A method, system, terminal and medium for on-line monitoring of conductivity probe performance and prediction of remaining useful life

Abstract

The application discloses a kind of conductivity probe performance online monitoring and residual service life prediction method, system, terminal and medium, method includes: obtaining voltage transient response slope reference range and gas phase voltage amplitude reference range;The voltage signal time series data collected is handled to sliding time window, and the voltage transient response slope average value and gas phase voltage amplitude average value of all bubble events in current sliding time window are obtained;Based on voltage transient response slope reference range, gas phase voltage amplitude reference range, voltage transient response slope average value and gas phase voltage amplitude average value, obtain health index;Based on health index, form health index time series, predict health index time series by time series regression prediction model, execute failure time detection and residual service life calculation, obtain conductivity probe residual service life.Through the above steps, the application evaluates the health status of conductivity probe and predicts its residual service life.

Description

Technical Field

[0001] This invention relates to the field of reliability engineering, and in particular to a method, system, terminal, and medium for online monitoring of conductivity probe performance and prediction of remaining service life. Background Technology

[0002] Against the backdrop of iterative upgrades in advanced nuclear energy technology and the large-scale development of clean energy, fourth-generation nuclear energy systems, with their advantages of high safety, high fuel utilization, and low radioactive waste production, have become the core direction for overcoming the safety shortcomings of traditional nuclear energy, improving uranium resource utilization efficiency, and promoting the green and low-carbon development of nuclear energy. They are also a key area for current nuclear energy research and engineering implementation. Lead-cooled fast reactors, as the reactor type with the greatest potential for large-scale commercialization among fourth-generation nuclear energy systems, rely on the excellent physicochemical properties of liquid lead-based alloy coolants—high boiling point, atmospheric pressure operation, and strong chemical inertness—to fundamentally avoid safety risks such as coolant boiling and violent chemical reactions. Simultaneously, they can achieve nuclear fuel breeding and long-life transmutation of high-level actinide nuclides, improving uranium resource utilization and reducing the hazards of nuclear waste. This represents the core development path for the next generation of advanced nuclear energy. Lead-cooled fast reactors abandon the intermediate isolation loop of traditional reactor types, adopting a pool-type integrated structure design. The main heat exchanger / steam generator is directly placed within the primary loop lead-based alloy molten pool, simplifying the system structure and reducing equipment operation and maintenance costs. However, this also makes the heat transfer tubes the sole barrier separating the high-temperature, low-pressure lead-based alloy in the primary loop from the high-pressure heat exchanger in the secondary loop. Under long-term corrosion, erosion, flow-induced vibration, and thermal stress coupling conditions, the heat transfer tubes are prone to damage and cracking, leading to a heat exchanger tube rupture (HXTR / Steam Generator Tube Rupture, SGTR) accident. After an accident, the secondary loop working fluid will be injected into the primary loop under pressure differential, forming a large amount of non-condensable gas after flash evaporation or phase change. Ultimately, a two-phase flow of liquid lead-based alloy and non-condensable gas will form within the primary loop lead-based alloy molten pool. These two-phase flow bubbles can enter the reactor core with the coolant, causing problems such as abnormal core reactivity and decreased local heat transfer performance, seriously threatening the operational safety of lead-cooled fast reactors. Therefore, research on the accident mechanism and prevention technology of HXTR / SGTR is the core key to the engineering implementation of lead-cooled fast reactors. The core difficulty in HXTR / SGTR accident research lies in the accurate measurement of two-phase flow parameters of liquid lead-based alloys. Key two-phase flow parameters such as the transport behavior of non-condensable gases, void fraction, interfacial area concentration, and interphase velocity are the core data support for analyzing the accident evolution law, assessing the core safety status, and formulating prevention and control strategies. Due to the special physicochemical properties of liquid lead-based alloys, such as opacity, high density, strong corrosion, and high-temperature oxidation, conventional optical measurement methods are not applicable, and parameter detection can only be carried out by relying on non-optical measurement techniques. The conductivity probe method is the most mature and widely used non-optical detection method in the field of gas-liquid two-phase flow measurement. It has been applied on a large scale in the ambient temperature water medium scenario. Its detection principle of judging the flow state and extracting the characteristic parameters of two-phase flow based on the difference in conductivity between gas and liquid is mature and reliable. It is the preferred technical solution for the detection of parameters of liquid lead-based alloy two-phase flow at this stage.Current conductivity probe detection technology can effectively adapt to conventional water environments with normal temperature, normal pressure, and no strong corrosion. It can accurately collect two-phase flow parameters such as cavitation fraction, bubble crossing frequency, and interfacial area concentration, basically meeting the testing requirements of two-phase flow under normal operating conditions. However, when applied to the extreme conditions of lead-cooled fast reactors, such as high temperature, strong corrosion, and high-density scouring, there are still inherent technical defects that are difficult to avoid. It cannot meet the long-term, stable, and accurate testing requirements of lead-based alloy two-phase flow, which seriously restricts the research on the accident mechanism of HXTR / SGTR and the engineering process of lead-cooled fast reactors. Specifically, the existing conductivity probe application technology has three major technical shortcomings: First, the performance degradation of probes under extreme conditions is gradual and covert. The high-temperature oxidation and strong corrosion of liquid lead-based alloys will gradually cause distortion of probe electrode morphology and attenuation of conductivity. At the same time, it will cause aging and deterioration of the insulating coating and degradation of insulation performance, resulting in decreased probe sensitivity, distortion of detection signal, and reduction of signal-to-noise ratio. Moreover, this type of performance degradation does not have instantaneous failure characteristics and shows a gradual evolution trend. Existing technologies lack targeted online monitoring and real-time evaluation methods, making it impossible to determine the probe measurement accuracy decay status in real time, identify signal failure thresholds, and distinguish between the effective and failed ranges of experimental data. This easily leads to systematic biases in experimental data, resulting in distorted conclusions in HXTR / SGTR accident mechanism research. Secondly, existing probe performance evaluation systems rely solely on short-term initial performance tests, failing to characterize the long-term performance evolution and lifetime characteristics of different probe materials, packaging structures, and fabrication processes under extreme lead-based alloy conditions. They also struggle to quantitatively compare the advantages and disadvantages of different probe schemes, and cannot provide reliable data support for probe structure optimization, process iteration, and material selection, thus hindering the development and iteration of dedicated probes adapted to lead-based alloy conditions. Finally, existing technologies lack the ability to predict the remaining probe lifetime, cannot accurately define the effective working time and failure stage of the probe, and cannot avoid the problem of invalid data acquisition in the probe's semi-failure state. This fundamentally fails to guarantee the reliability and validity of two-phase flow test data for lead-based alloys, making it difficult to support the accurate assessment and safety control technology development for lead-cooled fast reactor HXTR / SGTR accidents.

[0003] In summary, existing conductivity probe detection technologies are only suitable for conventional and mild operating conditions. They have not established an online performance monitoring, condition assessment, and lifetime prediction system for the extreme service environment of liquid lead-based alloys. They cannot solve the core problems of probe performance degradation and unreliable test data during long-term service. They are insufficient to meet the core needs of accident mechanism research, core safety analysis, and engineering applications of lead-cooled fast reactors HXTR / SGTR. There is an urgent need to develop online conductivity probe performance monitoring and lifetime prediction technologies that are suitable for high-temperature and highly corrosive lead-based alloy operating conditions.

[0004] Therefore, existing technologies still need to be improved and enhanced. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a method, system, terminal and medium for online monitoring of conductivity probe performance and prediction of remaining service life, addressing the above-mentioned deficiencies of the prior art. This invention aims to solve the problem that the existing technology lacks effective means for online performance monitoring, quantitative degradation assessment and remaining service life prediction of conductivity probes under high temperature and strong corrosive liquid lead-based alloy conditions, which makes it impossible to guarantee the reliability of two-phase flow test data of lead-cooled fast reactors and difficult to support reactor safety research and engineering applications.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides a method for online monitoring of conductivity probe performance and prediction of remaining service life, wherein the method includes: Obtain the reference range of voltage transient response slope and gas phase voltage amplitude of the conductivity probe under normal conditions; The collected voltage signal time series data is processed by a sliding time window to obtain the average slope of the voltage transient response and the average amplitude of the gas phase voltage for all bubble events within the current sliding time window. The health index is obtained based on the voltage transient response slope reference range, the gas phase voltage amplitude reference range, the average voltage transient response slope, and the average gas phase voltage amplitude. Based on the health index, the health index time series is formed by accumulating and arranging the window numbers in sequence. The health index time series is predicted by a time series regression prediction model, and failure time detection and remaining service life calculation are performed to obtain the remaining service life of the conductivity probe.

[0007] In one implementation, acquiring the voltage transient response slope reference range and the gas phase voltage amplitude reference range of the conductivity probe under normal conditions includes: Based on the fabricated conductivity probe, continuous measurements of bubble events were performed for a preset time in a water environment at room temperature and pressure, resulting in a dataset of voltage transient response slope and gas phase voltage amplitude of the conductivity probe in a water environment at room temperature and pressure. The reference range of voltage transient response slope under normal conditions is obtained by using the preset quantile of the voltage transient response slope dataset as the lower limit and the maximum value of the voltage transient response slope dataset as the upper limit; By using the preset quantile of the gas phase voltage amplitude dataset as the lower limit and the maximum value of the gas phase voltage amplitude dataset as the upper limit, the reference range of the gas phase voltage amplitude of the conductivity probe under normal conditions is obtained.

[0008] In one implementation, the step of processing the acquired voltage signal time series data through a sliding time window to obtain the average slope of the voltage transient response and the average amplitude of the gas phase voltage for all bubble events within the current sliding time window includes: The collected voltage signal time series data is processed by a sliding time window to obtain the voltage transient response slope and gas phase voltage amplitude of all bubble events within the current sliding time window; The average voltage transient response slope is obtained by arithmetically averaging the voltage transient response slope based on the number of bubble events, and the average gas phase voltage amplitude is obtained by arithmetically averaging the gas phase voltage amplitude based on the number of bubble events.

[0009] In one implementation, obtaining the health index based on the voltage transient response slope reference range, the gas phase voltage amplitude reference range, the average voltage transient response slope, and the average gas phase voltage amplitude includes: The average voltage transient response slope is compared with the voltage transient response slope reference range to obtain the voltage transient response slope achievement rate; The average value of the gas phase voltage amplitude is compared with the reference range of the gas phase voltage amplitude to obtain the gas phase voltage amplitude achievement rate; The health index is obtained by weighting and fusing the voltage transient response slope achievement rate and the gas phase voltage amplitude achievement rate.

[0010] In one implementation, comparing the average voltage transient response slope with a reference range of voltage transient response slope to obtain the voltage transient response slope achievement rate includes: If the average value of the voltage transient response slope is greater than or equal to the lower limit of the voltage transient response slope reference range, then the voltage transient response slope achievement rate is 1. If the average voltage transient response slope is less than the lower limit of the voltage transient response slope reference range, then the voltage transient response slope achievement rate is the ratio of the average voltage transient response slope to the lower limit of the voltage transient response slope reference range.

[0011] In one implementation, comparing the average value of the gas phase voltage amplitude with the reference range of the gas phase voltage amplitude to obtain the gas phase voltage amplitude achievement rate includes: If the average value of the gas phase voltage amplitude is greater than or equal to the lower limit of the reference range of the gas phase voltage amplitude, then the gas phase voltage amplitude achievement rate is 1. If the average value of the gas phase voltage amplitude is less than the lower limit of the reference range of the gas phase voltage amplitude, then the gas phase voltage amplitude achievement rate is the ratio of the average value of the gas phase voltage amplitude to the lower limit of the reference range of the gas phase voltage amplitude.

[0012] In one implementation, the step of predicting the health index time series using a time-series regression prediction model and performing failure time detection and remaining service life calculation to obtain the remaining service life of the conductivity probe includes: When the cumulative sliding time window number is less than the preset value, the health index time series is predicted by the Gaussian process regression prediction model, and failure time detection and remaining service life calculation are performed to obtain the remaining service life of the conductivity probe. When the cumulative sliding time window number is greater than or equal to the preset value, the health index time series is predicted by a bidirectional long short-term memory autoregressive prediction model, and failure time detection and remaining service life calculation are performed to obtain the remaining service life of the conductivity probe.

