtained by fitting the static reference value of the clean sensing rope under different humidity conditions to ensure that the reduction only reflects the influence of coolant residue.

[0046] The valid judgment rule for impedance drop recovery events is clearly defined as follows: an impedance value that remains below the leakage impedance threshold for more than 1 second is considered a valid drop event, excluding invalid triggers caused by transient electromagnetic interference. After the impedance recovers to the normal range, the stabilization judgment time is dynamically determined based on the current relative humidity of the environment. The fluctuation range of the static impedance value of the sensor rope is used as the basis for judging dryness stability: when the relative humidity of the environment is ≤40%, the fluctuation range of the static impedance value of the sensor rope does not exceed ±5% within 30 minutes, which is considered as dryness stability; when the relative humidity of the environment is 40%~70%, the stabilization judgment time is extended to 60 minutes, and the fluctuation range of the static impedance value of the sensor rope must not exceed ±5%; when the relative humidity of the environment is ≥70%, the stabilization judgment time is extended to 120 minutes, and the fluctuation range of the static impedance value of the sensor rope must not exceed ±5%. During this period, the surface drying of the sensor rope can be accelerated by jogging heating, with the heating power set to 2W and the heating time not exceeding 30 minutes as an example. When the ambient humidity fluctuates within ±2%RH of the interval boundary, the stability judgment time corresponding to the higher humidity interval is adopted; for example, when the ambient humidity fluctuates between 38% and 42%RH, the stability judgment time corresponding to the 40% to 70% interval is adopted. After the fluctuation amplitude requirement under the corresponding humidity condition is met, it is considered that the impedance drop recovery event has ended, and the static reference value measurement stage begins. The normal interval refers to the state interval where the impedance value is higher than the leakage impedance threshold. Irreversible conductive residues refer to solid or viscous residues formed on the surface of the sensor rope insulation layer after the evaporation of water, such as ethylene glycol and corrosion inhibitors in the coolant. These residues are hygroscopic and ionicly conductive. Even if the sensor rope surface is dry, the residues will still form microscopic conductive channels on the surface of the insulation layer, resulting in a permanent increase in the baseline leakage current of the sensor rope. Reversible offset refers to the type of static reference value change caused by the evaporation of condensate. The condensate is mainly composed of pure water or water containing trace impurities. After evaporation, no residue adheres to the sensor rope surface. The static reference value recovers to an acceptable range near the initial static reference value as the water evaporates completely. The acceptable range is defined as within ±10% of the initial static reference value. Irreversible offset refers to the type of static reference value change caused by coolant evaporation. Its typical characteristic is that after the event ends and the sensor rope surface dries, the static reference value remains within a certain range below the initial static reference value and does not completely return to the initial level. Offset type refers to the determination of the category of static reference value change, including irreversible offset and reversible offset. Preset duration refers to an observation time window calculated from the end of the impedance drop recovery event. Within this window, the system continuously monitors the trend of static reference value changes to determine the offset type. An example value is 24 hours. The recovery range refers to the range of static reference value recovery used to determine reversible offset, i.e., within ±10% of the initial static reference value.The offset threshold is the critical value for determining whether the reduction in the static reference value constitutes irreversible residue. When the reduction exceeds this threshold, the system determines that there is irreversible conductive residue on the surface of the sensing rope. The formula for calculating the reduction is: Reduction = (Initial static reference value - Current static reference value) / Initial static reference value × 100%.

[0047] The identification of offset types is based on the following physical mechanism: Ethylene glycol and corrosion inhibitor components in the coolant irreversibly adhere to the surface of the sensing rope after water evaporation, forming hygroscopic conductive residues. These residues, even in a dry state, create additional leakage current channels, causing irreversible offsets in the static reference value. Conversely, after condensate evaporates, no residue remains, and the static reference value can return to its initial state. Based on these mechanistic differences, the criteria for determining irreversible offset are: within a preset time period, a static reference value is collected every 2 hours; the current static reference value remains consistently lower than the initial static reference value; and the impedance value decreases with increasing ambient humidity and increases with decreasing ambient humidity, exhibiting a humidity response pattern consistent with the characteristics of hygroscopic residues, and cannot fall back to within ±10% of the initial static reference value. The criteria for determining reversible offset are: the static reference value gradually falls back to the range near the initial static reference value as ambient humidity decreases. If the static reference value offset continues to increase and there is no clear time correspondence with the impedance drop recovery event, it is determined to be baseline drift caused by natural aging of the sensing rope or environmental dust or oil pollution, and will not be included in the leakage history tracing results to avoid false warnings caused by non-leakage factors.

