A method for automatic detection and determination of valve seal failure

Abstract

The application discloses an automatic detection and judgment method for valve sealing failure, relates to the field of high-pressure fluid control equipment testing, and aims at the problem that traditional static leak detection methods fail to detect intermittent self-sealing of high-pressure fluid valves caused by particle embedding. The method synchronously collects pressure and temperature measurement sequences after establishing a closed volume cavity; a command level symmetry strategy is used to apply a slight swing excitation to the system; a real gas state equation is used to transform the measurement sequences into equivalent mass sequences; a timestamp is used to calculate a half-way mass change rate, and a dimensionless self-sealing reversibility direction characteristic parameter is constructed; then, a first level symbol consistency judgment based on polarity statistics and a second level reset consistency judgment based on one-dimensional data clustering are respectively performed; finally, two types of logical flag bits are fused, and a final qualitative conclusion is output. The application excludes nonlinear compression and thermodynamic noise interference, and realizes accurate online discrimination of valve micro-dynamic leakage.

Description

Technical Field

[0001] This invention relates to the field of high-pressure fluid control equipment testing, specifically to an automatic detection and judgment method for valve sealing failure. Background Technology

[0002] With the comprehensive rollout of hydrogen energy infrastructure construction, the 70 MPa ultra-high pressure hydrogen valves widely deployed in hydrogen refueling stations, as core components ensuring controlled fluid transmission and pressure safety isolation, directly determine the inherent safety of the equipment in their long-term shut-off state. Due to the small molecular weight of hydrogen and its tendency to diffuse and permeate, even micron-level fluid channels at the sealing interface can cause severe internal leakage. In actual high-frequency refueling and station pressure regulation operations, valves are subjected to frequent and intense start-stop actions and the strong impact loads generated by the instantaneous ultra-high pressure hydrogen flow. These extreme conditions cause the sealing surfaces of the valve core and seat to continuously endure mechanical impacts and alternating stresses, inducing localized wear and material spalling at the microscopic level. Simultaneously, metal materials are highly susceptible to hydrogen embrittlement and degradation under long-term high-pressure hydrogen environments, leading to a non-linear propagation trend of microcracks on the sealing surface, further increasing the uncertainty of seal failure. Therefore, achieving accurate online assessment of the sealing health of valves in their shut-off state without disassembling them has become an indispensable support for the safe operation and maintenance of high-pressure hydrogen energy equipment.

[0003] For online detection of internal leakage in high-pressure valves, existing methods mainly rely on acoustic monitoring signal extraction or single static pressure attenuation comparison. However, the valve core impact energy wave obtained by existing acoustic monitoring and the high-frequency disturbance of microchannel leakage suffer from severe signal aliasing in the spatiotemporal domain, making it difficult to extract the true leakage response. Conventional static pressure comparison methods assume that leakage is a monotonically evolving or quasi-stationary continuous process, judging it only by the absolute pressure drop slope over a fixed period and a set threshold. Under 70 MPa ultra-high pressure conditions, tiny particles or foreign objects worn off from the sealing surface are easily embedded in the leakage microchannel by high-speed fluid scouring and valve core reset action, producing a hidden intermittent self-sealing phenomenon. These foreign particles undergo reversible displacement within the microchannel with pressure differential disturbances, causing the sealing surface to exhibit random switching between blockage and conduction at the same damage level. Existing pressure-holding detection logic is limited by the interference of nonlinear gas compression and environmental heat transfer fluctuations, making it unable to identify such non-stationary switching processes. It is prone to misjudging short-term leak-free phenomena caused by foreign object embedding as successful sealing, or misinterpreting the sudden increase in flow after particles are pushed open as irreversible severe deterioration. Therefore, overcoming the inherent defect of existing technologies in misjudging reversible intermittent self-sealing as a static process under strong thermodynamic background noise interference, and constructing a controlled dynamic test path to transform the random conduction state of microchannels into a repeatable and verifiable deterministic evolution trend, thereby accurately identifying the true type of valve sealing failure, is a core technical challenge that urgently needs to be solved in this field. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes an automatic detection and judgment method for valve sealing failure, which solves the problem that under strong thermodynamic background noise interference, the microscopic reversible intermittent self-blocking phenomenon is easily misjudged as a static evolution process, thus failing to accurately identify the true type of valve failure.

[0005] To achieve the above objectives, the present invention provides the following technical solution: Establish a closed volume cavity under shutdown conditions and obtain the equivalent volume, and simultaneously acquire pressure measurement sequences and temperature measurement sequences; A micro-oscillation excitation is applied using an instruction-level symmetric strategy, and the segmented interval timestamps and steady-state reset extraction times are extracted. The dynamic compressibility factor sequence is obtained, and the pressure measurement sequence and temperature measurement sequence are transformed into an equivalent mass sequence based on the real gas equation of state, combined with the equivalent volume and the dynamic compressibility factor sequence. The rate of change of mass during the upper half of the swing and the rate of change of mass during the lower half of the swing are calculated based on the segmented interval timestamps and the equivalent mass sequence. The equivalent reset mass is extracted based on the steady-state reset extraction time and the equivalent mass sequence. Based on the mass change rate of the upper half-stroke and the mass change rate of the lower half-stroke, a dimensionless self-blocking reversible directional characteristic parameter is constructed, and a first-level sign consistency determination based on polarity statistics is performed to output the logical state flag bit of the directional system deviation. Perform a second-level reset consistency determination based on data clustering on the equivalent reset quality to output the logical status flag bits of the two types of residual reset states; The logical state flags of the integrated directional system deviation and the two types of residual states (reset) are subjected to a logical AND operation to output a qualitative diagnostic conclusion.