[0013] Secondly, embodiments of the present invention also provide an online monitoring system for the performance of a conductivity probe and a prediction system for its remaining service life, wherein the system includes: The reference range acquisition module is used to acquire the reference range of voltage transient response slope and gas phase voltage amplitude of the conductivity probe under normal conditions. The voltage transient response slope average and gas phase voltage amplitude average acquisition module is used to process the collected voltage signal time series data through a sliding time window to obtain the voltage transient response slope average and gas phase voltage amplitude average of all bubble events within the current sliding time window. The health index acquisition module is used to obtain the health index based on the voltage transient response slope reference range, the gas phase voltage amplitude reference range, the average voltage transient response slope, and the average gas phase voltage amplitude. The remaining service life acquisition module is used to accumulate and arrange the health index time series according to the window number based on the health index, predict the health index time series through the time series regression prediction model, and perform failure time detection and remaining service life calculation to obtain the remaining service life of the conductivity probe.

[0014] Thirdly, embodiments of the present invention also provide a terminal, wherein the terminal includes a memory, a processor, and a conductivity probe performance online monitoring and remaining service life prediction program stored in the memory and executable on the processor. When the processor executes the conductivity probe performance online monitoring and remaining service life prediction program, it implements the steps of the conductivity probe performance online monitoring and remaining service life prediction method described in any of the above schemes.

[0015] Fourthly, embodiments of the present invention also provide a computer-readable storage medium, wherein the computer-readable storage medium stores a program for online monitoring of conductivity probe performance and prediction of remaining service life, and when the program for online monitoring of conductivity probe performance and prediction of remaining service life is executed by a processor, it implements the steps of the method for online monitoring of conductivity probe performance and prediction of remaining service life as described in any of the above schemes.

[0016] Beneficial Effects: This invention provides a method for online monitoring of conductivity probe performance and prediction of remaining service life. Compared with existing technologies, this invention first obtains the reference range of voltage transient response slope and gas phase voltage amplitude of the conductivity probe under normal conditions, establishing a standardized and accurate reference for subsequent probe state discrimination and performance quantification evaluation. This effectively avoids the problems of traditional detection methods, such as lack of unified evaluation criteria, subjective state judgment, and large errors. It can adapt to the basic performance evaluation standards of probes under different operating conditions, ensuring the accuracy of the reference for subsequent state monitoring and lifespan prediction. Next, the collected voltage signal time series data is processed by a sliding time window to obtain the average voltage transient response slope and average gas phase voltage amplitude of all bubble events within the current sliding time window. This achieves noise reduction and refined decomposition of the original monitoring data, eliminating the drawback of ignoring local operating condition fluctuations in overall data statistics. It can accurately capture subtle voltage parameter changes caused by bubble interference and performance degradation during probe operation, making the collected state data more consistent with the actual situation. The probe's real-time actual working status is monitored, improving the effectiveness and relevance of data representation. Then, based on the voltage transient response slope reference range, the gas phase voltage amplitude reference range, the average voltage transient response slope, and the average gas phase voltage amplitude, a health index is obtained. This transforms the abstract probe aging and performance degradation status into quantifiable numerical indicators, enabling digital, visual, and precise assessment of the probe's operational health status. This overcomes the limitations of traditional methods that rely on manual experience or single parameters to determine equipment status, and comprehensively reflects the probe's current working performance and wear level. Finally, based on the health index, a health index time series is formed by accumulating and arranging the data sequentially according to window numbers. The health index time series is then predicted using a time-series regression prediction model, and failure time detection and remaining service life calculation are performed to obtain the remaining service life of the conductivity probe. This achieves intelligent processing throughout the entire process, from real-time status monitoring to trend prediction and life prediction, eliminating the passive mode of traditional periodic maintenance and post-fault repair, and allowing for early prediction of probe failure risks. Overall, this invention constructs a complete technical system from benchmark calibration, real-time data refinement, health status quantitative assessment to time-series lifetime prediction. Through multi-dimensional parameter fusion and time-series data analysis, it improves the accuracy and stability of conductivity probe status monitoring and remaining lifetime prediction, effectively avoiding detection errors caused by operating condition fluctuations and environmental interference. It can monitor the probe's entire life cycle operating status in real time and dynamically, accurately predict equipment failure points and remaining usable time, and provide reliable data support for preventive maintenance, on-demand replacement, and equipment operation and maintenance scheduling. It avoids problems such as monitoring data distortion and production operation hazards caused by premature probe failure and performance drift, and eliminates the waste of operation and maintenance costs caused by excessive replacement and blind maintenance. It improves the intelligence level of conductivity probe operating condition monitoring and the economy, stability, and safety of industrial production operation and maintenance. Attached Figure Description

[0017] Figure 1 A flowchart illustrating the specific implementation of the online monitoring and remaining service life prediction method for conductivity probes provided in this invention.

[0018] Figure 2 This is a flowchart illustrating the online performance monitoring and remaining service life prediction method for conductivity probes used in high-temperature, highly corrosive liquid lead-based alloy environments, as provided in the embodiments of the present invention.

[0019] Figure 3 This is a schematic diagram illustrating the performance evaluation index defined based on the physical mechanism of conductivity probe failure in the online monitoring and remaining service life prediction method for conductivity probes provided in this embodiment of the invention.

[0020] Figure 4 This is a schematic diagram of the sliding time window processing of the conductivity probe measurement voltage signal in the online monitoring and remaining service life prediction method for conductivity probe performance provided in this embodiment of the invention.

[0021] Figure 5 This is a schematic diagram of a calibration device for the characteristic benchmark value of conductivity probe performance evaluation index in a normal temperature and pressure water environment, which is part of the online monitoring and remaining service life prediction method for conductivity probe performance provided in this embodiment of the invention.

[0022] Figure 6 The schematic diagram shows the system principle of the online monitoring and life assessment application device for conductivity probe performance in a pool-type high-temperature liquid lead-based alloy environment, which is part of the online monitoring and remaining service life prediction method for conductivity probe performance in an embodiment of the present invention.

[0023] Figure 7 The histogram of gas phase voltage amplitude distribution in the online monitoring and remaining service life prediction method for conductivity probe performance provided in this embodiment of the invention.

[0024] Figure 8 The voltage transient response slope distribution histogram is shown in the online monitoring and remaining service life prediction method for conductivity probes provided in this embodiment of the invention.

[0025] Figure 9 The real-time performance evaluation and life prediction curves of a four-probe conductivity probe with a sliding time window of 451 are shown in the online performance monitoring and remaining service life prediction method for conductivity probes provided in this embodiment of the invention.

[0026] Figure 10 The real-time performance evaluation and life prediction curves of a four-probe conductivity probe with a sliding time window of 702 are shown in the online performance monitoring and remaining service life prediction method for conductivity probes provided in this embodiment of the invention.

[0027] Figure 11 This is a schematic diagram of the online performance monitoring and remaining service life prediction system for conductivity probes provided in this embodiment of the invention.

[0028] Figure 12 This is a block diagram illustrating the internal structure of a terminal provided in an embodiment of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0030] Against the backdrop of iterative upgrades in advanced nuclear energy technology and the large-scale development of clean energy, fourth-generation nuclear energy systems, with their advantages of high safety, high fuel utilization, and low radioactive waste production, have become the core direction for overcoming the safety shortcomings of traditional nuclear energy, improving uranium resource utilization efficiency, and promoting the green and low-carbon development of nuclear energy. They are also a key area for current nuclear energy research and engineering implementation. Lead-cooled fast reactors, as the reactor type with the greatest potential for large-scale commercialization among fourth-generation nuclear energy systems, rely on the excellent physicochemical properties of liquid lead-based alloy coolants—high boiling point, atmospheric pressure operation, and strong chemical inertness—to fundamentally avoid safety risks such as coolant boiling and violent chemical reactions. Simultaneously, they can achieve nuclear fuel breeding and long-life transmutation of high-level actinide nuclides, improving uranium resource utilization and reducing the hazards of nuclear waste. This represents the core development path for the next generation of advanced nuclear energy. Lead-cooled fast reactors abandon the intermediate isolation loop of traditional reactor types, adopting a pool-type integrated structure design. The main heat exchanger / steam generator is directly placed within the primary loop lead-based alloy molten pool, simplifying the system structure and reducing equipment operation and maintenance costs. However, this also makes the heat transfer tubes the sole barrier separating the high-temperature, low-pressure lead-based alloy in the primary loop from the high-pressure heat exchanger in the secondary loop. Under long-term corrosion, erosion, flow-induced vibration, and thermal stress coupling conditions, the heat transfer tubes are prone to damage and cracking, leading to a heat exchanger tube rupture (HXTR / Steam Generator Tube Rupture, SGTR) accident. After an accident, the secondary loop working fluid will be injected into the primary loop under pressure differential, forming a large amount of non-condensable gas after flash evaporation or phase change. Ultimately, a two-phase flow of liquid lead-based alloy and non-condensable gas will form within the primary loop lead-based alloy molten pool. These two-phase flow bubbles can enter the reactor core with the coolant, causing problems such as abnormal core reactivity and decreased local heat transfer performance, seriously threatening the operational safety of lead-cooled fast reactors. Therefore, research on the accident mechanism and prevention technology of HXTR / SGTR is the core key to the engineering implementation of lead-cooled fast reactors. The core difficulty in HXTR / SGTR accident research lies in the accurate measurement of two-phase flow parameters of liquid lead-based alloys. Key two-phase flow parameters such as the transport behavior of non-condensable gases, void fraction, interfacial area concentration, and interphase velocity are the core data support for analyzing the accident evolution law, assessing the core safety status, and formulating prevention and control strategies. Due to the special physicochemical properties of liquid lead-based alloys, such as opacity, high density, strong corrosion, and high-temperature oxidation, conventional optical measurement methods are not applicable, and parameter detection can only be carried out by relying on non-optical measurement techniques. The conductivity probe method is the most mature and widely used non-optical detection method in the field of gas-liquid two-phase flow measurement. It has been applied on a large scale in the ambient temperature water medium scenario. Its detection principle of judging the flow state and extracting the characteristic parameters of two-phase flow based on the difference in conductivity between gas and liquid is mature and reliable. It is the preferred technical solution for the detection of parameters of liquid lead-based alloy two-phase flow at this stage.Current conductivity probe detection technology can effectively adapt to conventional water environments with normal temperature, normal pressure, and no strong corrosion. It can accurately collect two-phase flow parameters such as cavitation fraction, bubble crossing frequency, and interfacial area concentration, basically meeting the testing requirements of two-phase flow under normal operating conditions. However, when applied to the extreme conditions of lead-cooled fast reactors, such as high temperature, strong corrosion, and high-density scouring, there are still inherent technical defects that are difficult to avoid. It cannot meet the long-term, stable, and accurate testing requirements of lead-based alloy two-phase flow, which seriously restricts the research on the accident mechanism of HXTR / SGTR and the engineering process of lead-cooled fast reactors. Specifically, the existing conductivity probe application technology has three major technical shortcomings: First, the performance degradation of probes under extreme conditions is gradual and covert. The high-temperature oxidation and strong corrosion of liquid lead-based alloys will gradually cause distortion of probe electrode morphology and attenuation of conductivity. At the same time, it will cause aging and deterioration of the insulating coating and degradation of insulation performance, resulting in decreased probe sensitivity, distortion of detection signal, and reduction of signal-to-noise ratio. Moreover, this type of performance degradation does not have instantaneous failure characteristics and shows a gradual evolution trend. Existing technologies lack targeted online monitoring and real-time evaluation methods, making it impossible to determine the probe measurement accuracy decay status in real time, identify signal failure thresholds, and distinguish between the effective and failed ranges of experimental data. This easily leads to systematic biases in experimental data, resulting in distorted conclusions in HXTR / SGTR accident mechanism research. Secondly, existing probe performance evaluation systems rely solely on short-term initial performance tests, failing to characterize the long-term performance evolution and lifetime characteristics of different probe materials, packaging structures, and fabrication processes under extreme lead-based alloy conditions. They also struggle to quantitatively compare the advantages and disadvantages of different probe schemes, and cannot provide reliable data support for probe structure optimization, process iteration, and material selection, thus hindering the development and iteration of dedicated probes adapted to lead-based alloy conditions. Finally, existing technologies lack the ability to predict the remaining probe lifetime, cannot accurately define the effective working time and failure stage of the probe, and cannot avoid the problem of invalid data acquisition in the probe's semi-failure state. This fundamentally fails to guarantee the reliability and validity of two-phase flow test data for lead-based alloys, making it difficult to support the accurate assessment and safety control technology development for lead-cooled fast reactor HXTR / SGTR accidents.