[0048] The offset threshold is calibrated as follows: it is pre-calibrated based on the sensor rope type and coolant component concentration, and the ratio of the offset threshold to the initial static reference value is the preset proportional coefficient. For a coolant with a 25%:75% volume ratio of ethylene glycol to deionized water, the exemplary range of the preset proportional coefficient is 1.5~2.0. For coolants with different ethylene glycol concentrations, the preset proportional coefficient can be adjusted according to the content of non-volatile components: the higher the ethylene glycol concentration in the coolant, the higher the content of non-volatile components, and the greater the amount of residue after water evaporation, the larger the corresponding value of the preset proportional coefficient. The exemplary correspondence is: 30% ethylene glycol coolant corresponds to a proportional coefficient of 2.0~2.5, 40% ethylene glycol coolant corresponds to a proportional coefficient of 2.5~3.0, and 50% ethylene glycol coolant corresponds to a proportional coefficient of 3.0~3.5. For different types of sensor ropes, due to differences in insulation material and surface structure, their sensitivity to residues varies, and the corresponding preset proportional coefficient can be determined through experimental calibration.

[0049] This embodiment defines an initial static baseline value as a reference point and an impedance drop recovery event as a prerequisite for triggering historical tracing. Based on the inherent physical difference between coolant and condensate after evaporation, it establishes a logic to distinguish between irreversible and reversible offsets. By clarifying the effective event judgment rules, the reduction calculation formula, and the quantitative judgment standard for offset type, it monitors the trend of static baseline value changes after the impedance drop recovery event ends. The combination of the baseline's continuous reduction characteristic and humidity response characteristics serves as the core basis for determining irreversible offsets, effectively distinguishing between coolant drying residue and condensate evaporation scenarios. This solution endows the monitoring system with the ability to trace dried-out leakage events, expanding the monitoring dimension beyond existing technologies that only monitor real-time impedance status. It allows maintenance personnel to know if a server has experienced a minor leak, enabling targeted preventative maintenance. Simultaneously, the offset threshold calibration method considers differences in sensor rope type and coolant component concentration, allowing the historical tracing function to adapt to different specifications of sensor ropes and different coolant formulations, enhancing the solution's versatility and engineering feasibility.

[0050] In one feasible implementation, step S4, which involves determining the fusion liquid type and comparing the historical traceability results, includes the following specific steps: S41: Establish a fusion judgment logic hierarchy, wherein the liquid type judgment result is used to distinguish between condensate events and coolant events in real time, and the historical traceability result is used to identify the state of residues already existing on the surface of the sensing rope. S42: When the historical tracing result is positive, the leakage impedance threshold is adjusted downward to compensate for the reduction in the baseline impedance of the sensing rope caused by residue and to maintain the leakage detection sensitivity; the adjustment is provided with a preset lower limit to prevent over-adjustment.

[0051] In step S42, the downward adjustment range of the leakage impedance threshold is dynamically determined based on the graded and quantified residue accumulation level identified by the historical traceability results. The higher the residue accumulation level, the greater the adjustment range. The adjusted leakage impedance threshold is not lower than the preset lower limit. When the leakage impedance threshold has been adjusted to the preset lower limit but still needs further compensation, a sensor rope contamination replacement warning is triggered.

[0052] In this embodiment, the fusion judgment logic hierarchy refers to establishing two logic levels with information feedback relationship between liquid type differentiation judgment and historical traceability judgment. The first level is the liquid type differentiation level, which is used to analyze the impedance drop event that occurs in real time and solve the problem of whether the impedance drop is caused by condensate or coolant. Its output includes three categories: condensate judgment result, coolant judgment result, and priority judgment result for the transition interval. The priority judgment result for the transition interval is processed as medium risk priority in the fusion level, triggering early warning first and continuously monitoring the trend. The second level is the historical traceability level, which is used to record and analyze the historical state of the sensor rope and solve the problem of whether there are irreversible conductive residues on the surface of the sensor rope due to the drying of coolant. Its output is the historical traceability result. An information feedback channel is set between the two levels. The output of the historical traceability level can be fed back to the judgment link of the real-time leakage alarm to adjust the leakage impedance threshold and form a closed-loop adaptive adjustment mechanism.