[0006] Compared with existing technologies, it has the following advantages: This solution proposes an automatic detection and judgment method for valve sealing failure, abandoning the passive judgment logic of traditional leak detection technology that relies on static pressure holding and fixed thresholds. It innovatively introduces a command-level symmetrical micro-oscillation excitation mechanism. Addressing the intermittent self-sealing illusion caused by foreign particles embedded in the leakage channel under high-pressure hydrogen conditions, this solution actively breaks the short-term mechanical sealing state of the particles by applying controlled dynamic micro-disturbances to the fluid system. Through an active test path, the originally random and uncontrollable changes in the sealing surface conductivity are transformed into a deterministic response process that evolves synchronously with the test command, effectively eliminating the unavoidable system coupling interference in conventional detection. By combining the real gas state equation with an equivalent mass transformation of the acquired sequence, common-mode errors caused by nonlinear compression of ultra-high-pressure fluids and environmental heat transfer fluctuations are eliminated, clearly revealing the evolution trend of micro-leakage hidden in strong background noise. This fundamentally solves the core problem of existing technologies easily misjudging intermittent leak-free phenomena as successful sealing.

[0007] Building upon this foundation, this invention further decouples the constraints of basic test pressure levels and absolute leakage volume on the judgment results by constructing a dimensionless self-sealing reversibility direction parameter. This parameter utilizes the relative difference ratio of the mass change rate during the upper and lower swing halves to transform absolute numerical evaluation into relative evolution ratio evaluation, enabling the system to achieve adaptive diagnosis without pre-setting complex absolute calibration thresholds for different equipment specifications. Combined with polarity statistics' sign consistency test and one-dimensional data clustering separation degree calculation, this scheme can accurately capture the discrete residence distribution of the equivalent conductivity of the leakage channel between the blocked and open states, effectively filtering out system water hammer oscillations and single-shot occasional acquisition errors. Based on composite judgment logic, it can not only accurately output the conclusion of whether the valve has actually leaked, but also keenly identify whether the specific type of leakage belongs to reversible intermittent failure caused by particle migration, significantly improving the robustness and accuracy of online monitoring of high-pressure equipment sealing safety. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0009] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0010] Please see Figure 1 This application provides an automatic detection and judgment method for valve sealing failure; The method specifically includes the following steps: Step 1: The core objective of this invention is to solve the problem of identifying sealing failure in 70 MPa high-pressure hydrogen valves under shut-off conditions, particularly targeting the detection of reversible intermittent self-blocking caused by the reversible embedding or detachment of particles from the microchannels at the sealing interface. This non-stationary leakage manifests as random switching between conduction and closure of the effective conduction cross-section of the microchannel under the same damage level due to temporary blockage by foreign objects and subsequent fluid scouring, making conventional single-pressure decay leak detection techniques unable to provide a stable judgment. To transform this microscopic random phenomenon into macroscopic and objectively clear test data, the primary prerequisite of this method is to establish a closed boundary that conforms to fluid testing standards and obtain multidimensional state parameters with a unified time reference, transforming the degradation problem of valve mechanical seal performance into a thermodynamic state evolution problem within a controlled closed volume. The specific process is as follows: Establish the closed volume test boundary under the closed condition of the valve under test and complete the measurement system configuration.

[0011] Specifically, the fundamental parameters involved in this step include the equivalent volume V of the closed cavity being measured, the hydrogen specific gas constant R, and the compressibility factor of high-pressure hydrogen. The unit of the equivalent volume V is cubic meters. The hydrogen specific gas constant is a known physical constant, with units of joules per kilogram Kelvin. The compressibility factor of high-pressure hydrogen is a dynamic dimensionless parameter used to characterize the degree to which the real gas deviates from the ideal gas and changes in real time with the state; it is denoted as Z(t) with a time independent variable. This step specifically includes the following sub-steps: S101, perform a shut-off action on the target valve under test, and select the valve-side pipeline to construct a closed volume cavity with clear boundaries and constant volume.

[0012] Specifically, a shut-off command is issued to the target valve in the high-pressure hydrogen system of the hydrogen refueling station. After confirming that the valve is fully shut off via valve position feedback signal or mechanical shut-off switch, a fixed pipe section downstream of the target valve is selected to construct a test space that is strictly isolated from the outside world as a closed volume cavity. During the test, all bypass valves, vent valves, and sampling branches connected to the outside world in this closed volume cavity must be in a forced closed or effectively isolated state to determine the equivalent volume V of the closed volume cavity. The above operation cuts off the exchange of unknown fluids with the outside world and establishes a benchmark constraint boundary for mass and energy conservation analysis. This makes the change in gas mass in the cavity strongly correlated only with the leakage behavior of the target valve. Thus, the microchannel leakage of the valve seat sealing surface, which was originally invisible and difficult to measure directly, is transformed into a macroscopic process of material accumulation or loss in the closed space. This provides a unique and constant geometric constraint for subsequent equivalent mass conversion, greatly improving the detection limit of micro-leakage.

[0013] S102, sensors are arranged in the closed volume cavity and upstream of the valve to synchronously collect time series data including pressure and temperature.

[0014] Specifically, a pressure sensor is positioned near the target valve's main flow area within the enclosed cavity to continuously acquire dynamic sequence data of pressure changes over time, denoted as the pressure measurement sequence P(t). A temperature sensor is positioned within the enclosed cavity to avoid fluid dead zones, continuously acquiring dynamic sequence data of temperature changes over time, denoted as the temperature measurement sequence T(t). Simultaneously, an existing pressure measuring point upstream of the valve is utilized to continuously acquire dynamic sequence data of the upstream driving hydrogen gas changes over time, denoted as the upstream pressure measurement sequence U(t). The acquisition system forces the use of the same clock source as the sampling time reference for the above three sensor data to achieve absolute timestamp alignment. Based on the combined strategy of synchronously acquiring the three-dimensional state sequences at the same frequency, the system fully considers and compensates for the significant throttling expansion heat effect when high-pressure hydrogen passes through the microchannel and the interference caused by ambient temperature fluctuations. The three data streams are highly coupled in both spatial and temporal senses, eliminating phase errors caused by differences in the sampling frequencies of different instruments, and providing extremely high data fidelity for accurately capturing the asymmetric weak response under the self-closing mechanism.