[0031] In summary, existing conductivity probe detection technologies are only suitable for conventional and mild operating conditions. They have not established an online performance monitoring, condition assessment, and lifetime prediction system for the extreme service environment of liquid lead-based alloys. They cannot solve the core problems of probe performance degradation and unreliable test data during long-term service. They are insufficient to meet the core needs of accident mechanism research, core safety analysis, and engineering applications of lead-cooled fast reactors HXTR / SGTR. There is an urgent need to develop online conductivity probe performance monitoring and lifetime prediction technologies that are suitable for high-temperature and highly corrosive lead-based alloy operating conditions.

[0032] To address the aforementioned issues, this embodiment provides a method for online monitoring of conductivity probe performance and prediction of remaining service life. Specifically, this embodiment first obtains the reference range of the voltage transient response slope and the reference range of the gas phase voltage amplitude of the conductivity probe under normal conditions. This establishes a standardized and accurate reference for subsequent probe state discrimination and performance quantification evaluation, effectively avoiding the problems of traditional detection methods such as lack of a unified evaluation standard, subjective state judgment, and large errors. It can adapt to the basic performance evaluation standards of probes under different operating conditions, ensuring the accuracy of the reference for subsequent state monitoring and lifespan prediction. Next, the collected voltage signal time series data is processed using a sliding time window to obtain the average voltage transient response slope and the average gas phase voltage amplitude of all bubble events within the current sliding time window. This achieves noise reduction and refined decomposition of the original monitoring data, eliminating the drawback of ignoring local operating condition fluctuations in overall data statistics. It can accurately capture subtle voltage parameter changes caused by bubble interference and performance degradation during probe operation, making the collected state data more accurate and precise. By combining the real-time actual working status of the probe, the effectiveness and relevance of data representation are improved. Then, based on the voltage transient response slope benchmark range, the gas phase voltage amplitude benchmark range, the average voltage transient response slope, and the average gas phase voltage amplitude, a health index is obtained. This transforms the abstract probe aging and performance degradation status into a quantifiable numerical indicator, realizing the digital, visual, and accurate assessment of the probe's operational health status. This overcomes the limitations of traditional methods that rely on manual experience or single parameters to determine equipment status, and can comprehensively reflect the current working performance and wear level of the probe. Finally, based on the health index, a health index time series is formed by accumulating and arranging the window numbers sequentially. The health index time series is then predicted using a time series regression prediction model, and failure time detection and remaining service life calculation are performed to obtain the remaining service life of the conductivity probe. This achieves intelligent processing of the entire process from real-time status monitoring to trend prediction and life prediction, getting rid of the passive mode of traditional periodic maintenance and post-fault repair, and can predict the risk of probe failure in advance. Overall, this invention constructs a complete technical system from benchmark calibration, real-time data refinement, health status quantitative assessment to time-series lifetime prediction. Through multi-dimensional parameter fusion and time-series data analysis, it improves the accuracy and stability of conductivity probe status monitoring and remaining lifetime prediction, effectively avoiding detection errors caused by operating condition fluctuations and environmental interference. It can monitor the probe's entire life cycle operating status in real time and dynamically, accurately predict equipment failure points and remaining usable time, and provide reliable data support for preventive maintenance, on-demand replacement, and equipment operation and maintenance scheduling. It avoids problems such as monitoring data distortion and production operation hazards caused by premature probe failure and performance drift, and eliminates the waste of operation and maintenance costs caused by excessive replacement and blind maintenance. It improves the intelligence level of conductivity probe operating condition monitoring and the economy, stability, and safety of industrial production operation and maintenance.

[0033] The online monitoring and remaining lifespan prediction method for conductivity probes provided in this embodiment can be applied to smart terminals, such as... Figure 1 As shown, the specific steps include the following: Step S100: Obtain the reference range of voltage transient response slope and gas phase voltage amplitude of the conductivity probe under normal conditions.

[0034] In this embodiment, a reference range for the voltage transient response slope of the conductivity probe under normal operating conditions is obtained. The voltage transient response slope can intuitively reflect the voltage change rate of the conductivity probe at the moment of power-on, accurately reflecting the probe's response sensitivity and operational stability. Establishing this reference range can provide a core slope reference standard for subsequent probe fault diagnosis and performance degradation detection, effectively distinguishing between the probe's normal response and abnormal distortion states, and avoiding deviations in conductivity detection data caused by abnormal response rates. Furthermore, a reference range for the gas phase voltage amplitude under normal probe conditions is obtained. The gas phase voltage amplitude is a core voltage characteristic parameter of the probe under gas phase conditions, corresponding to the probe's basic electrical operating state under no-load and gas phase environments. This reference range can calibrate the acceptable range of the probe's basic electrical performance, and investigate voltage amplitude deviation problems caused by electrode aging, poor circuit contact, or insulation abnormalities. The entire parameter benchmark acquisition process establishes a complete standard parameter system for the normal operating conditions of conductivity probes from two core dimensions: dynamic transient response performance and static gas phase voltage performance. It covers both the performance indicators of the probe's dynamic detection response and the criteria for judging the basic electrical operating conditions. This provides accurate and comprehensive benchmark references for subsequent online probe monitoring, fault diagnosis, and performance calibration, improving the accuracy and reliability of conductivity probe operating condition detection, ensuring the accuracy and effectiveness of subsequent dielectric conductivity detection data, and avoiding equipment monitoring errors and misjudgments of operating conditions caused by probe performance abnormalities.

[0035] Specifically, step S100 includes the following steps: Step S101: Based on the completed conductivity probe, continuously measure the bubble event for a preset time in a normal temperature and pressure water environment to obtain the voltage transient response slope data set and the gas phase voltage amplitude data set of the conductivity probe in a normal temperature and pressure water environment. Step S102: Using the preset quantile of the voltage transient response slope dataset as the lower limit and the maximum value of the voltage transient response slope dataset as the upper limit, the reference range of the voltage transient response slope of the conductivity probe under normal conditions is obtained. Step S103: Using the preset quantile of the gas phase voltage amplitude dataset as the lower limit and the maximum value of the gas phase voltage amplitude dataset as the upper limit, the reference range of the gas phase voltage amplitude of the conductivity probe under normal conditions is obtained.

[0036] In one implementation, such as Figure 2 As shown, firstly, the performance evaluation index based on the physical mechanism of conductivity probe failure is defined. Liquid lead-based alloys have high temperature and strong corrosive properties. Their chemical corrosion of the probe tip will lead to the attenuation of electrode conductivity, thereby reducing the probe's ability to distinguish between gas and liquid phases, manifested as a decrease in probe sensitivity. Simultaneously, the aging and deterioration of the insulating coating material in a high-temperature corrosive environment will lead to the degradation of insulation performance, causing distortion of the measured voltage signal, a decrease in signal-to-noise ratio, and a reduction in measurement accuracy. Therefore, there are two main physical mechanisms for the performance degradation or even failure of conductivity probes in high-temperature, highly corrosive liquid lead-based alloy environments, and this invention defines two performance evaluation indices accordingly. These are described below: Firstly, corrosion of the probe tip leads to contamination and oxide layer thickening, resulting in a decrease in conductivity and thus affecting the probe's sensitivity to distinguish between gas and liquid phases. This mechanism will lead to a decrease in the transient rate of voltage change at the moment the bubble contacts the probe tip. Ideally, at the instant the bubble contacts the probe tip, the voltage of the probe measurement circuit will change instantaneously from near 0 (corresponding to the liquid phase) to near the DC drive voltage (corresponding to the gas phase). Figure 3 The upper part shows: The left side is a schematic diagram of the measurement circuit. The signal acquisition unit, DC drive voltage, and resistor constitute the measurement loop. The probe tip electrode is immersed in liquid lead-based alloy, but at this time, the tip is corroded and passivated, and the rate of voltage change from 0 to DC drive voltage slows down; The right side is a schematic diagram of the voltage-time waveform. In the liquid phase, the voltage is close to zero (liquid phase signal), and in the gas phase, the voltage is close to the DC drive voltage (gas phase signal). At this time, the time change of the voltage in the probe measurement loop from close to 0 (corresponding to the liquid phase) to close to the DC drive voltage (corresponding to the gas phase) is... τ. Therefore, the first performance evaluation index is defined as "voltage transient response slope (τ). The "voltage" refers to the average rate of change (in V / ms) of the voltage at the probe's contact with a bubble, from near 0 (corresponding to the liquid phase) to near constant DC driving voltage (corresponding to the gas phase). It quantifies the probe's dynamic response performance during the transition between the gas and liquid phases. This indicator directly reflects the conductivity response speed of the electrode surface. When the probe experiences surface contamination, oxide layer thickening, or increased contact impedance, the response slope will significantly decrease, making it a sensitive characteristic of probe dynamic performance degradation.

[0037] Secondly, the insulating coating of the needle body can be damaged by corrosion and erosion from the high-temperature liquid lead-based alloy, leading to a decrease in insulation performance and causing abnormalities such as distortion of the gas phase voltage signal and a reduction in the signal-to-noise ratio. In the ideal situation, as long as the probe tip is still encased in the bubble, the probe measurement circuit should be in a nearly open state, and the measured voltage value should always be close to a constant DC drive voltage (corresponding to the gas phase). For example... Figure 3The lower half shows: the left side is a schematic diagram of the measurement circuit. When bubble ② covers the needle tip, the measurement circuit cannot be completely disconnected due to the damage to the needle's insulating coating. The right side shows the voltage signal at this time, and the gas phase voltage value decreases. Therefore, the second performance evaluation index, "gas phase voltage amplitude ( The voltage rise (in volts) after the probe encounters a bubble indicates the maximum voltage increase reached by the probe upon contact with the bubble, measuring the remaining ability of the probe to distinguish between the gas and liquid phases. Under normal circumstances, the voltage difference in the probe measurement circuit is significant and the distinction is clear when the probe tip is covered by a bubble (representing the probe is measuring the gas phase) versus when the probe tip is completely immersed in liquid lead-based alloy (representing the probe is measuring the liquid phase). However, as the probe is used over time, the insulating layer of the probe body is corroded and eroded, its insulation properties change, and some insulation may fail. This can lead to the probe measurement circuit not being completely disconnected even when the probe tip is covered by a bubble, resulting in a measured voltage that is not close to the ideal constant DC drive voltage and may be significantly lower. Therefore, this indicator directly quantifies the degree of decline in the probe's ability to distinguish between gas and liquid phases and is a core indicator reflecting the degradation of the probe's static performance.