[0053] The residue status refers to the historical tracking level's identification result regarding the presence of irreversible conductive residues on the sensor rope surface, including negative and positive statuses. A negative status indicates no historical residues on the sensor rope surface, and the deviation between the current static baseline value and the initial static baseline value does not exceed a preset deviation threshold. A positive status indicates the presence of historical residues on the sensor rope surface, and the deviation between the current static baseline value and the initial static baseline value has exceeded a preset deviation threshold. In the positive status, further information on the degree of residue accumulation is included.

[0054] The quantitative grading rules for the degree of residue accumulation are clearly defined as follows: Quantitative grading is based on the magnitude of the decrease in the current static baseline value relative to the initial static baseline value. Level 1 accumulation: the decrease exceeds a preset offset threshold but is less than or equal to 15% of the initial value; Level 2 accumulation: the decrease is 15% to 30% of the initial value; Level 3 accumulation: the decrease exceeds 30% of the initial value. The grading results are directly used for dynamic matching of the downward adjustment range of the leakage resistance threshold.

[0055] An exemplary mapping relationship is as follows: When the first level of accumulation (offset of 10% of the initial value) occurs, the leakage impedance threshold is adjusted downwards to 80% of the original threshold; when the second level of accumulation (offset of 20% of the initial value) occurs, the leakage impedance threshold is adjusted downwards to 65% of the original threshold. To prevent over-adjustment from misjudging environmental noise as leakage, a preset lower limit is set for the adjustment range. The exemplary lower limit is 50% of the original threshold, meaning the adjusted leakage impedance threshold is not lower than 50% of the original threshold. Supplementary processing rules for the extreme pollution condition of the third level of accumulation: When the leakage impedance threshold has been adjusted to the preset lower limit, but the degree of residue accumulation continues to deepen, a sensor rope contamination replacement warning is triggered, prompting maintenance personnel to check and replace the sensor rope to avoid the risk of missed detection due to long-term severe contamination.

[0056] This embodiment establishes a fusion-based judgment logic hierarchy, constructing liquid type differentiation judgment and historical tracing judgment as two logical levels with information feedback relationships, clarifying the functional positioning and interrelationships of each level. By identifying the state of residue and the degree of accumulation of quantitative classification at the historical tracing level, and feeding this information back to the judgment stage of real-time leakage alarm, when the historical tracing result is positive, the leakage impedance threshold is adjusted downward to compensate for the reduction in sensor rope baseline impedance caused by residue and maintain leakage detection sensitivity. The adjustment range is dynamically determined according to the degree of residue accumulation, so that the compensation intensity matches the degree of contamination, avoiding the undercompensation or overcompensation problems that may be caused by fixed compensation, and supplementing the rules for triggering sensor rope contamination replacement warnings under extreme contamination conditions. This fusion linkage mechanism deeply coordinates the two core invention points of liquid type differentiation and historical tracing, enabling the monitoring system to have the ability to autonomously perceive changes in sensor rope state and adaptively adjust the leakage detection threshold, eliminating the risk of subsequent minor leakage underreporting caused by residue accumulation, and further improving the monitoring reliability under long-term operating conditions.

[0057] For example, the present invention employs a graded triggering logic based on progressive risk levels, outputting warning signals and alarm signals sequentially from low to high according to the degree of harm determined by the judgment result, and matching corresponding handling actions: Level 1 Warning (Low Risk): The trigger condition is an independent event where the liquid type is determined to be condensate. The criteria for determining an independent event is that the interval between two impedance drops exceeds 1 hour. The system only generates a warning log and records the event time, ambient temperature and humidity, and location information, without triggering any hardware operation. When 3 or more independent Level 1 warnings occur in the same monitoring area within 7 days, it will automatically be upgraded to Level 2 warning. The cancellation condition is that the condensate evaporates and the sensor rope impedance returns to the normal range and remains so for 10 minutes before it is automatically cancelled.