[0015] It should be noted that, considering the extremely non-ideal gas characteristics of hydrogen under 70 MPa ultra-high pressure conditions, this invention abandons the conventional ideal gas equation of state in its underlying principle of state transformation, and instead introduces a real gas equation of state for rigorous calculation. The real gas equation of state states that the product of the pressure inside the closed cavity and the equivalent volume equals the dynamic compressibility factor, the equivalent mass of hydrogen inside the closed cavity at the current moment, and the combined product of the hydrogen specific gas constant and the cavity temperature. Its dynamic evolution relationship satisfies the following equation: In the formula, P(t) is the pressure inside the closed cavity at the current moment; V is the equivalent volume of the closed cavity; Z(t) is the dynamic compressibility factor at the current moment; m(t) is the equivalent mass of hydrogen in the closed cavity at the current moment; R is the hydrogen-to-gas constant; and T(t) is the temperature inside the closed cavity at the current moment.

[0016] Preferably, this step can be based on the temperature and pressure lookup table interpolation algorithm of an authoritative physical property database, or it can be performed in real time using the ultra-high pressure hydrogen state equation to dynamically obtain the dynamic compressibility factor sequence Z(t) that matches the current pressure measurement sequence and temperature measurement sequence. This completely eliminates the systematic error caused by gas compressibility and ensures the absolute accuracy of subsequent calculations of minute mass changes.

[0017] It should be noted that, in order to accurately capture the macroscopic mass change when the system re-reaches steady state after particle embedding and detachment, the sampling frequency setting of the data acquisition system in this step must meet specific functional boundary conditions. Specifically, the lower limit of the high-frequency synchronous sampling frequency must be strictly greater than the single-step command output frequency of the controller during subsequent differential pressure micro-oscillation. Preferably, the sampling frequency is not lower than 10 Hz and not higher than 500 Hz to ensure that sufficient effective discrete data points can be acquired within a single actuator action cycle, thereby avoiding feature omissions due to undersampling or high-frequency data redundancy caused by oversampling.

[0018] It should be noted that, given the initial stage after the valve is closed and the closed volume cavity is newly established, the high-pressure hydrogen gas will experience a rapid temperature change and intense heat exchange with the pipe wall due to adiabatic compression or expansion. If the test is started immediately at this point, it will introduce a significant thermodynamic non-equilibrium baseline drift. Therefore, this scheme forcibly introduces a thermodynamic quasi-steady-state settling period determination mechanism after step S102 is completed. That is, before extracting leakage pressure drop data, the measurement and control system needs to monitor the absolute value of the first derivative of the temperature measurement sequence T(t) in real time. Only when the absolute value of the first derivative of the temperature measurement sequence is continuously lower than the set temperature fluctuation threshold (e.g., temperature fluctuation less than 0.1 Kelvin within ten consecutive seconds) can the closed volume cavity be determined to have entered a thermodynamic quasi-steady state. This static moment is then used as the effective starting point for subsequent micro-oscillation excitation and equivalent mass conversion.

[0019] It should be noted that for enclosed cavities with a large length-to-diameter ratio or complex internal structures, a single measuring point is insufficient to represent the true overall temperature. In such cases, multiple temperature sensors can be arranged along the pipeline axis or at different spatial heights, and the spatially weighted average or volume fraction integral of the temperatures at each measuring point can be used as the final temperature measurement sequence T(t). This spatial compensation embodiment effectively eliminates the error caused by local thermal stratification in the overall mass conversion, further enhancing the applicability of the solution in complex irregular pipe networks.

[0020] It should be noted that, regarding the methods for obtaining the equivalent volume V, in addition to standard geometric calculations or looking up tables based on drawings, when faced with complex working conditions where the pipeline space at a hydrogen refueling station is extremely irregular, the water injection volume measurement method known in industrial fluid testing can be used, or the volume value can be extracted after scanning and modeling the on-site pipeline using three-dimensional computer-aided design software.

[0021] It should be noted that for the unavoidable micro-connection paths such as instrument capillaries within the enclosed volume cavity on-site, this solution does not require forced removal. Instead, during the aforementioned thermodynamic quasi-steady-state settling period, the fluid resistance state of this path is confirmed to be constant by observing whether the natural decay slope of the pressure measurement sequence P(t) is stable. This operation can treat this fixed micro-emission as part of the system's inherent leakage boundary, and thus, in subsequent symmetrical differential pressure micro-oscillations, it is naturally canceled out as a common-mode background by the adaptive differential algorithm.

[0022] Specifically, after executing the above steps and steady-state determination, Step 1 ultimately outputs the equivalent volume V and the hydrogen-to-gas ratio constant R. It also outputs the pressure measurement sequence P(t), temperature measurement sequence T(t), upstream pressure measurement sequence U(t), and dynamic compressibility factor sequence Z(t) calculated based on real-time temperature and pressure synchronization, all aligned with a unified timestamp and having entered a thermodynamic quasi-steady state. This combination of parameters and sequences will serve as a lossless and complete raw material, fully transmitted to subsequent steps. It will be directly used to calculate the pressure difference time series across the valve, serving as the feedback control basis for executing the pressure difference symmetrical micro-oscillation command, and providing an analytical foundation for subsequent real-time calculation of the equivalent mass change under micro-oscillation excitation.

[0023] Step Two: Following the operations in Step One, the test system has established a highly controlled closed experimental boundary and confirmed that the system has entered a thermodynamic quasi-steady state, while acquiring a three-dimensional state sequence with absolute time alignment. However, static pressure-holding observation alone cannot effectively stimulate the reversible embedding and detachment mechanism of foreign matter within the microchannels of the sealing interface. To make the hidden reversible intermittent self-blocking phenomenon manifest as a macroscopic feature that can be clearly captured by the algorithm, the core technical logic of Step Two lies in applying a mild and highly symmetrical dynamic differential pressure disturbance to the target valve. By constructing a strictly traceable command-level excitation segment and thermodynamic reset benchmark, this step transforms the originally random failure switching process of the valve into a deterministic response sequence under controlled excitation, thereby providing a crucial and unambiguous temporal framework for subsequent extraction of directional features and reset residual traces. The specific process is as follows: Perform a differential pressure symmetrical micro-oscillation test with the valve kept closed, and define the segmented intervals and reset times using the control command log.