[0038] Voltage transient response slope and gas phase voltage amplitude are two core indicators for evaluating the performance of conductivity probes in detecting bubble events. Voltage transient response slope characterizes the dynamic performance of the probe in detecting bubble events, while gas phase voltage amplitude characterizes the static performance. When conducting online monitoring of conductivity probes used in high-temperature, highly corrosive liquid lead-based alloy environments, the primary prerequisite is to accurately calibrate the characteristic benchmark values ​​of the probe's performance under normal operating conditions. This provides a reliable basis for determining the effectiveness of the probe's performance under subsequent online operating conditions. Since the application technology of conductivity probes in ambient temperature and pressure water environments is mature, and existing published literature confirms the excellent service stability of the probes under these conditions, an ambient temperature and pressure water environment is selected as the calibration environment for the characteristic benchmark values. Figure 2As shown, the performance evaluation index characteristic benchmark values ​​of the conductivity probe in a normal temperature and pressure water environment are calibrated. The corresponding calibration device is made of transparent plexiglass, using clean water as the calibration medium and operating under normal temperature and pressure conditions throughout the process. The specific calibration procedure is as follows: After the conductivity probe is fabricated, it is placed in a normal temperature and pressure water environment to conduct continuous measurements of bubble events for a preset time, and noise reduction preprocessing is performed to obtain the voltage transient response slope dataset and the gas phase voltage amplitude dataset of the conductivity probe in a normal temperature and pressure water environment. The preset time required for calibration is "sufficiently long time", with the sliding time window used for online monitoring as the reference benchmark. If the sliding time window for online monitoring is set to 30 seconds, then the continuous measurement time for this calibration is set to at least 30 minutes to ensure the comprehensiveness and representativeness of the calibration data. At the same time, the probe performance characteristic benchmark values ​​obtained through the above-mentioned long-term continuous calibration measurement are not single fixed values, but correspond to reasonable value ranges, forming voltage transient response slope datasets { } and the gas phase voltage amplitude dataset { Long-term continuous data acquisition and calibration ensures that the two datasets can completely cover all performance value ranges of the conductivity probe under normal operating conditions. Within the corresponding dataset range, higher voltage transient response slope values ​​and higher gas phase voltage amplitude values ​​indicate that the conductivity probe's performance is closer to its ideal operating state. Considering that the harsh service environment of high-temperature, highly corrosive liquid lead-based alloys can cause certain reversible changes in the physical properties of the conductivity probe, resulting in inherent performance degradation compared to calibration in a normal temperature and pressure water environment, a reasonable engineering tolerance must be reserved when determining the normal performance benchmark range of the probe under high-temperature, highly corrosive lead-based alloy conditions. This tolerance is based on the voltage transient response slope dataset { } and gas phase voltage amplitude dataset { This invention extracts feature values ​​for conductivity probe performance evaluation indicators. It abandons the traditional conservative approach of using the minimum value as the lower limit of the benchmark, and instead uses a preset quantile of the dataset (e.g., the 40th quantile) as the lower limit and the maximum value of the dataset as the upper limit. , in , These correspond to the sets of voltage transient response slopes { }, Set of gas phase voltage amplitudes { The 40th percentile of} , These represent the lower and upper limits of the reference values ​​for the voltage transient response slope of the conductivity probe under normal conditions in a high-temperature, highly corrosive lead-based alloy environment. , These represent the lower and upper limits of the gas-phase voltage amplitude reference value under normal conditions for a conductivity probe in a high-temperature, highly corrosive lead-based alloy environment. This method effectively avoids false alarms caused by minor fluctuations in the probe's physical characteristics due to high-temperature environments, eliminating extreme abnormal data while maintaining reasonable engineering tolerance for performance fluctuations under normal operating conditions. Based on the above value rules, the final performance reference range of the probe under high-temperature conditions can be determined, with the voltage transient response slope reference range being: The reference range for gas phase voltage amplitude is This refers to the confidence interval of the output performance evaluation index, providing a basis for subsequent probe status assessment. During subsequent online monitoring in a high-temperature, highly corrosive liquid lead-based alloy environment, the voltage transient response slope and gas phase voltage amplitude of the conductivity probe are continuously acquired through a real-time sliding time window. If both performance index values ​​obtained in real-time are within the corresponding benchmark range, it is determined that the conductivity probe's performance has not abnormally deteriorated and remains in a fully usable normal working state, enabling stable online monitoring of bubble events.

[0039] This invention obtains the reference range of voltage transient response slope and gas phase voltage amplitude of the conductivity probe under normal conditions, establishing a standardized and accurate reference for subsequent probe status discrimination and performance quantification evaluation. It effectively avoids the problems of traditional detection methods, such as lack of unified evaluation criteria, subjective status judgment, and large errors. It can adapt to the basic performance evaluation standards of probes under different working conditions, ensuring the accuracy of the benchmark for subsequent status monitoring and lifespan prediction.

[0040] Step S200: Process the collected voltage signal time series data by a sliding time window to obtain the average slope of the voltage transient response and the average amplitude of the gas phase voltage for all bubble events within the current sliding time window.

[0041] In this embodiment, the acquired bubble voltage signal time series data is first processed by sliding time window segmentation. The continuous time series voltage data is segmented according to the set window duration and sliding step size, breaking down the extremely long continuous time series data into multiple interconnected and sequentially continuous local data windows. This segmentation method avoids the problems of large overall time series data span and low instantaneous feature recognition, accurately locking the local bubble voltage response data corresponding to different moments, ensuring that subsequent feature analysis has time-specificity and real-time performance. After completing the sliding window data segmentation, the voltage transient response data corresponding to all bubble events within the current window are extracted one by one. The slope parameter of the single bubble voltage transient response is calculated, and then the average value of the slope data of all bubbles within the window is calculated. This operation can effectively offset the numerical fluctuations caused by accidental noise and small detection errors during single bubble detection, objectively reflecting the overall speed characteristics of the bubble voltage transient change within the current time series interval, and accurately characterizing the dynamic change law of voltage during bubble breakdown and disturbance. Simultaneously, the gas phase voltage amplitude data of each bubble event within the current sliding time window are extracted, and the average gas phase voltage amplitude of the window is calculated by mean value. This integrates the voltage amplitude characteristics of all bubbles within the window, weakens the discrete differences in the amplitude of a single bubble, and stably presents the overall amplitude level of the gas phase discharge voltage in the current period. The overall processing method, which combines segmented sliding time window with dual-feature mean statistics, not only preserves the continuity of the voltage signal's temporal evolution but also achieves refined and stable extraction of the transient response and amplitude characteristics of bubble voltage. It effectively filters out invalid interference information in the time series data, accurately quantifies the electrical characteristics of bubble discharge in different periods, and provides stable, reliable, and time-representative core feature data for subsequent bubble state identification, discharge characteristic analysis, and operating condition judgment.

[0042] Specifically, step S200 includes the following steps: Step S201: Process the collected voltage signal time series data by sliding time window to obtain the voltage transient response slope and gas phase voltage amplitude of all bubble events within the current sliding time window; Step S202: Calculate the arithmetic mean of the voltage transient response slope based on the number of bubble events to obtain the average voltage transient response slope, and calculate the arithmetic mean of the gas phase voltage amplitude based on the number of bubble events to obtain the average gas phase voltage amplitude.

[0043] In one implementation, the voltage transient response slope and the gas phase voltage amplitude are two performance evaluation indicators that reflect the dynamic and static performance of the conductivity probe for each bubble event it detects, respectively. If performance evaluation is based solely on a single or a few bubble measurement events, it is impossible to accurately quantify the overall performance of the conductivity probe. Therefore, sufficient bubble event data needs to be collected and statistically analyzed to scientifically evaluate the probe's performance. Based on this principle, this invention employs a fixed-size sliding time window processing method for the real-time voltage signal time series data acquired by the conductivity probe system. Considering the high sampling frequency of the conductivity probe system (up to 10kHz), this scheme sets the sliding time window size to half a minute. This window duration can encompass hundreds of effective bubble events under typical operating conditions, providing sufficient data support for probe performance evaluation. It should be noted that 300,000 data points can be collected within a half-minute time window at a 10kHz sampling frequency. The number of data points is not directly equivalent to the number of bubble events; in this scheme, a bubble event specifically refers to the number of bubbles effectively detected and identified by the probe within the corresponding time window. The above-mentioned sliding time window processing can be achieved through... Figure 4 Intuitive display, Figure 4 This diagram illustrates the sliding time window processing of voltage signal time series data measured by a conductivity probe. Based on a voltage-time waveform, it presents the position and recursive relationship of three consecutive sliding time windows. The horizontal axis of the waveform represents time in seconds (s), and the vertical axis represents voltage in volts (V). In the diagram, a single blue pulse corresponds to one bubble event. The three sub-graphs sequentially show the progressive update process of the sliding windows from top to bottom. The range of each window is marked with a red rectangle, and each sliding time window independently completes the index calculation. At the same time, a certain proportion of overlap is set between two adjacent consecutive sliding time windows, so that the sliding time windows can be iteratively updated in real time during online monitoring. The probe performance evaluation index calculated based on the time windows can also be updated synchronously in real time. This enables real-time online tracking of the performance trend of conductivity probes with the service time under high temperature and highly corrosive liquid lead-based alloy conditions, providing a reliable data foundation for real-time evaluation of probe performance and accurate prediction of service life. After completing the sliding time window truncation processing of the voltage signal time series data, the voltage transient response slope and gas phase voltage amplitude corresponding to all valid bubble events within the current sliding time window can be extracted. Then, the random error of single bubble measurement is eliminated by a statistical averaging algorithm to obtain the overall performance characterization parameters within the window. Specifically, the calculation method is to calculate the average voltage transient response slope of the current window by arithmetically averaging the voltage transient response slope of all bubbles within the window based on the number of bubble events, and simultaneously calculate the average gas phase voltage amplitude of the current window by arithmetically averaging the gas phase voltage amplitude of all bubbles within the window based on the number of bubble events. The formula for calculating the average voltage transient response slope is as follows: The formula for calculating the average amplitude of the gas phase voltage is: In the two formulas Each represents the total number of bubble events effectively measured by the probe within the current sliding time window. Represents the number of times within the current sliding time window. The slope of the voltage transient response corresponding to each bubble. Represents the number of times within the current sliding time window. The gas phase voltage amplitude corresponding to each bubble is finally calculated. This represents the average slope of the voltage transient response for all bubble events within the current sliding time window. This represents the average vapor phase voltage amplitude of all bubble events within the current sliding time window. This scheme, through high-frequency sampling combined with a statistical evaluation method using a fixed-duration overlapping sliding time window, eliminates the randomness and bias of traditional single-point, small-scale bubble event evaluations. Relying on the statistical averaging of sufficient effective bubble samples, it effectively avoids the interference of single-measurement errors on performance evaluation results, improving the stability and accuracy of dynamic and static performance evaluation results for conductivity probes. Simultaneously, leveraging the real-time iterative update characteristic of the sliding window, it can continuously track the dynamic decay changes of probe performance under harsh operating conditions, achieving online, real-time, and continuous quantitative evaluation of probe performance. This solves the technical pain point of traditional static sampling methods being unable to capture the gradual trend of probe performance changes, accurately reflecting the performance evolution pattern of probes during long-term service. It provides accurate, continuous, and reliable data support for probe fault prediction, precise lifespan estimation, and equipment maintenance strategy optimization, improving the timeliness and comprehensiveness of bubble detection probe performance evaluation under lead-based alloy conditions.

[0044] Step S300: Based on the voltage transient response slope reference range, the gas phase voltage amplitude reference range, the average voltage transient response slope, and the average gas phase voltage amplitude, the health index is obtained.

[0045] In this embodiment, a standardized evaluation reference system is established based on the voltage transient response slope benchmark range and the gas phase voltage amplitude benchmark range. This provides a unified quantitative basis for defining the quality and deviation of subsequent equipment status data, avoiding the problems of standard confusion and judgment deviation caused by the lack of reference evaluation, and ensuring the standardization and uniformity of subsequent data analysis. Furthermore, by combining the calculated average voltage transient response slope and the average gas phase voltage amplitude, the actual voltage transient change characteristics and gas phase voltage fluctuation characteristics of the equipment are quantified and grounded in reality. This accurately restores the electrical state characteristics of the equipment in real time, compensating for the randomness and one-sidedness of single instantaneous data, and ensuring the authenticity and representativeness of the data representation. Finally, the equipment health index is calculated by combining the standardized benchmark range and the actual operating average data. This index integrates the status information of the two core dimensions of voltage transient response and gas phase voltage operation, comprehensively and three-dimensionally reflecting the overall operating conditions and health status of the equipment. It enables accurate identification of potential equipment operation hazards and performance degradation, providing an intuitive and reliable quantitative basis for equipment status assessment, fault prediction, and maintenance decisions, effectively improving the accuracy and comprehensiveness of equipment status evaluation.

[0046] Specifically, step S300 includes the following steps: Step S301: Compare the average voltage transient response slope with the voltage transient response slope reference range to obtain the voltage transient response slope achievement rate; Step S302: Compare the average value of the gas phase voltage amplitude with the reference range of the gas phase voltage amplitude to obtain the gas phase voltage amplitude achievement rate; Step S303: Weighted fusion of voltage transient response slope achievement rate and gas phase voltage amplitude achievement rate to obtain the health index.