[0058] Level 2 warning (medium risk): The triggering condition is that any of the following conditions are met: (1) The degree of accumulation of sensor rope residue is determined to be Level 1 or Level 2; (2) The cumulative number of Level 1 warning events reaches the standard for upgrading; (3) The liquid type determination result is pending review priority; The system generates an operation and maintenance warning and pushes it to the management platform. When triggered by condition (1), the operation and maintenance personnel are prompted to check the cold plate joint and pipeline sealing of the corresponding location during the next routine inspection. When triggered by condition (2), the operation and maintenance personnel are prompted to check the dehumidification capacity of the computer room air conditioner and the sealing status of the server chassis. When triggered by condition (3), the operation and maintenance personnel are prompted to pay close attention to the leakage risk of the monitored area and there is no need to immediately shut down the machine for disposal. The upgrading rule is: When the degree of accumulation of sensor rope residue rises to Level 3 under condition (1), it will automatically upgrade to Level 2 alarm. If the temperature fluctuation is clearly determined to be coolant under condition (3), it will automatically upgrade to Level 1 alarm. The downgrade / removal rules are as follows: When triggered by situation (1), the maintenance personnel manually confirm that there are no hidden dangers or replace the sensor rope and reset the initial static baseline value before the downgrade is lifted; when triggered by situation (2), the first-level warning is automatically lifted if no further first-level warning occurs within the next 7 days; when triggered by situation (3), if it is subsequently determined to be condensate, the downgrade is recorded as a first-level warning log. If no effective temperature fluctuation is completed within 24 hours, the second-level warning status is maintained and a manual on-site inspection is prompted for confirmation.

[0059] Level 1 Alarm (High Risk): The trigger condition is a valid real-time coolant leakage determination result; the system immediately triggers a local audible and visual alarm, and simultaneously pushes the leakage location, leakage degree, and estimated impact range to the mobile terminal of the maintenance personnel, supporting linkage triggering of server power reduction protection; when the leakage duration exceeds 30 seconds and the overall impedance value drops by more than 20% within 30 seconds, it automatically upgrades to a Level 2 alarm; the release condition is that the leakage fault is resolved, the sensor rope is dry, and the impedance returns to the normal range, and the alarm is released after manual confirmation by the maintenance personnel.

[0060] Level 2 alarm (extremely high risk): The triggering condition is that any of the following situations are met: (1) The degree of accumulation of sensor rope residue is judged to be Level 3 and a prompt is made that it needs to be replaced immediately; (2) Level 1 alarm event is upgraded; the system triggers the highest priority audible and visual alarm, forcibly pushes an emergency maintenance work order, and can trigger the server to shut down safely according to the preset strategy to avoid hardware short circuit damage; the release condition is that after the fault is eliminated and the contaminated sensor rope is replaced and the initial static baseline value of the system is reset, the release is confirmed manually by the maintenance manager.

[0061] When the same monitoring area has both historical traceability positive results and residue accumulation at level one or two, and real-time coolant leakage determination results, a level two alarm is triggered directly, skipping the early warning and level one alarm processes, ensuring that high-risk events are responded to first.

[0062] When the same level of early warning / alarm is triggered multiple times in the same monitoring area within 24 hours, the system merges them into a single event push, updating only the number of times the event occurred and its latest status to avoid information overload.

[0063] This invention also provides a liquid leakage monitoring system for cold plate liquid-cooled servers based on multiple physical characteristics, such as... Figure 4 As shown, it includes: an impedance acquisition module 10, a temperature acquisition module 20, a memory 30, a processor 40, and a multi-physical feature-based liquid cooling server leakage monitoring program stored in the memory 30 and capable of running on the processor 40. The impedance acquisition module 10 is used to acquire the real-time impedance characteristics of the leakage sensing rope. The temperature acquisition module 20 is used to acquire the real-time temperature characteristics of the monitoring area corresponding to the leakage sensing rope; The processor 40 is connected to the impedance acquisition module 10, the temperature acquisition module 20, and the memory 30 respectively. The liquid leakage monitoring program for the cold plate liquid-cooled server based on multiple physical characteristics is configured to implement the steps of the liquid leakage monitoring method for the cold plate liquid-cooled server based on multiple physical characteristics.