[0024] Specifically, the control and timing characteristic parameters involved in this step include the voltage regulation control step size s, the number of consecutive single-half-stroke applications k, the total number of micro-oscillation executions N, and the segmentation node time of each action. The voltage regulation control step size, denoted as s, characterizes the control amplitude of a single discrete action or equivalent analog quantity change of the voltage regulation actuator. The number of consecutive single-half-stroke applications is the cumulative base value of the directional commands set by the control strategy, denoted as k. The total number of micro-oscillation executions is a preset positive integer for cyclic verification, denoted as N. For the i-th micro-oscillation iteration, the corresponding start time of the upward swing command, end time of the upward swing command, start time of the downward swing command, end time of the downward swing command, end time of the reset command, and steady-state reset extraction time are denoted as follows: , , , , and This step specifically includes the following sub-steps.

[0025] S201 calculates the pressure difference sequence based on the upstream and intracavity pressures and applies pressure difference micro-oscillation excitation using a command-level symmetric strategy.

[0026] Specifically, based on the upstream pressure measurement sequence U(t) and pressure measurement sequence P(t) obtained in step one, the monitoring and control system calculates the pressure difference time series across the valve in real time, denoted as D(t). The calculation relationship satisfies the following equation: In the formula, D(t) is the dynamic pressure difference across the valve at the current moment; U(t) is the pressure of the driving hydrogen upstream of the valve obtained in step one at the current moment; and P(t) is the pressure inside the closed volume cavity obtained in step one at the current moment.

[0027] Under the condition that the target valve is kept closed and the system is in a thermodynamic quasi-steady state, a small and slow symmetrical micro-oscillation excitation is applied to the upstream pressure by finely adjusting the pressure regulating actuator upstream of the target valve. A single micro-oscillation consists of a complete reciprocating motion of first swinging upward, then swinging downward to cross the zero point and finally resetting to the reference pressure. This step adopts a command-level symmetrical control strategy, that is, with the current reference pressure as the initial zero point, k pressure regulating control steps s in the pressure increasing direction are continuously sent to the pressure regulating actuator during the first half of the upward swing, so that the total control command amplitude of the upward swing is the product of k and s; then, during the second half of the downward swing, twice the pressure regulating control steps s in the pressure decreasing direction are continuously sent to the actuator, at which point the system crosses the initial zero point and reaches the maximum reverse disturbance amplitude; finally, k reset commands in the pressure increasing direction are sent to make it accurately return to the initial reference setting state.

[0028] The core advantage of the aforementioned command-level symmetry feature lies in replacing the conventional approach in existing fluid testing techniques that relies on observing actual pressure thresholds to determine when the actuator reverses. Conventional response-level symmetry is highly susceptible to interference from the superposition of nonlinear noise from fluid sensors and pressure attenuation caused by minute leaks in the target valve, resulting in an inherent asymmetry in the applied excitation boundary. This step defines symmetry through the absolute output of the actuator's underlying controller. Regardless of whether this output is a discrete digital step pulse or a continuous equivalent analog voltage ramp, it ensures that the initial perturbation potential energy applied to the valve sealing interface is strictly equal in amplitude at the source. This synergistic effect guarantees that any asymmetric mass loss observed in subsequent steps originates purely from the reversible rearrangement of microchannel particles within the valve, and is not a deviation in the testing method itself, significantly improving the anti-interference capability of the detection algorithm.

[0029] S202 synchronously collects the controller's action logs and strictly extracts the segmented intervals and reset times in conjunction with the fluid steady-state feedback mechanism.

[0030] Specifically, during the differential pressure micro-oscillation of S201, the measurement and control system synchronously records the underlying action log of the pressure regulating actuator and extracts the precise generation or completion timestamp of each instruction to achieve time-series segmentation of the data. For the i-th micro-oscillation, the precise clock points of the first instruction issuance and the last instruction completion of the k consecutive upward oscillation instruction sequences are extracted and defined as the start time of the upward oscillation instruction, respectively. End time of the swing command Extract the precise clock points of the first instruction issuance and the last instruction completion from a sequence of twice k consecutive downward instruction sequences, and define them as the start times of the downward instruction. With the end time of the lowering command Extract the precise clock point at which the last instruction in a sequence of k consecutive reset instructions completes, and define it as the reset instruction end time. It is essential to ensure that all extracted time data uses the same clock source determined in step one as the timing alignment reference.

[0031] It should be noted that, considering that after high-pressure hydrogen undergoes pressure differential micro-oscillation disturbance, the gas pressure diffusion and heat transfer within the closed volume cavity require a specific amount of time to re-establish equilibrium, this step is performed after the reset command is completed at the lowest level and the reset command end time is obtained. Then, an adaptive closed-loop fluid stabilization delay mechanism is forcibly superimposed. The duration of this fluid stabilization delay is not a fixed dead time, but is determined by the system's adaptive closed-loop mechanism.

[0032] Specifically, at the end of the reset command Subsequently, the measurement and control system needs to monitor the first derivative of the intracavity pressure measurement sequence P(t) or temperature measurement sequence T(t) in real time. Only when the rate of change falls back to within the set small background noise threshold range for a preset time can the dual equilibrium of fluid dynamics and thermodynamics be confirmed, and this point is locked as the steady-state reset extraction moment. The threshold range for this small background noise can be adaptively calibrated based on the variance of the sensor's background noise in an absolutely static state. It is typically set within a range of plus or minus three times the standard deviation of the static rate of change, with a recommended preset time of 3 to 5 seconds. This bridges the time lag between the completion of mechanical control commands and the stabilization of the fluid's physical state, ensuring timely extraction during steady-state reset. The extracted state variables completely eliminated the interference of dynamic water hammer effect and transient heat conduction.

[0033] The synergistic effect of this step lies in establishing a rigid time-series segmentation benchmark that integrates underlying execution logic and macroscopic fluid feedback characteristics. Through absolute time intervals, it allows for seamless nesting with the continuously acquired 3D state sequences from Step 1. This enables accurate extraction of effective integral data segments during subsequent dynamic feature extraction without the need for computationally complex waveform extremum point finding algorithms. This significantly reduces the computational cost of data processing and fundamentally avoids the erroneous induction of time segmentation points by non-stationary fluid noise under 70 MPa conditions.