[0047] In one implementation, the degradation of both voltage transient response slope and gas phase voltage amplitude stems from tip corrosion and insulation coating damage that occur during long-term probe service. Furthermore, these two deteriorating physical mechanisms occur simultaneously and jointly affect probe performance during service. To achieve a precise and comprehensive quantitative assessment of the overall performance status of conductivity probes, this invention integrates the two aforementioned individual performance parameters to construct a dimensionless comprehensive probe performance evaluation index, namely the Health Index (HI), for use in... Figure 2The diagram illustrates online monitoring of conductivity probes in a high-temperature liquid lead-based alloy environment. The online calculation and performance evaluation of this health index uses a sliding time window as the basic calculation unit. Each independent time window can complete a full health index solution. The specific implementation steps are as follows: First, load the voltage transient response slope reference range and the gas phase voltage amplitude reference range under normal operating conditions of the conductivity probe (i.e., read the feature value file and obtain the confidence interval of the performance evaluation index), providing a standard basis for subsequent performance comparison and compliance judgment; second, extract the average voltage transient response slope and the average gas phase voltage amplitude obtained from the real-time acquisition of conductivity probe signals and time window processing monitoring; subsequently, compare the two measured performance parameters (average voltage transient response slope and average gas phase voltage amplitude) obtained in the current time window with the corresponding reference parameter ranges (voltage transient response slope reference range and gas phase voltage amplitude reference range), and calculate the voltage transient response slope achievement rate and the gas phase voltage amplitude achievement rate accordingly. , , The specific judgment and calculation rules of the above formula are as follows: If the average voltage transient response slope of the current sliding time window is greater than or equal to the lower limit of the voltage transient response slope reference range, the dynamic performance index is judged to be completely normal, and the voltage transient response slope achievement rate is directly assigned to 1. If the average voltage transient response slope is less than the lower limit of the voltage transient response slope reference range, it indicates that the probe dynamic performance has deteriorated but has not completely failed. The corresponding transient response slope achievement rate is calculated dimensionlessly with the lower limit of the reference range as the reference standard. That is, the voltage transient response slope achievement rate is the ratio of the average voltage transient response slope to the lower limit of the voltage transient response slope reference range. The determination and calculation logic for the gas phase voltage amplitude achievement rate is completely consistent with that for the transient response slope achievement rate, allowing for simultaneous quantitative evaluation of probe static performance compliance. Specifically: if the average gas phase voltage amplitude is greater than or equal to the lower limit of the gas phase voltage amplitude reference range, the gas phase voltage amplitude achievement rate is 1; if the average gas phase voltage amplitude is less than the lower limit of the gas phase voltage amplitude reference range, the gas phase voltage amplitude achievement rate is the ratio of the average gas phase voltage amplitude to the lower limit of the gas phase voltage amplitude reference range. After calculating the achievement rates of the two indicators, the probe health index HI for the current time window is solved using a weighted fusion method. The specific calculation formula is as follows: ,in, The slope achievement rate of voltage transient response. The gas phase voltage amplitude achievement rate is represented by α and β, which are the weighting coefficients corresponding to the two performance indicators, satisfying the weighting constraint α + β = 1. In this embodiment, α = 0.4 and β = 0.6 are preferably set. The health index HI is a dimensionless parameter, ranging from 0 to 1. The closer the value is to 1, the closer the conductivity probe performance is to a normal and usable state; the lower the value, the more severe the degradation of probe performance. Simultaneously, based on the real-time calculated window health index, online grading evaluation of probe performance can be completed, such as... Figure 2 As shown, the real-time performance evaluation of the conductivity probe in a specific high-temperature liquid lead-based alloy environment is as follows: when When the probe is deemed completely healthy, with all performance characteristics within the normal baseline range, it can operate stably and reliably; when At this point, the probe is determined to be in a state of mild degradation, with some performance characteristics beginning to deviate from the baseline standard, and overall performance showing a slight decline; when When the probe is in a moderately degraded state, its performance characteristics deviate significantly from the reference range, and its detection accuracy and stability decrease markedly, close monitoring and continuous monitoring of the probe's operating status are required; when When the probe is determined to be in a severely degraded state, with its performance approaching the critical failure point, it cannot meet the requirements for accurate detection and must be replaced promptly. Overall, this online monitoring and evaluation method for conductivity probe performance based on a health index effectively integrates both dynamic and static core performance evaluation dimensions. It overcomes the shortcomings of traditional single-index evaluation, which cannot comprehensively consider multi-dimensional performance degradation and suffers from insufficient accuracy. It can accurately capture the progressive performance degradation caused by probe tip corrosion and insulation coating damage under special conditions of high temperature and strong corrosion. Through time-sequential, quantitative, and graded evaluation methods, it achieves real-time and accurate perception of probe operating status, refined judgment of degradation degree, and effective prediction of remaining lifespan. It can provide reliable data support for operation and maintenance decisions, early maintenance, and failure warnings for conductivity probes under liquid lead-based alloy conditions, effectively avoiding problems such as data distortion and operational monitoring errors caused by probe failure. This improves the stability of the probe during service and the reliability of the entire monitoring system, adapting to the long-term online monitoring needs under harsh industrial conditions.

[0048] Step S400: Based on the health index, the health index time series is formed by accumulating and arranging the window numbers in sequence. The health index time series is predicted by the time series regression prediction model, and failure time detection and remaining service life calculation are performed to obtain the remaining service life of the conductivity probe.

[0049] In this embodiment, firstly, the health index of the conductivity probe is sequentially accumulated and arranged according to the window number to construct a standardized health index time series. This operation can orderly connect discrete health data of the probe in different monitoring periods according to the time dimension, completely reconstructing the trajectory of health degradation during the long-term operation of the probe, avoiding the problems of fragmented discrete data and lack of temporal correlation, and providing a continuous, regular, and analyzable basic data carrier for subsequent trend prediction. Subsequently, trend prediction is carried out on the constructed health index time series based on the time series regression prediction model. Taking advantage of the time series regression model's ability to discover potential change patterns in time series data and fit the dynamic evolution trend of parameters, the decay trend of the conductivity probe health index in future periods is accurately deduced, accurately capturing changes in operating status that are difficult to identify through instantaneous data, such as slow degradation of probe performance and hidden losses. On this basis, failure time detection and remaining service life calculation are carried out. By benchmarking against the failure judgment criteria of probe equipment and combining the predicted future change curve of the health index, the time node when the probe performance decays to the failure threshold is accurately located. At the same time, the usable time of the conductivity probe from the current operating state to complete failure is accurately calculated, quantifying the remaining service life of the conductivity probe. The overall process, through a closed-loop logic of time-series data normalization, intelligent trend prediction, and accurate lifespan calculation, achieves dynamic tracking of the health status of conductivity probes, prediction of degradation trends, and accurate lifespan assessment. It breaks away from the passive mode of traditional equipment periodic maintenance and post-failure maintenance, and can predict equipment failure risks in advance, providing accurate data support for preventive maintenance of probes, advance preparation of spare parts, and optimization of operation and maintenance plans.

[0050] Specifically, step S400 includes the following steps: Step S401: When the cumulative sliding time window number is less than the preset value, the health index time series is predicted by the Gaussian process regression prediction model, and failure time detection and remaining service life calculation are performed to obtain the remaining service life of the conductivity probe. Step S402: When the cumulative sliding time window number is greater than or equal to the preset value, the health index time series is predicted by the bidirectional long short-term memory autoregressive prediction model, and failure time detection and remaining service life calculation are performed to obtain the remaining service life of the conductivity probe.

[0051] In one implementation, the conductivity probe health index (HI) serves as the core indicator for quantifying probe degradation. This index exhibits unique time-series accumulation characteristics in the online monitoring scenario of high-temperature liquid lead-based alloys. Specifically, the data volume monotonically increases with probe service time, new data shows a strong time-series correlation with historical data, and probe performance degradation is driven by multiple coupled physical mechanisms, exhibiting a non-steady-state degradation pattern without fixed linear or exponential decay laws. Furthermore, the monitoring process is characterized by scarce small-sample data in the early stages and complex degradation patterns in the later stages with large samples. This makes traditional single prediction models unsuitable for the prediction needs of the entire data lifecycle. High-capacity models are prone to overfitting in small-sample scenarios, while low-capacity models are prone to underfitting in large-sample scenarios, making it difficult to simultaneously satisfy the dual constraints of robustness in small samples and expressiveness in large samples. Based on these technical pain points, such as... Figure 2 As shown, this paper proposes an adaptive dual-model switching strategy driven by the amount of data, with the number of cumulative sliding time windows as the preset value as the switching threshold. This strategy is used to predict the lifetime of conductivity probes in high-temperature liquid lead-based alloy environments. The Gaussian Process Regression (GPR) prediction model for small sample scenarios and the Bidirectional Long Short-Term Memory Autoregressive (LSTM-AR) prediction model for large sample scenarios are matched respectively. The switching threshold (preset value) is determined by three core constraints: First, the lower limit of training convergence of the LSTM-AR model. When the amount of data is less than the preset value (500), the number of training samples after the sliding window construction is small (less than 480), which cannot support the model to stably learn the degradation mode. After reaching the preset value (500), Huber loss, weight decay and early stopping mechanism can be used to achieve effective convergence. Second, the upper limit of effective application of the GPR model. When the amount of data exceeds 800, the kernel matrix dimension is reduced. The surge in data points can lead to the accumulation of numerical errors, resulting in non-physical fluctuations and decreased stability in the prediction curve. Thirdly, there are real-time response time requirements for the online monitoring system. GPR can achieve second-level inference within 800 data points, meeting the real-time requirements. However, the inference time exceeds the limit after exceeding the limit. Considering the three constraints, the feasible range for model switching is [500, 800]. The lower bound of the range, 500, is selected as the switching threshold (i.e., the preset value). In the small sample stage, GPR can be used to ensure the robustness of prediction, while in the large sample stage, LSTM-AR can be used to fully explore complex degradation patterns and improve prediction accuracy.

[0052] In this scheme, the Remaining Useful Life (RUL) is defined as the remaining service time from the current moment when the probe's health index decays to the failure threshold. In specific implementation, firstly, based on the sliding time window processing mechanism of the conductivity probe online monitoring system, the health index (HI data points) is calculated window by window from the continuously acquired raw voltage signal. These are then accumulated and arranged sequentially according to the window number to form a health index time series, i.e., the HI time series. ,in, Let be the index of the i-th sliding time window. The corresponding health index is N, which is the total number of current accumulated sliding time windows, providing common input data for dual-model prediction.

[0053] When the total number of cumulative sliding time windows N is less than a preset value, a Gaussian process regression prediction model is used to complete the HI time series prediction and uncertainty quantification. This model uses a window index vector n... With the corresponding HI vector h Assuming the input is a Gaussian prediction function f(·), and that the prior function follows a Gaussian process, satisfying the following conditions: In the formula, n, n' Let n be the sliding time window index, and k(n,n') be the kernel function, representing the covariance between window indices n and n'. For any test point n... The training set output h and the latent function value f at the test point The joint distribution of follows an (N+1)-dimensional Gaussian distribution, satisfying: ,in To train the health index vector, Let be the latent function value at the test point, and 0 be the (N+1)-dimensional zero vector. It is an N×N identity matrix. For the training set kernel matrix, ; Let be the covariance vector between the test points and the training points. The test point autocovariance (scalar) is given. To observe the noise variance, based on the conditional distribution properties of the multivariate Gaussian distribution, the test output in the joint distribution is analyzed. Find the conditional distribution and obtain the test points. The posterior prediction distribution of the latent function is: ,in The posterior predicted mean is essentially the HI observation value of the training set. A weighted linear combination, where the weights are determined by the covariance between the test points and the training set. and the inverse of the training set noise covariance Joint decision; The posterior prediction variance (measures the model's uncertainty about future HI values) is derived from the prior variance at the test points. The variance reduction obtained by subtracting the training set information is different from the observation noise mentioned earlier. (Measuring measurement error). To adapt to the non-steady-state degradation characteristics of the probe, accurately capture the plateau period, and accelerate the degradation inflection point, this scheme adopts the Matérn kernel, the specific expression of which is: ,in, is the Euclidean distance in the input space; l is the length scale parameter, which controls how fast the function changes. The signal variance determines the magnitude of the function. For smoothness parameters; It is the Gamma function; This refers to the modified Bessel function of the second kind. This represents the covariance between window numbers n and n'. This scheme sets... =1.5, corresponding to a first-order differentiable stochastic process, takes into account the following two requirements: first, the ability to capture non-monotonic degradation trends (such as plateau periods and inflection points of accelerated degradation); second, avoiding excessive assumptions about function smoothness, so as to prevent the true degradation fluctuations from being erased due to overly strong smoothing priors. Observation noise is obtained through the joint distribution mentioned above. The introduction of the term, which is physically equivalent to the White Kernel, explicitly models the observation noise caused by electromagnetic interference in the high-temperature liquid lead-based alloy environment, giving GPR an inherent noise filtering capability.

[0054] The search range for kernel function hyperparameters is set as follows: The initial value is set to The aforementioned interval covers all possible scenarios ranging from approximately linear degradation to severe nonlinear degradation. GPR kernel function hyperparameters The log-marginal likelihood (LML) expression is determined through optimization by maximizing the log-marginal likelihood (LML): ,in ,and The variance of the observation noise to be optimized. The jitter term, added to the diagonal of the kernel matrix, is used to improve numerical conditions. This invention employs the L-BFGS-B optimizer, performing optimization with 20 random restarts. Each restart uniformly and randomly samples initial values ​​from the hyperparameter search interval to avoid local optima. The mathematical structure of LML has a clear physical meaning: the first term... The second item measures how well the model fits the observed data; As a penalty for model complexity, it automatically suppresses overly complex kernel function configurations; the third term This is a normalization constant, independent of hyperparameters. The first two terms together constitute the Occam's razor mechanism built into the Bayesian framework, preventing overfitting without the need for external regularization coefficients.