[0064] The present invention provides a leakage monitoring system for cold-plate liquid-cooled servers based on multiple physical characteristics. Employing the leakage monitoring method for cold-plate liquid-cooled servers based on multiple physical characteristics described in the above embodiments, it can distinguish between misjudgments caused by impedance overlap between condensate and coolant in a specific temperature range, enabling the tracing of even dried-out trace leakage events. Furthermore, it adaptively compensates for changes in sensor rope sensitivity through a fusion linkage mechanism. Compared with existing technologies, the beneficial effects of the leakage monitoring system for cold-plate liquid-cooled servers based on multiple physical characteristics provided by the present invention are the same as those of the leakage monitoring method for cold-plate liquid-cooled servers based on multiple physical characteristics provided in the above embodiments. Other technical features of the leakage monitoring system for cold-plate liquid-cooled servers based on multiple physical characteristics are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0065] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for detecting leakage in a cold-plate liquid-cooled server based on multiple physical characteristics, characterized in that, Includes the following steps: S1: Synchronously collect multi-dimensional physical features to obtain the real-time impedance characteristics of the leakage sensing rope and the real-time temperature characteristics of the monitoring area corresponding to the leakage sensing rope. S2: Based on the impedance and temperature characteristics, identify the correlation characteristics of impedance changes with temperature, perform liquid type differentiation judgment, and generate a liquid type differentiation result that distinguishes between condensate and coolant. S3: Based on the impedance characteristics, monitor and identify the baseline characteristic offset of the sensing rope after the impedance drop recovery event, perform leakage history tracing judgment, and generate historical tracing results indicating that the dried-up trace leakage event has been generated. S4: Integrate the liquid type determination results with the historical traceability results to generate the final alarm and early warning signals for leakage monitoring.

2. The method for detecting leakage in a cold-plate liquid-cooled server based on multiple physical characteristics according to claim 1, characterized in that, In step S2, the specific steps for identifying the correlation characteristics of impedance changes with temperature and performing liquid type differentiation include: S21: When the absolute value of the temperature change within the preset time window is greater than or equal to the preset minimum temperature change threshold, calculate the correlation parameter characterizing the degree of correlation between the impedance change and the temperature change. The correlation parameter is the ratio of the relative impedance change to the temperature change. The relative impedance change is the ratio of the absolute value of the impedance change within the preset time window to the impedance value at the beginning of the time window. S22: Match the associated parameters with the preset first feature interval and second feature interval, wherein the first feature interval corresponds to the high temperature coefficient characteristic of conductivity, and the second feature interval corresponds to the low temperature coefficient characteristic of conductivity. S23: When the associated parameter falls into the first characteristic range, it is determined that the impedance drop is caused by condensate; when the associated parameter falls into the second characteristic range, it is determined that the impedance drop is caused by coolant.

3. The method for monitoring leakage in a cold-plate liquid-cooled server based on multiple physical characteristics according to claim 2, characterized in that, In step S22, the determination of the first feature interval and the second feature interval includes: Based on the physical differences between condensate and coolant in terms of conductivity temperature coefficient, the first characteristic interval corresponds to the range of values ​​of the associated parameter where the conductivity of condensate increases exponentially with increasing temperature, and the second characteristic interval corresponds to the range of values ​​of the associated parameter where the rate of increase of coolant conductivity with increasing temperature is significantly lower than the rate of increase of condensate conductivity.

4. The method for detecting leakage in a cold-plate liquid-cooled server based on multiple physical characteristics according to claim 2, characterized in that, In step S23, after determining that the impedance drop is caused by the coolant, the following steps are also included: S24: Compare the real-time impedance value with the preset leakage impedance threshold, and perform joint verification by combining the impedance amplitude condition and the duration condition; when the liquid type is determined to be coolant and the impedance value is continuously lower than the leakage impedance threshold for more than the preset time threshold, mark the determination result as a valid coolant leakage determination result and include it in the valid output item of the liquid type determination result.