[0034] It should be noted that the setting mechanism for the total number of micro-oscillation executions N can be set to a fixed positive integer constant greater than or equal to 2 to ensure basic statistical repeatability verification capability. Preferably, to balance the online detection efficiency in industrial settings with the robustness of the conclusions, the total number of micro-oscillation executions N can adopt a system-level adaptive dynamic stopping strategy. That is, after the system cyclically executes subsequent quality evolution calculation steps, if the monitoring system determines that the self-blocking directional characteristic conclusions obtained from two consecutive micro-oscillations no longer exhibit polarity reversal, it issues an interrupt command to the measurement and control system, automatically stopping the additional micro-oscillation test and fixing the current total number of micro-oscillation executions N. This provides the solution with extremely high engineering flexibility while ensuring compliance with action timing regulations.

[0035] It should be noted that the product of the basic cumulative amplitude of the control command, k and s, must not only meet the principle of being greater than the hysteresis dead zone of the valve sealing surface micro-friction and less than the valve opening critical pressure difference, but also be coordinated with the equivalent volume of the closed cavity. If the equivalent volume is too large, the gas volume change caused by the small command amplitude will be completely absorbed by the fluid volume elasticity of the pipeline. Therefore, the pressure regulation control step size s and the number of consecutive applications k in the single half-stroke need to be adaptively calculated and matched based on the actual upstream gas supply pressure and the volume of the closed cavity to ensure that macroscopic fluctuations with a signal-to-noise ratio that meet the requirements can be observed on the cavity pressure measurement sequence P(t).

[0036] Specifically, after executing the aforementioned command-level symmetric strategy and introducing a fluid steady-state feedback mechanism, the final output parameters of step two are the pressure regulation control step size *s*, the number of consecutive single-half-stroke applications *k*, and the total number of executions *N*. The output dynamic dataset is the pressure difference time series *D(t)*, as well as the start and end times of the upward swing command, the start and end times of the downward swing command, the end time of the reset command, and the steady-state reset extraction time for the *i*th micro-oscillation. These timing nodes and control parameters will serve as observation windows with absolute time anchors, and will be fully transmitted to subsequent steps, directly guiding step three to perform absolutely accurate data extraction and equivalent mass difference evolution calculations on the continuous thermodynamic state sequence.

[0037] Step 3: Following the controlled execution of Step 2, the test system has successfully applied a command-level symmetrical differential pressure micro-oscillation disturbance to the target valve and acquired the segmented node times of each stage with absolute time anchors. However, in the 70 MPa ultra-high pressure fluid system, directly observing the fluctuations of the pressure sequence is prone to misjudgment. The throttling expansion of high-pressure hydrogen in the microchannel and the surrounding heat transfer will cause strong thermodynamic state fluctuations, making simple pressure changes no longer equivalent to actual leakage. To eliminate common-mode interference caused by temperature fluctuations and nonlinear changes in gas compressibility, Step 3 will introduce the laws of energy and mass conservation. By rigorously projecting the multidimensional thermodynamic state sequence obtained in Step 1 onto the mass dimension and performing precise segmentation and time normalization based on the time window in Step 2, this step can extract the microscopic mass loss evolution characteristics purely caused by the reversible rearrangement of particles. The specific process is as follows: The state variables are converted into equivalent mass sequences based on the segmented intervals, and the half-range mass change rate and reset mass are calculated.

[0038] Specifically, the state evolution and feature extraction parameters involved in this step include the equivalent hydrogen mass sequence within the closed cavity, the mass change rate during the upper half of the swing, the mass change rate during the lower half of the swing, the number of smoothed sampling points, the uniform sampling period, and the equivalent reset mass. The equivalent hydrogen mass sequence within the closed cavity represents the absolute hydrogen mass remaining in the cavity over time, denoted as m(t) with a time independent variable. The mass change rates during the upper and lower half of the swing represent the time-normalized mass loss rates during the two halves of the dynamic excitation, respectively. For the i-th micro-oscillation iteration, they are denoted as follows: and The number of smoothed sampling points is a preset positive integer parameter for the algorithm, denoted as J. The uniform sampling period is the time interval for the measurement and control system hardware to acquire discrete data, denoted as h. The equivalent reset mass characterizes the absolute mass residence level of the fluid after it reaches thermodynamic steady state again, denoted as... This step specifically includes the following sub-steps: S301 integrates multi-dimensional dynamic data sequences to transform the intracavitary thermodynamic state point by point into an equivalent mass sequence.

[0039] Specifically, based on the real gas equation of state established in step one, the measurement and control system performs joint calculations on the pressure measurement sequence P(t), temperature measurement sequence T(t), and the synchronously matched dynamic compressibility factor sequence Z(t) acquired in step one at each discrete sampling point on the time axis. This joint calculation shields the single-dimensional parameter fluctuations caused by environmental heat transfer or local throttling, transforming the invisible microchannel leakage into a quantifiable equivalent mass sequence m(t). Its point-by-point calculation relationship satisfies the following equation: In the formula, m(t) is the equivalent mass of hydrogen in the closed cavity at the current moment, in kilograms; P(t) is the pressure in the closed cavity at the current moment obtained in step one, in Pascals; V is the equivalent volume of the closed cavity determined in step one, in cubic meters; Z(t) is the dynamic compressibility factor at the current moment obtained in step one, which is a dimensionless parameter; R is the hydrogen-to-gas constant determined in step one, in joules per kilogram in Kelvin; and T(t) is the temperature in the closed cavity at the current moment obtained in step one, in Kelvin.

[0040] The core of this step lies in introducing a real gas model containing the dynamic compressibility factor sequence Z(t), which completely decouples the nonlinear distortion between pressure and mass caused by the sudden density change of ultra-high pressure hydrogen. When reversible detachment of particles at the sealing interface leads to a trace escape of hydrogen, even if the instantaneous expansion and heat absorption mask the pressure decay, the equivalent mass sequence m(t) can still effectively reflect the absolute loss of material within the cavity, laying a solid analytical foundation for identifying extremely small leaks under non-destructive conditions.

[0041] S302 accurately captures evolution data based on the control action timing window and calculates the dynamic quality change rate and reset benchmark for each stage.