[0055] The training and prediction of GPR follows this process: First, based on the training data... By fitting a Gaussian process model, the optimized kernel function hyperparameters are obtained. Subsequently, back-substitution predictions were performed on historical time points to obtain the fitted mean of the training set. and fit standard deviation For future extrapolation, construct the future window sequence: ,in To predict the step size, perform posterior prediction on the above future window to obtain the predicted mean of future HI. and the predicted standard deviation Finally, the historical data fit is combined with the future prediction to obtain the complete prediction curve: Based on the complete prediction curve described above, the output window number sequence is determined. Predicting the mean sequence and the predicted standard deviation series This serves as the input for subsequent Remaining Useful Life (RUL) calculations. As can be seen from the above prediction extrapolation, GPR simultaneously outputs the predicted mean for each future window. The standard deviation σ is the prediction standard deviation. This standard deviation is not a traditional fitting residual, but a posterior uncertainty measure with strict Bayesian significance. The core advantage of GPR is that the predicted output naturally includes an uncertainty measure. Posterior variance It possesses the following properties, among which The posterior prediction standard deviation is defined as follows: Variance is small near the training data points, indicating high model confidence; variance monotonically increases in extrapolation regions far from the training data, intuitively reflecting increasing uncertainty as the prediction progresses; variance is large in sparse data regions, and the model automatically maintains a conservative estimate. Based on the posterior distribution, the analytical expression for the 95% confidence interval is: This closed-loop computation is more efficient than the Bootstrap resampling or MCMC methods, making it suitable for online real-time prediction scenarios. The prediction results with confidence intervals provided above offer a basis for uncertainty quantification in subsequent remaining useful life calculations.

[0056] When the total number of cumulative sliding time windows N ≥ a preset value, the system automatically switches to a bidirectional long short-term memory autoregressive prediction model for prediction. The input to this model is the same as that of the GPR model, which is a real-time accumulated HI time series. The network architecture uses a 1-dimensional input, 128-dimensional hidden layers, and a stacked 2-layer bidirectional LSTM structure with a dropout rate of 0.3 between layers and bidirectional encoding. The specific configuration of each layer is as follows: Bidirectional LSTM layer: 1 input dimension, 128 hidden dimensions, 2 stacked layers, dropout rate of 0.3 between layers, and bidirectional encoding enabled. The forward LSTM processes the HI sequence in ascending time order. Reverse LSTM processing in reverse time sequence Finally, the hidden states at the last time step in both directions are concatenated, and the output dimension is 2H=256. Layer normalization (LayerNorm): The 256-dimensional vector output by the LSTM is normalized to stabilize the training of deep networks. Fully connected projection layers: FC1: 256→128 (activation function LeakyReLU(0.1)), FC2: 128→64 (activation function LeakyReLU(0.1)), FC3: 64→1 (linear output). Dropout (p=0.3) is inserted between layers to prevent overfitting. The total number of parameters of this network is approximately 5.71×105, and the specific structure is as follows: Bidirectional LSTM parameter count The number of parameters for the second-layer bidirectional LSTM is 4 × (256 + 128 + 2) × 128 × 2 = 395264; the total number of LSTM parameters is 529408; the number of normalized parameters for each layer is 256 × 2 = 512; the number of parameters for the three fully connected layers is (256 × 128 + 128) + (128 × 64 + 64) + (64 × 1 + 1) = 41217; the total number of network parameters is 529408 + 512 + 41217 = 571137.

[0057] The bidirectional long short-term memory autoregressive prediction model reconstructs HI prediction into an autoregressive time-series learning task, assuming the probe health index sequence is... The sequence length L is adaptively determined based on the amount of data: That is, the lower bound of the sequence length is 8, the upper bound is 20, and the default value is one-quarter of the data volume. Training samples are constructed using a sliding time window: let the sample index... The input sequence is The target value is This construction method transforms N original data points into NL training samples. This sliding window sampling method is essentially standard time series resampling, without introducing additional random transformations or interpolation operations. Before training, the HI sequence is Min-Max normalized: ,in The minimum value of the health index series, This represents the maximum value of the health index sequence. The predicted output is mapped back to the original scale via an inverse transformation.

[0058] The following training configuration was used, with all parameters being the actual values ​​set in the code. Loss function: Huber loss. , Let h be the true HI value and h' be the model prediction value. This loss exhibits a quadratic loss (L2) when the absolute value of the error is less than 0.1, and degenerates into a linear loss (L1) when it is greater than 0.1, balancing convergence speed and robustness to outliers. Optimizer: AdamW, initial learning rate... Weight decay coefficient AdamW decouples weight decay from L2 regularization to avoid interference from the adaptive learning rate on the weight decay effect. Learning rate scheduling: Cosine annealing, period T. max =100 epochs, let t be the current epoch number (t=0, 1, ..., The learning rate is calculated using a cosine function from the initial value. =0.003 decays to 0: Gradient clipping: Maximum gradient norm of 5.0 to prevent gradient explosion during backpropagation. Early stopping: Patience value of 15 epochs. If the training loss does not reach a new low for 15 consecutive epochs, training is terminated and the model parameters are rolled back to the historical best. Training mode: Full-batch training, the entire training set is fed into the network at once, and training is conducted for a maximum of 100 epochs (subject to the early stopping mechanism).

[0059] During the inference phase, the MC Dropout ensemble strategy is employed to reduce prediction variance. The specific process is as follows: During the inference phase, the Dropout layer remains active, and M=5 random forward propagations are performed on the same input sequence (5 being an empirically optimal value that balances prediction stability and computational efficiency), and the mean prediction is taken. As a point estimate, the forecast standard deviation As a measure of uncertainty. Let... Let be the predicted output of the m-th random forward propagation, then: The update rule for autoregressive extrapolation is as follows: using the last L values ​​of the normalized HI sequence as initial input, after each prediction step, the output value is fed back to the end of the input sequence (discarding the oldest value), and this process is repeated for T steps, where T is the prediction step size. Finally, Min-Max inverse normalization is performed on the prediction results. ,in The predicted mean after inverse normalization. The standard deviation is the predicted value after inversion. This inversion maps the model's predicted output at the normalized scale back to the original physical scale of HI, ensuring the prediction results are comparable to historical HI sequences and failure thresholds. The LSTM does not output the standard deviation for historical segments, only for future segments. Based on the predicted mean sequence and its standard deviation after inversion, the output sequence includes the window number sequence, the predicted mean sequence, and the predicted standard deviation sequence.

[0060] After completing the HI time series prediction, failure time detection and remaining useful life calculation are performed based on the window number sequence, predicted mean sequence, and predicted standard deviation sequence output by the model, with a preset fixed HI failure threshold. Let the last window number of the prediction curve be... =N+T, corresponding to the HI prediction value. =h( Suppose the prediction model outputs a sequence of future window numbers. (T is the prediction step size) and the corresponding HI prediction value The detection process is as follows: Step 1: Let the index of the future window sequence be... Corresponding window number and HI predicted value The index m is used to find the first index m that satisfies the condition. The index m. If m=1 (the first predicted point is below the threshold), the probe is considered invalid, and RUL=0. Step 2: If m>1, use linear interpolation to calculate the precise cross-window index: In the formula: The sequence number of the adjacent window before and after the intersection; ， This corresponds to the HI predicted value. Step 3: If no value is found after traversing the entire prediction sequence... ≤ (If HI remains above the threshold within the prediction range), the system will then display "No failure trend detected within the prediction range; the estimated safe usage time will be output." The system records the prediction endpoint status (window number, prediction mean, and prediction standard deviation) and outputs the corresponding window number, prediction mean, and prediction standard deviation as input for subsequent remaining useful life (RUL) calculations.

[0061] Remaining Useful Life (RUL) is defined as the difference between the time of failure and the current time. When Step 2 successfully detects an intersection, or Step 3 outputs the predicted endpoint, the RUL point window size is: N is the current accumulated number of windows. The number of failure windows (determined by interpolation in Step 2, and in Step 3) Since the sliding time window step is 15 seconds, the window number form of RUL is converted to minutes (min): The uncertainty of RUL propagates through the prediction standard deviation. Let the prediction standard deviation at failure index m be... (That is, the m-th element of the predicted standard deviation sequence output by GPR or LSTM-AR, and the average standard deviation of the farthest predicted segment is taken during extrapolation in Step 3), then the RUL confidence interval is converted through error propagation: Convert to minutes: The value of 1.96 represents the 97.5th percentile of the standard normal distribution, corresponding to a 95% confidence level. The lower confidence limit is truncated to 0 to avoid non-physical explanations of negative lifetime. The above formula directly maps the prediction uncertainty in the HI domain to the RUL confidence interval in the time domain. Its physical meaning is: if the actual fluctuation range of HI is greater than the predicted mean, the failure time may be earlier, and vice versa.

[0062] In one implementation, Figure 5 This is a schematic diagram of a calibration device for the characteristic benchmark values ​​of conductivity probe performance evaluation indicators in a normal temperature and pressure water environment. Figure 6 Taking a pool-type high-temperature liquid lead-bismuth environment as an example, a schematic diagram of an application scenario for the online conductivity probe monitoring and lifetime prediction method proposed in this invention is shown. It should be noted that... Figure 5 and Figure 6 The protective gas used in the device is nitrogen, but the method proposed in this invention is also applicable to other types of non-condensable gases. Meanwhile... Figure 6This is just one feasible application scenario of the present invention. The method of the present invention can be applied to liquid lead-bismuth environment, liquid pure lead environment, pool-type liquid lead-based alloy environment with zero liquid lead-based alloy flow rate, and loop-type liquid lead-based alloy environment with greater than zero liquid lead-based alloy flow rate.The entire experimental setup mainly consists of a conductivity probe subsystem, a nitrogen injection subsystem, an experimental vessel, and a storage tank system. The conductivity probe subsystem includes: 1. a dedicated control computer for the conductivity probe measurement system; 2. a dedicated high-frequency signal acquisition unit for the conductivity probe with a sampling frequency greater than 10kHz; and 3. a conductivity probe (this is only a schematic diagram; the actual conductivity probe can be configured with multiple probes according to actual usage requirements). The nitrogen injection subsystem includes: 4. a high-pressure nitrogen cylinder; 5. a high-pressure nitrogen cylinder pressure reducing valve; 6. a peristaltic pump for gas injection; 7. a nitrogen gas flow meter; and 8. a nitrogen gas injection conduit, which is immersed below the surface of the liquid lead-bismuth. Nitrogen bubbles are continuously introduced into the liquid lead-bismuth pool from the conduit outlet. The experimental vessel and storage tank system includes: 9. Experimental apparatus power control cabinet; 10. Experimental apparatus signal acquisition unit, which is responsible for collecting signals from all measuring components in the experimental apparatus except for the conductivity probe, including the measurement signals from the gas flow meter, thermocouple, and level gauge; 11. Liquid lead-bismuth level gauge for measuring the liquid lead-bismuth level in the experimental vessel; 12. Thermocouple for measuring the temperature of the liquid lead-bismuth in the experimental vessel; 13. Experimental vessel heating coil; and 14. Experimental vessel body with a lid. The experimental vessel body is made of high-temperature resistant and lead-bismuth corrosion-resistant stainless steel, and the vessel opening is equipped with a sealing cap to prevent the evaporation of liquid lead-bismuth and to block the flow of liquid lead-bismuth. External impurities enter; 15. Liquid lead-bismuth is contained inside the experimental vessel. The liquid lead-bismuth can be transferred between the experimental vessel and the storage tank according to experimental needs; 16. The experimental vessel is covered with a pressure gauge for gas pressure measurement; 17. Experimental vessel exhaust valve; 18. Vacuum pump; 19. Main exhaust valve; 20. Gas path differential pressure regulating valve. The gas path differential pressure regulating valve switches the gas path in conjunction with a three-way valve to regulate the gas pressure difference in the vessel to drive the liquid lead-bismuth to flow and transport between the experimental vessel and the storage tank; 21. Three-way valve; 22. High-pressure argon-hydrogen mixed gas cylinder. The gas cylinder generally uses an argon-hydrogen mixed gas of 95% argon and 5% hydrogen to reduce the oxides inside the liquid lead-bismuth. 3. A valve connecting the experimental tank and the storage tank, which is a liquid metal pipeline valve; 24. A heating coil for the storage tank; 25. The storage tank body with a lid; 26. Liquid lead-bismuth contained inside the storage tank, which can be transferred between the experimental tank and the storage tank according to experimental needs; 27. A thermocouple for measuring the temperature of the liquid lead-bismuth inside the storage tank; 28. A pressure gauge for measuring the pressure of the covered gas inside the storage tank; 29. ​​An exhaust valve for the storage tank. In addition, there is a 30 ambient temperature and pressure experimental tank, which is used to conduct ambient temperature and pressure visual control experiments. Compared with the 14 experimental tank body, its tank material is replaced with plexiglass (Polymethyl Methacrylate, PMMA), which can clearly identify the internal bubble morphology. The medium inside the tank is ambient temperature and pressure liquid water.