5. The method for detecting leakage in a cold-plate liquid-cooled server based on multiple physical characteristics according to claim 1, characterized in that, In step S3, the specific steps for monitoring and identifying the baseline characteristic shift of the sensing rope after the impedance drop recovery event and performing leakage history tracing determination include: S31: When the sensing rope is in an initial clean state, measure and store the initial static reference value of the sensing rope. S32: Monitor the occurrence and end of impedance drop recovery events. The impedance drop recovery event refers to the complete process in which the impedance value drops from the normal range to below the preset leakage impedance threshold and continues to exceed the preset effective drop time threshold, and then rises back to the normal range. The normal range refers to the state range in which the impedance value is higher than the leakage impedance threshold. S33: After the impedance drop recovery event ends and the sensor rope surface is dry and stable, measure the current static reference value of the sensor rope. S34: Calculate the decrease of the current static reference value relative to the initial static reference value. When the decrease exceeds the preset offset threshold, determine that there are irreversible conductive residues on the surface of the sensing rope due to the evaporation of coolant, and generate a historical traceability warning for the dried coolant leakage.

6. The method for detecting leakage in a cold-plate liquid-cooled server based on multiple physical characteristics according to claim 5, characterized in that, In step S34, the specific steps for determining whether there are irreversible conductive residues on the surface of the sensing rope due to coolant evaporation include: S341: Within a preset time period after the impedance drop recovery event ends, collect the static reference value of the sensing rope under at least two different ambient humidity conditions, and identify the type of deviation of the current static reference value relative to the initial static reference value: If the current static reference value is continuously lower than the initial static reference value, and its value decreases as the ambient humidity increases and increases as the ambient humidity decreases, it is an irreversible deviation caused by hygroscopic conductive residues formed after the evaporation of non-volatile components in the coolant; if the current static reference value gradually rises back to near the initial static reference value as the ambient humidity decreases, it is a reversible deviation caused by the evaporation of condensate. S342: When an irreversible offset is identified and the reduction exceeds a preset offset threshold, it is determined that there are irreversible conductive residues on the surface of the sensing rope, and a historical traceability warning for dried coolant leakage is generated.

7. The method for detecting leakage in a cold-plate liquid-cooled server based on multiple physical characteristics according to claim 5, characterized in that, In step S34, the offset threshold is pre-calibrated based on the sensor rope type and coolant component concentration. The ratio of the offset threshold to the initial static reference value is a preset proportional coefficient. The range of the preset proportional coefficient is determined based on the content of non-volatile components in the coolant.

8. The method for detecting leakage in a cold-plate liquid-cooled server based on multiple physical characteristics according to claim 1, characterized in that, In step S4, the specific steps for determining the fusion liquid type and tracing historical results include: S41: Establish a fusion judgment logic hierarchy, wherein the liquid type judgment result is used to distinguish between condensate events and coolant events in real time, and the historical traceability result is used to identify the state of residues already existing on the surface of the sensing rope. S42: When the historical tracing result is positive, the leakage impedance threshold is adjusted downward to compensate for the reduction in the baseline impedance of the sensing rope caused by residue and to maintain the leakage detection sensitivity; the adjustment is provided with a preset lower limit to prevent over-adjustment.

9. The method for detecting leakage in a cold-plate liquid-cooled server based on multiple physical characteristics according to claim 8, characterized in that, In step S42, the downward adjustment range of the leakage impedance threshold is dynamically determined based on the graded and quantified residue accumulation level identified by the historical traceability results. The higher the residue accumulation level, the greater the adjustment range. The adjusted leakage impedance threshold is not lower than the preset lower limit. When the leakage impedance threshold has been adjusted to the preset lower limit but still needs further compensation, a sensor rope contamination replacement warning is triggered.

10. A liquid leakage monitoring system for cold-plate type liquid-cooled servers based on multiple physical characteristics, characterized in that, include: Impedance acquisition module, temperature acquisition module, memory, processor, and a multi-physical feature-based liquid cooling server leakage monitoring program stored in the memory and capable of running on the processor; The impedance acquisition module is used to acquire the real-time impedance characteristics of the leakage sensing rope. The temperature acquisition module is used to acquire the real-time temperature characteristics of the monitoring area corresponding to the leakage sensing rope. The processor is connected to the impedance acquisition module, the temperature acquisition module, and the memory respectively. The liquid leakage monitoring program for cold plate liquid-cooled servers based on multiple physical characteristics is configured to implement the steps of the liquid leakage monitoring method for cold plate liquid-cooled servers based on multiple physical characteristics as described in any one of claims 1 to 9.

Images