[0042] Specifically, based on the equivalent mass sequence m(t) generated by S301, and combined with the segmented interval timestamps extracted in step two, the measurement and control system begins closed-loop data interception and differential calculation. Since there is an objective time difference in duration between the upward swing command sequence and the downward swing command sequence set in step two, simply comparing the absolute mass change will introduce a system error due to integration time asymmetry. Therefore, time normalization must be performed in this step. For the i-th micro-oscillation, the difference between the equivalent mass corresponding to the end of the upward swing command and the equivalent mass corresponding to the start of the upward swing command is calculated and divided by the upward swing duration, defined as the rate of change of the mass of the i-th micro-oscillation during the half-stroke of the upward swing. Similarly, the difference between the equivalent mass at the end of the swing command and the equivalent mass at the beginning of the swing command is calculated and divided by the swing duration. This difference is defined as the rate of change of mass of the half-stroke of the i-th micro-oscillation. Its normalized calculation relationship satisfies the following equation: In the formula, Let be the rate of change of mass during the first half of the upper swing of the i-th micro-oscillation, in kilograms per second; The equivalent mass at the moment the swing command ends; The equivalent mass corresponding to the start of the upward swing command; Let be the rate of change of mass of the i-th micro-oscillation during the half-stroke of the lower swing, in kilograms per second; The equivalent mass corresponding to the end of the swing command; The equivalent mass corresponding to the start of the swing command; After extracting the mass change rate for both the first and second halves of the process, in order to establish the absolute mass level after the system returns to the baseline setting state, the measurement and control system extracts the corresponding equivalent reset mass based on the steady-state reset extraction time determined in step two. Considering the high-frequency mechanical vibration and electromagnetic interference noise present in the 70 MPa pipeline network in the industrial site, extracting only the value of a single sampling point at the steady-state reset extraction time is highly susceptible to interference from random sudden deviations, leading to severe drift in subsequent judgment criteria. Therefore, this scheme forcibly introduces a low-pass smoothing mechanism when extracting the equivalent reset mass. The measurement and control system uses the steady-state reset extraction time as the calculation starting point, continuously extracts a preset set of uppercase letter J smoothed sampling points on the time axis, and defines the arithmetic mean of their equivalent masses as the final equivalent reset mass. Its smoothing calculation relationship satisfies the following equation: In the formula, denoted as , where is the equivalent reset mass after the i-th micro-oscillation returns to the steady-state reference; J is the preset number of smooth sampling points; x is the discrete step rate variable summed; and h is the uniform sampling period of the measurement and control data acquisition system.

[0043] The core of the above time normalization and smoothing calculation lies in eliminating the time integral differences caused by different control command lengths, thus reducing the rate of change of mass during the half-stroke of the upper swing. With the rate of change of mass at halfway point of the hem It is given a rigorous and equal premise for comparison. The comparison here does not pursue absolute numerical equality, but rather provides a generalized asymmetric state envelope containing complete hysteresis characteristics for the construction of dimensionless directional feature parameters in step four. At the same time, the multi-point averaging reset quality extraction logic enhances the numerical stability of the algorithm in harsh industrial environments, ensuring the absolute reliability of subsequent reset trace classification and judgment.

[0044] It should be noted that the selection of the smoothing sampling point number J should be adaptively configured based on the uniform sampling period h and the characteristic spectrum of the high-frequency noise in the field. The engineering physics principle for selection is that the product of the smoothing sampling point number J and the uniform sampling period h is the smoothing time window. This time window must cover at least one complete high-frequency noise oscillation cycle, but must be strictly less than the time constant of macroscopic baseline drift caused by the fluid re-emerging from natural leakage. Typically, the smoothing sampling point number J can be set to a positive integer range of five to twenty to achieve the best algorithmic balance between filtering out random burst deviations and preserving true thermodynamic residence characteristics.

[0045] Specifically, after executing the equivalent state transformation and time-series extraction sub-steps, the final dynamic feature output of step three is the rate of change of mass during the upper half-stroke, the rate of change of mass during the lower half-stroke, and the equivalent reset mass for each iteration. These core feature quantities characterizing the evolution of microscopic leakage will be fully transferred to subsequent steps, directly driving step four to construct the self-sealing reversibility directional feature parameters, which will then serve as the data kernel for programmatic objective classification and determination.

[0046] Step Four: Following the in-depth data processing in Step Three, the test system has successfully eliminated system errors caused by different control command integration times and high-frequency thermodynamic noise in the high-pressure fluid system. It has extracted the pure, time-normalized mass change rates of the upper and lower swing halves and obtained the smoothed equivalent reset mass. However, these isolated numerical values ​​are significantly affected by the overall wear of the target valve and the baseline test pressure. If a conventional judgment is made by directly setting an absolute leakage threshold, misjudgments are easily generated on valves of different diameters or with different initial damage states. To completely eliminate the dependence on absolute values, Step Four introduces a relative ratio and statistical consistency test. By constructing a dimensionless relative characteristic parameter insensitive to background leakage and combining it with the macroscopic repeatability under multiple cyclic excitations, this step can isolate the phenomenon caused by reversible particle rearrangement from complex background leakage, providing a decisive logical criterion for final fault diagnosis. The specific process is as follows: Construct dimensionless directional characteristic parameters and perform multi-level physical law consistency determination.

[0047] Specifically, the logical operations and feature determination parameters involved in this step include a numerical stability term, a self-blocking reversibility directional feature parameter, a minimum significance dead zone threshold, a first-level directionality determination threshold, and a second-level cluster separation degree. The numerical stability term is a preset minimum positive number used to prevent division by zero overflow, denoted as e. The self-blocking reversibility directional feature parameter characterizes the directional shift of the leakage rate within a single micro-oscillation cycle; for the i-th micro-oscillation iteration, this parameter is denoted as... The minimum significance dead zone threshold, denoted as d, is used to filter out background random noise near the zero point. The first-level directionality determination threshold, denoted as p, is used to evaluate the consistency probability of polarity across multiple iterations. The second-level cluster separation degree, denoted as g, is used to quantify the bistate discreteness of the reset quality distribution. This step specifically includes the following sub-steps.