[0063] In one implementation, the boundary conditions of this invention are as follows: A four-probe conductivity probe is selected to measure the conductivity. The method proposed in this invention can be independently applied to each of the four probes. Since for multi-probe probes, performance degradation or even functional failure of any probe means that the data from that probe is no longer reliable, it is essential to conduct online performance monitoring and lifespan prediction for each probe in a multi-probe probe system. The conductivity probe measurement system uses a 10V DC drive voltage, sets the sampling frequency to 10kHz, and uses liquid lead-bismuth as the liquid lead-based alloy as the experimental medium. Firstly... Figure 5 The ambient temperature and pressure water environment device shown completes the calibration of the characteristic benchmark values ​​of the conductivity probe performance evaluation index, and then uses... Figure 6 The high-temperature liquid lead-bismuth pool-type experimental device shown is used to carry out online monitoring and lifetime prediction of the conductivity probe in the high-temperature liquid lead-based alloy environment. During the online monitoring process, the sliding time window length is defined as 30s and the step size is 15s, and the overlap rate of two adjacent sliding time windows reaches 50%.

[0064] In one implementation, the calibration of the performance evaluation index characteristic benchmark value of the conductivity probe in a normal temperature and pressure water environment is specifically implemented as follows: First, the conductivity probe to be tested is installed in the normal temperature and pressure experimental container, ensuring that the probe tip electrode is completely immersed in the water. Gas is continuously injected into the water through a gas injection subsystem. A transparent plexiglass device allows visual or high-speed imaging to confirm that the injected bubbles can be effectively detected by the probe. Next, the voltage signal output by the probe is continuously acquired for 30 minutes to ensure sufficient bubble events are collected for statistical analysis. Then, the acquired data is processed to calculate the "voltage transient response slope" and "gas phase voltage amplitude" of all bubble events within the acquisition time. Taking the performance evaluation index of any probe in the four-channel conductivity probe as an example, its voltage transient response slope and gas phase voltage amplitude histogram distributions are as follows: Figure 7 and Figure 8 As shown in the figure, the 40th percentile (shown by the red dashed line in the figure) is taken as the lower limit of the confidence interval for the performance index characteristic value, and the maximum value is taken as the upper limit.

[0065] In one implementation method, the specific implementation process of online monitoring of conductivity probe in a high-temperature liquid lead-based alloy environment is as follows: First, carry out device sealing and equipment debugging. Precisely connect the lid of experimental tank No. 14 to the tank opening and tighten the sealing bolts evenly to ensure the sealing performance of the tank. Only the inert gas inlet and outlet valves are kept open. All other interfaces, such as the probe mounting hole, are sealed with sealing plugs to prevent air from entering the tank. Complete the wiring connection between the data acquisition device, computer and probe. Start the acquisition device and debug the core parameters. Set the sampling frequency to 10kHz to be consistent with the parameters of the subsequent data processing program. Verify that the device can stably acquire probe voltage signals. After confirming that there are no errors, fix the probe and other components. Subsequently, an airtightness test was conducted on the system consisting of experimental tank No. 14, connecting pipelines, and storage tank No. 25. An initial static pressure of 0.31 MPa (gauge pressure) was first applied to the closed experimental system. After pressurization, the absolute pressure data of the system was continuously recorded at a fixed sampling period of 30 minutes. The pressure decay rate ΔP / Δt per unit time was calculated based on the pressure-time series data. When the decay rate was continuously lower than the threshold of 0.002 MPa / h, the airtightness of the system was deemed to meet the standard. After the device is sealed and the equipment is debugged to pass inspection, the lead-bismuth introduction and equipment start-up operations are carried out. Beforehand, check the connecting pipeline between storage tank No. 25 and experimental tank No. 14 to confirm that the pipeline is sealed without damage and the interface is tightly connected. First, close the gas outlet valve of storage tank No. 25, and then open the gas inlet valve 21 and the valve 23 between storage tank No. 25 and experimental tank No. 14 in sequence. Gas is introduced into storage tank No. 25 through the gas inlet. The pre-treated liquid lead-bismuth is pushed into experimental tank No. 14 by the gas pressure difference. The liquid level gauge signal is continuously observed during the pushing process. After the lead-bismuth liquid level reaches the set position, the connecting valve and the gas inlet valve of the storage tank are closed in sequence to complete the feeding. The feed port of the experimental tank is sealed to maintain the system's sealing state. The heating and temperature control system of experimental tank No. 14 is started. The heating power is adjusted to stabilize the ambient temperature inside the tank at the set temperature. After the temperature stabilizes, the pressure reducing valve 5 is opened. After setting the gas flow rate, gas is injected through the gas injection pipe 8. After the operating conditions are stable, the data recording work can be started. Afterwards, conductivity probe data acquisition was carried out. The data acquisition equipment was started and the probe voltage signal was continuously acquired at a sampling frequency of 10kHz. Environmental maintenance and regular inspection were carried out simultaneously throughout the experiment. In terms of temperature control, thermocouples and temperature sensors were used to monitor and adjust the heating power in real time to keep the temperature inside the tank stable at 350℃ with a temperature fluctuation of ≤±5℃. In terms of liquid level control, if the lead-bismuth liquid level dropped due to volatilization, a small amount of liquid lead-bismuth could be added as long as the oxygen content inside the tank was ≤0.05%. In terms of pressure control, the pressure inside the tank was monitored in real time through the top pressure gauge. When the pressure was lower than 0.02MPa, argon gas was slowly added through the gas injection device to maintain a slightly positive pressure environment inside the tank.The standard procedure for regular inspections is to record the temperature, liquid level, and probe operating status inside the tank every hour, and to check the pressure of the inert gas cylinder, the gas injection device, and the vacuum device every two hours. Abnormal handling rules should also be established: the sealing cap must not be opened during the entire experiment; if the thermocouple temperature exceeds 450℃, the heating device should be immediately shut off, and the fault should be investigated after the tank has cooled to room temperature; if the pressure gauge shows abnormal pressure fluctuations, the sealing condition and the gas injection device should be checked promptly and the parameters adjusted; if the probe signal experiences interruptions, sudden amplitude changes, or increased noise, the time of occurrence, the phenomenon, and the corresponding parameters such as temperature, pressure, and oxygen content inside the tank should be recorded; if necessary, the experiment should be stopped and the probe status checked. After continuous voltage signal acquisition, real-time performance evaluation and lifespan prediction of the conductivity probe can be performed. The online monitoring system automatically calculates the Health Index (HI) after acquiring the voltage signal, continuously updates the historical HI curve, and achieves real-time probe performance evaluation. Once at least 10 sliding time windows of HI data have been accumulated, the online monitoring system activates the adaptive lifespan prediction module. The system automatically adapts the prediction model based on the current accumulated time window number N. When N < 500 (preset value), Gaussian process regression (GPR) is used for lifespan prediction; when N ≥ 500 (preset value), it automatically switches to an autoregressive prediction model based on bidirectional LSTM (LSTM-AR) for lifespan prediction. Figure 9 As shown, when the number of sliding time windows is 451, the GPR model is used for prediction. The horizontal axis represents the sliding window number, and the vertical axis represents the health index HI. At this time, the health indices HI of the four probes of the conductivity probe are CH0: 0.940, CH1: 0.857, CH2: 0.724, and CH3: 0.701, respectively. The HI values ​​of each channel are all higher than the mild degradation threshold of 0.7, and all are in a usable and normal state. The health index HI is set to drop to 0.5 or below as the failure threshold. The model predicts that the expected remaining lifespan of the four probes is at least 19.5 minutes. According to the conservative safety principle, the minimum expected remaining lifespan of the four probes is selected as the expected lifespan of the entire conductivity probe, i.e., 19.5 minutes. Figure 10 As shown, when the cumulative sliding time window reaches 702, the system automatically switches to the LSTM-AR model for lifetime prediction. At this time, the health indices (HI) of the four probes are CH0: 0.830, CH1: 0.756, CH2: 0.712, and CH3: 0.695, respectively, all within the normal usable range. Within the prediction range of the model's 702-window duration, no probe HI was detected to drop to the 0.5 failure threshold, and none of the probes in each channel showed a significant degradation trend, maintaining a normal usable state. The above online monitoring, performance evaluation, and lifetime prediction process will be continuously executed in a loop along with the conductivity probe measurement process. Each health index evaluation result is automatically archived to the database, ultimately forming a complete full lifecycle health record for the conductivity probe.

[0066] In summary, this embodiment first obtains the reference range of voltage transient response slope and gas phase voltage amplitude of the conductivity probe under normal conditions. This establishes a standardized and accurate reference for subsequent probe state discrimination and performance quantification evaluation, effectively avoiding the problems of traditional detection methods such as lack of unified evaluation criteria, subjective state judgment, and large errors. It can adapt to the basic performance evaluation standards of probes under different operating conditions, ensuring the accuracy of the reference for subsequent state monitoring and lifespan extrapolation. Next, the collected voltage signal time series data is processed by a sliding time window to obtain the average voltage transient response slope and average gas phase voltage amplitude of all bubble events within the current sliding time window. This achieves noise reduction and refined decomposition of the original monitoring data, eliminating the drawback of ignoring local operating condition fluctuations in overall data statistics. It can accurately capture subtle voltage parameter changes caused by bubble interference and performance degradation during probe operation, making the collected state data more consistent with the real-time actual working state of the probe and improving the effectiveness of data representation. The system is designed for both specificity and relevance. Then, based on the voltage transient response slope reference range, the gas phase voltage amplitude reference range, the average voltage transient response slope, and the average gas phase voltage amplitude, a health index is obtained. This transforms the abstract state of probe aging and performance degradation into a quantifiable numerical indicator, enabling digital, visual, and precise assessment of the probe's operational health status. This overcomes the limitations of traditional methods relying on manual experience or single parameters to determine equipment status, comprehensively reflecting the probe's current working performance and wear level. Finally, based on the health index, a health index time series is formed by sequentially accumulating and arranging the window numbers. A time-series regression prediction model is used to predict the health index time series, and failure time detection and remaining service life calculation are performed to obtain the remaining service life of the conductivity probe. This achieves intelligent processing throughout the entire process, from real-time status monitoring to trend prediction and lifespan forecasting, eliminating the passive mode of traditional periodic maintenance and post-fault repair, and allowing for early prediction of probe failure risks. Overall, this invention constructs a complete technical system from benchmark calibration, real-time data refinement, health status quantitative assessment to time-series lifetime prediction. Through multi-dimensional parameter fusion and time-series data analysis, it improves the accuracy and stability of conductivity probe status monitoring and remaining lifetime prediction, effectively avoiding detection errors caused by operating condition fluctuations and environmental interference. It can monitor the probe's entire life cycle operating status in real time and dynamically, accurately predict equipment failure points and remaining usable time, and provide reliable data support for preventive maintenance, on-demand replacement, and equipment operation and maintenance scheduling. It avoids problems such as monitoring data distortion and production operation hazards caused by premature probe failure and performance drift, and eliminates the waste of operation and maintenance costs caused by excessive replacement and blind maintenance. It improves the intelligence level of conductivity probe operating condition monitoring and the economy, stability, and safety of industrial production operation and maintenance.