[0048] S401, constructing a directional characteristic parameter that eliminates the absolute dimension based on the half-range mass change rate.

[0049] Specifically, for the i-th micro-oscillation iteration, the measurement and control system extracts the mass change rate of the upper half-stroke and the mass change rate of the lower half-stroke from the output of step three, and takes their absolute values ​​respectively. Then, the difference between the absolute values ​​of the upper and lower swing characteristics is calculated as the numerator; the sum of the absolute values ​​of the upper and lower swing characteristics and the numerical stability term e is calculated as the denominator. Dividing the numerator by the denominator yields the self-blocking reversibility direction characteristic parameter of the i-th micro-oscillation. The computational relationship satisfies the following equation: In the formula, Let be the directional characteristic parameter of self-blocking reversibility obtained from the i-th micro-oscillation iteration, which is a dimensionless parameter and its value range is strictly defined between negative one and positive one; The mass change rate of the i-th micro-oscillation during the upper half of the swing extracted in step three is expressed in kilograms per second. denoted as , which is the rate of change of mass during the half-stroke of the i-th micro-oscillation extracted in step three, in kilograms per second; e is a preset numerical stability term, in kilograms per second, whose magnitude is usually set to the equivalent mass loss rate corresponding to the minimum resolution of the system's analog-to-digital converter, in order to maintain the stability of mathematical calculations when the fluid is absolutely still, i.e., when both the numerator and denominator are close to zero.

[0050] The core advantage of this step lies in constructing a dimensionless parameter that only reflects the relative changes in the geometry of the microchannel. If the fluid channel morphology inside the target valve is stable, i.e., there is no particle self-blocking phenomenon, the leakage characteristics of the upper and lower swing halves will be highly symmetrical, and the directional characteristic parameter... It will approach zero infinitely; conversely, if there are foreign particles at the sealing interface, the change in fluid flow direction and the alternation of pressure difference will cause path-dependent micro-displacement of the particles, making the loss in a certain direction significantly larger, and thus leading to changes in the directional characteristic parameters. It exhibits a significant positive or negative polarity deviating from zero. This characteristic parameter completely decouples the interference between the basic test pressure and the absolute damage size of the valve, and is the mathematical core of this scheme that enables the identification of extremely weak faults.

[0051] S402, performs a progressive multi-level consistency determination based on the feature sequence and the reset sequence.

[0052] Specifically, the measurement and control system, based on the total number of micro-oscillation executions N set in step two, fully collects the data generated in all iteration rounds, and performs a progressive first-level symbol consistency determination and a second-level reset consistency determination.

[0053] In the first-level symbol consistency determination, to prevent polarity misjudgment caused by system background noise, the measurement and control system first checks the direction characteristic parameter. The absolute value is compared with the dimensionless minimum significance dead zone threshold d. Due to the directional feature parameter... For a normalized dimensionless value, the minimum significance dead zone threshold d is typically calibrated to a range of 0.05 to 0.15 in engineering applications. Its specific value can be adaptively compensated based on the inherent fluid resistance asymmetry of the control pipeline. If its absolute value is less than the dead zone threshold d, the oscillation is considered to have no significant directional characteristic. Only when its absolute value is greater than or equal to the dead zone threshold d is the system included in the valid polarity sample and its mathematical sign (positive or negative) extracted. The system then counts the number of parameters with the same mathematical sign in all valid polarity samples. If the proportion of a single absolute dominant symbol exceeds the preset first-level directional judgment threshold p (the recommended value of this judgment threshold p is 70% to 90%, and the specific value is adaptively adjusted according to the scale of the total number of micro-oscillations, for example, 80%), the system will draw and output a preliminary judgment conclusion that there is a directional system deviation; conversely, if positive and negative symbols appear randomly or a large number of samples fall within the dead zone threshold, it indicates that the asymmetry is caused by random turbulence of the fluid or that the system is completely sealed, and the system will draw a preliminary judgment conclusion that there is no directional feature.

[0054] In the second-level reset consistency determination, the system checks the equivalent reset quality sequence output in step three, which is from... arrive Since the reversible self-sealing phenomenon physically corresponds to two macroscopic mechanical states of stagnation—blockage and unobstructed flow—the measurement and control system uses a one-dimensional data clustering algorithm, such as K-means clustering, to extract the center points of the sequence. If the clustering algorithm identifies two cluster centers with a clear distribution gap in the sequence, and the absolute difference between the values ​​of the two centers is greater than the preset second-level cluster separation degree g, then a high-level judgment conclusion is drawn that two types of residual states of reset exist; if all equivalent reset masses are randomly scattered or completely clustered near a single center, then a high-level judgment conclusion is drawn that a single reset state exists. To prevent misjudging the background noise as a two-state distribution, the second-level cluster separation degree g must be set strictly greater than the natural fluctuation range or three times the standard deviation of the equivalent reset mass of the system in an absolutely static state, with units in kilograms. In actual engineering calibration, the threshold of this cluster separation degree g can also be directly set to 5% to 10% of the maximum allowable equivalent mass loss rate of the target valve under the current test pressure.

[0055] Specifically, the aforementioned progressive multi-level judgment mechanism serves to ensure that leaks caused by particulate self-sealing are not one-off, random, sudden events, but rather repetitive phenomena with mechanical path dependence. The first level, based on dead-zone immunity and sign consistency statistics, effectively filters out occasional asymmetries caused by single valve water hammer vibrations or instrument background noise. The second level, based on clustering algorithms for objective mathematical identification of the dual-state characteristics of equivalent reset mass, fully maps the deductive process of the movement of the two ends of foreign particles from the perspective of static residence. The two judgment levels are independent yet mutually reinforcing, ensuring that the algorithm has extremely high fault tolerance and zero false alarm rate under any harsh testing conditions.

[0056] Specifically, after executing the aforementioned feature construction and progressive consistency determination steps, step four ultimately outputs a logic state flag indicating whether there is a directional system deviation, and a logic state flag indicating whether there are two types of residual states: reset and reversibility. The output dynamic dataset is a complete sequence of self-blocking reversibility directional feature parameters. The aforementioned logic flag conclusions and feature sequences will serve as a highly condensed feature state matrix, directly driving step five to qualitatively declare the health status of the device under test and generate subsequent intervention strategies.