[0067] like Figure 11As shown in the illustration, this embodiment also provides an online monitoring and remaining service life prediction system for conductivity probe performance. This system includes: a reference range acquisition module 10, a module 20 for acquiring the average voltage transient response slope and the average gas phase voltage amplitude, a health index acquisition module 30, and a remaining service life acquisition module 40. Specifically, the reference range acquisition module 10 is used to obtain the reference range of the voltage transient response slope and the reference range of the gas phase voltage amplitude of the conductivity probe under normal conditions. The module 20 for acquiring the average voltage transient response slope and the average gas phase voltage amplitude is used to process the collected voltage signal time series data through a sliding time window to obtain the average voltage transient response slope and the average gas phase voltage amplitude of all bubble events within the current sliding time window. The health index acquisition module 30 is used to obtain a health index based on the reference range of the voltage transient response slope, the reference range of the gas phase voltage amplitude, the average voltage transient response slope, and the average gas phase voltage amplitude. The remaining service life acquisition module 40 is used to form a health index time series by accumulating and arranging the health index sequentially according to the window number based on the health index, predict the health index time series through a time series regression prediction model, and perform failure time detection and remaining service life calculation to obtain the remaining service life of the conductivity probe.

[0068] In one implementation, the reference range acquisition module 10 includes: The voltage transient response slope dataset and gas phase voltage amplitude dataset acquisition unit are used to continuously measure bubble events for a preset time in a normal temperature and pressure water environment based on the fabricated conductivity probe, and obtain the voltage transient response slope dataset and gas phase voltage amplitude dataset of the conductivity probe in a normal temperature and pressure water environment. The voltage transient response slope reference range acquisition unit is used to obtain the voltage transient response slope reference range of the conductivity probe under normal conditions by using the preset quantile of the voltage transient response slope dataset as the lower limit and the maximum value of the voltage transient response slope dataset as the upper limit. The gas phase voltage amplitude reference range acquisition unit is used to obtain the gas phase voltage amplitude reference range of the conductivity probe under normal conditions by using the preset quantile of the gas phase voltage amplitude dataset as the lower limit and the maximum value of the gas phase voltage amplitude dataset as the upper limit.

[0069] In one implementation, the voltage transient response slope average value and gas phase voltage amplitude average value acquisition module 20 includes: The voltage transient response slope and gas phase voltage amplitude acquisition unit is used to process the acquired voltage signal time series data through a sliding time window to obtain the voltage transient response slope and gas phase voltage amplitude of all bubble events within the current sliding time window. The voltage transient response slope average value and gas phase voltage amplitude average value acquisition unit are used to perform an arithmetic average of the voltage transient response slope based on the number of bubble events to obtain the voltage transient response slope average value, and to perform an arithmetic average of the gas phase voltage amplitude based on the number of bubble events to obtain the gas phase voltage amplitude average value.

[0070] In one implementation, the health index acquisition module 30 includes: The voltage transient response slope achievement rate acquisition unit is used to compare the average voltage transient response slope with the voltage transient response slope reference range to obtain the voltage transient response slope achievement rate; A gas phase voltage amplitude achievement rate acquisition unit is used to compare the average gas phase voltage amplitude with the gas phase voltage amplitude reference range to obtain the gas phase voltage amplitude achievement rate. The health index acquisition unit is used to weight and fuse the voltage transient response slope achievement rate and the gas phase voltage amplitude achievement rate to obtain the health index.

[0071] In one implementation, the voltage transient response slope achievement rate acquisition unit includes: The first determining subunit for voltage transient response slope achievement rate is used to determine the voltage transient response slope achievement rate as 1 if the average voltage transient response slope is greater than or equal to the lower limit of the voltage transient response slope reference range. The second determining subunit for voltage transient response slope achievement rate is used to determine the voltage transient response slope achievement rate as the ratio of the average voltage transient response slope to the lower limit of the voltage transient response slope reference range if the average voltage transient response slope is less than the lower limit of the voltage transient response slope reference range.

[0072] In one implementation, the gas phase voltage amplitude achievement rate acquisition unit includes: The first determining subunit for gas phase voltage amplitude achievement rate is used to determine the gas phase voltage amplitude achievement rate as 1 if the average value of the gas phase voltage amplitude is greater than or equal to the lower limit of the gas phase voltage amplitude reference range. The second determining subunit for gas phase voltage amplitude achievement rate is used to determine the gas phase voltage amplitude achievement rate as the ratio of the average gas phase voltage amplitude to the lower limit of the gas phase voltage amplitude reference range if the average gas phase voltage amplitude is less than the lower limit of the gas phase voltage amplitude reference range.

[0073] In one implementation, the remaining useful life acquisition module 40 includes: The first unit for obtaining the remaining service life of the conductivity probe is used to predict the health index time series through a Gaussian process regression prediction model when the cumulative sliding time window number is less than a preset value, and to perform failure time detection and remaining service life calculation to obtain the remaining service life of the conductivity probe. The second unit for obtaining the remaining lifespan of the conductivity probe is used to predict the health index time series through a bidirectional long short-term memory autoregressive prediction model when the cumulative sliding time window number is greater than or equal to a preset value, and to perform failure time detection and remaining lifespan calculation to obtain the remaining lifespan of the conductivity probe.

[0074] The working principle of each module in the online monitoring and remaining service life prediction system of conductivity probe performance in this embodiment is the same as that of each step in the above method embodiment, and will not be repeated here.

[0075] Based on the above embodiments, the present invention also provides a terminal, the principle block diagram of which can be as follows: Figure 12 As shown. The terminal may include one or more processors 100 ( Figure 12 (Only one is shown in the image), memory 101, and a computer program 102 stored in memory 101 and executable on one or more processors 100, such as a conductivity probe performance online monitoring and remaining lifespan prediction program. When one or more processors 100 execute computer program 102, they can implement the various steps in the embodiments of the conductivity probe performance online monitoring and remaining lifespan prediction method. Alternatively, when one or more processors 100 execute computer program 102, they can implement the functions of each module / unit in the embodiments of the conductivity probe performance online monitoring and remaining lifespan prediction method, which is not limited here.

[0076] In one embodiment, the processor 100 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0077] In one embodiment, memory 101 may be an internal storage unit of an electronic device, such as a hard drive or RAM. Memory 101 may also be an external storage device of the electronic device, such as a plug-in hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc. Furthermore, memory 101 may include both internal and external storage units. Memory 101 is used to store computer programs and other programs and data required by the terminal. Memory 101 can also be used to temporarily store data that has been output or will be output.

[0078] Those skilled in the art will understand that Figure 12 The block diagram shown is merely a partial structural diagram related to the present invention and does not constitute a limitation on the terminal to which the present invention is applied. A specific terminal may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0079] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, operational databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual operating data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for online monitoring of conductivity probe performance and prediction of remaining service life, characterized in that, The method includes: Obtain the reference range of voltage transient response slope and gas phase voltage amplitude of the conductivity probe under normal conditions; The collected voltage signal time series data is processed by a sliding time window to obtain the average slope of the voltage transient response and the average amplitude of the gas phase voltage for all bubble events within the current sliding time window. The health index is obtained based on the voltage transient response slope reference range, the gas phase voltage amplitude reference range, the average voltage transient response slope, and the average gas phase voltage amplitude. Based on the health index, the health index time series is formed by accumulating and arranging the window numbers in sequence. The health index time series is predicted by a time series regression prediction model, and failure time detection and remaining service life calculation are performed to obtain the remaining service life of the conductivity probe.

2. The method for online monitoring of conductivity probe performance and prediction of remaining service life according to claim 1, characterized in that, The reference range for the voltage transient response slope and the reference range for the gas phase voltage amplitude of the conductivity probe under normal conditions include: Based on the fabricated conductivity probe, continuous measurements of bubble events were performed for a preset time in a water environment at room temperature and pressure, resulting in a dataset of voltage transient response slope and gas phase voltage amplitude of the conductivity probe in a water environment at room temperature and pressure. The reference range of voltage transient response slope under normal conditions is obtained by using the preset quantile of the voltage transient response slope dataset as the lower limit and the maximum value of the voltage transient response slope dataset as the upper limit; By using the preset quantile of the gas phase voltage amplitude dataset as the lower limit and the maximum value of the gas phase voltage amplitude dataset as the upper limit, the reference range of the gas phase voltage amplitude of the conductivity probe under normal conditions is obtained.

3. The method for online monitoring of conductivity probe performance and prediction of remaining service life according to claim 1, characterized in that, The process of processing the acquired voltage signal time series data through a sliding time window to obtain the average slope of the voltage transient response and the average amplitude of the gas phase voltage for all bubble events within the current sliding time window includes: The collected voltage signal time series data is processed by a sliding time window to obtain the voltage transient response slope and gas phase voltage amplitude of all bubble events within the current sliding time window; The average voltage transient response slope is obtained by arithmetically averaging the voltage transient response slope based on the number of bubble events, and the average gas phase voltage amplitude is obtained by arithmetically averaging the gas phase voltage amplitude based on the number of bubble events.

4. The method for online monitoring of conductivity probe performance and prediction of remaining service life according to claim 1, characterized in that, The health index, derived based on the voltage transient response slope reference range, the gas phase voltage amplitude reference range, the average voltage transient response slope, and the average gas phase voltage amplitude, includes: The average voltage transient response slope is compared with the voltage transient response slope reference range to obtain the voltage transient response slope achievement rate; The average value of the gas phase voltage amplitude is compared with the reference range of the gas phase voltage amplitude to obtain the gas phase voltage amplitude achievement rate; The health index is obtained by weighting and fusing the voltage transient response slope achievement rate and the gas phase voltage amplitude achievement rate.

5. The method for online monitoring of conductivity probe performance and prediction of remaining service life according to claim 4, characterized in that, The step of comparing the average voltage transient response slope with the voltage transient response slope reference range to obtain the voltage transient response slope achievement rate includes: If the average value of the voltage transient response slope is greater than or equal to the lower limit of the voltage transient response slope reference range, then the voltage transient response slope achievement rate is 1. If the average voltage transient response slope is less than the lower limit of the voltage transient response slope reference range, then the voltage transient response slope achievement rate is the ratio of the average voltage transient response slope to the lower limit of the voltage transient response slope reference range.

6. The method for online monitoring of conductivity probe performance and prediction of remaining service life according to claim 4, characterized in that, The step of comparing the average value of the gas phase voltage amplitude with the reference range of the gas phase voltage amplitude to obtain the gas phase voltage amplitude achievement rate includes: If the average value of the gas phase voltage amplitude is greater than or equal to the lower limit of the reference range of the gas phase voltage amplitude, then the gas phase voltage amplitude achievement rate is 1. If the average value of the gas phase voltage amplitude is less than the lower limit of the reference range of the gas phase voltage amplitude, then the gas phase voltage amplitude achievement rate is the ratio of the average value of the gas phase voltage amplitude to the lower limit of the reference range of the gas phase voltage amplitude.

7. The method for online monitoring of conductivity probe performance and prediction of remaining service life according to claim 1, characterized in that, The process of predicting the health index time series using a time-series regression prediction model, and performing failure time detection and remaining service life calculation to obtain the remaining service life of the conductivity probe includes: When the cumulative sliding time window number is less than the preset value, the health index time series is predicted by the Gaussian process regression prediction model, and failure time detection and remaining service life calculation are performed to obtain the remaining service life of the conductivity probe. When the cumulative sliding time window number is greater than or equal to the preset value, the health index time series is predicted by a bidirectional long short-term memory autoregressive prediction model, and failure time detection and remaining service life calculation are performed to obtain the remaining service life of the conductivity probe.

8. A system for online monitoring of conductivity probe performance and prediction of remaining service life, characterized in that, The system includes: The reference range acquisition module is used to acquire the reference range of voltage transient response slope and gas phase voltage amplitude of the conductivity probe under normal conditions. The voltage transient response slope average and gas phase voltage amplitude average acquisition module is used to process the collected voltage signal time series data through a sliding time window to obtain the voltage transient response slope average and gas phase voltage amplitude average of all bubble events within the current sliding time window. The health index acquisition module is used to obtain the health index based on the voltage transient response slope reference range, the gas phase voltage amplitude reference range, the average voltage transient response slope, and the average gas phase voltage amplitude. The remaining service life acquisition module is used to accumulate and arrange the health index time series according to the window number based on the health index, predict the health index time series through the time series regression prediction model, and perform failure time detection and remaining service life calculation to obtain the remaining service life of the conductivity probe.

9. A terminal, characterized in that, The terminal includes a memory, a processor, and a conductivity probe performance online monitoring and remaining lifespan prediction program stored in the memory and executable on the processor. When the processor executes the conductivity probe performance online monitoring and remaining lifespan prediction program, it implements the steps of the conductivity probe performance online monitoring and remaining lifespan prediction method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program for online monitoring of conductivity probe performance and prediction of remaining service life. When the program is executed by a processor, it implements the steps of the method for online monitoring of conductivity probe performance and prediction of remaining service life as described in any one of claims 1-7.

Images