[0057] Step 5: After the in-depth processing in the preceding steps, the test system has successfully shielded the 70 MPa ultra-high pressure fluid system from various common-mode interferences introduced by environmental heat transfer, fluid compression nonlinearity, sensor background noise, and control timing asymmetry. It also successfully output a highly condensed logic status flag in Step 4. Isolated features often have multiple solutions. Even if conventional leak detection methods capture changes in leakage, they cannot qualitatively distinguish whether it is a stable attenuation caused by normal wear of the sealing surface or intermittent self-sealing caused by impurity particles. To achieve a definitive diagnosis, the core technical logic of Step 5 lies in constructing a rigorous multi-dimensional conditional Boolean logic gate. By fusing the directional deviation conclusion under dynamic disturbances with the dual-state residence conclusion under static reset, this step will complete the final qualitative assessment of the valve's microscopic failure mechanism and automatically generate traceable audit evidence, thereby triggering safety isolation and maintenance intervention in the industrial field, realizing a complete process from state perception to control, as detailed below: Based on the multi-level judgment results, the final qualitative conclusion is output and the system closed-loop operation is executed.

[0058] Specifically, the logic parameters and timing control parameters involved in this step follow those in the preceding steps, including the logic state flags for directional system deviation, the logic state flags for resetting the two types of residual states, the equivalent volume of the closed cavity, the cavity pressure measurement sequence, the self-sealing reversibility directional characteristic parameter sequence, the mass change rate of the upper half-stroke, the mass change rate of the lower half-stroke, the equivalent reset mass, the total number of micro-oscillation executions, and the total equivalent mass loss rate. The total equivalent mass loss rate is used to characterize the macroscopic decay rate within the global test cycle, denoted as... This step specifically includes the following sub-steps.

[0059] S501 integrates dynamic polarity characteristics and static resident flags to perform Boolean logic AND operations to output qualitative diagnostic conclusions.

[0060] Specifically, the main control unit of the measurement and control system reads the two independent logical state flags output in step four in parallel: the logical state flag of the directional system deviation and the logical state flag of the two residual states. It then performs a strict logical AND operation.

[0061] When both of the aforementioned logic state flags are determined to be true, meaning the system has determined in step four that both directional system deviation and reset residual states exist, the measurement and control system outputs the final qualitative conclusion of self-sealing intermittent seal failure. The calculation principle lies in the fact that the systematic polarity asymmetry appearing under dynamic micro-oscillation excitation proves the existence of movable particles at the sealing interface, and that these particles exhibit a path dependence sensitive to the direction of fluid pressure difference. Furthermore, the two significantly different mass residual levels appearing after steady-state reset statically confirm that the particles not only move but can also achieve stable retention at both the blocked and unobstructed mechanical extreme positions. This tight interlocking of dynamic and static characteristics constitutes the necessary and sufficient condition for the self-sealing phenomenon of microchannels.

[0062] When the combined condition of the above two logic state flags is not met (i.e., either flag is false or both are false), the measurement and control system outputs a preliminary qualitative conclusion that the system is in a non-self-sealing sealing state. Under this conclusion, the measurement and control system automatically calls the conventional absolute leakage rate threshold comparison algorithm. The measurement and control system extracts the start time of the first micro-oscillation upward swing command. The corresponding equivalent mass, and the steady-state reset extraction time of the Nth micro-oscillation. The corresponding equivalent mass. Calculate the absolute value of the difference between the two equivalent masses, and divide it by the time span of this period. and The difference is used to calculate the total equivalent quality loss rate. The computational relationship satisfies the following equation: In the formula, The total equivalent quality loss rate during the global testing cycle, expressed in kilograms per second. The equivalent mass corresponding to the steady-state reset extraction moment of the Nth micro-oscillation; The equivalent mass corresponding to the start of the first micro-oscillation upward swing command; This refers to the steady-state reset extraction time of the Nth micro-oscillation extracted in step two. This is the start time of the first micro-oscillation upward swing command extracted in step two.

[0063] If the total equivalent quality loss rate If the leakage rate exceeds the preset safety warning threshold, the system further classifies it as a stable irreversible leak caused by macroscopic scratches or severe wear on the sealing surface; if the total equivalent mass loss rate... If the leakage rate is less than the safety warning threshold, the valve is ultimately determined to be in a completely leak-free and healthy state. This safety warning threshold is set based on the equivalent mass loss rate derived from the maximum permissible leakage rate corresponding to the industry standard for the valve's diameter and pressure rating. For example, referring to the industry standard for hydrogen valves in hydrogen refueling stations, the permissible leakage volumetric flow rate under standard test conditions is rigorously converted into the absolute mass loss rate threshold under the current temperature and pressure conditions using the real gas state equation, and this threshold is used as the basis for judgment.

[0064] The core of this step lies in replacing the traditional judgment method that relies solely on alarms triggered by excessive absolute leakage values, and establishing a causal logic judgment chain based on microscopic failure mechanisms. Through a mechanism combining multidimensional cross-validation and backtracking comparison, the diagnostic algorithm is endowed with extremely strong resistance to false alarms, and can clearly distinguish it from intermittent leakage caused by self-sealing even in the presence of severe background leakage.

[0065] S502 generates a timestamp test record containing complete features and triggers safety intervention actions in the industrial field.

[0066] Specifically, after completing the qualitative diagnosis of S501, the measurement and control system immediately packages and encapsulates the constraint boundary of this test, namely the equivalent volume V of the closed cavity and the cavity pressure measurement sequence P(t), the total number of micro-oscillation executions N, the mass change rate of the upper half swing and the mass change rate of the lower half swing extracted in step three, as well as the equivalent reset mass, the complete self-sealing reversibility direction characteristic parameter sequence extracted in step four, and the final diagnostic conclusion, according to a unified absolute timestamp, to generate an unalterable leak audit record and transfer it to the database for long-term storage.

[0067] Based on this, the measurement and control system executes an automated hardware closed loop according to the qualitative conclusions output. When the final conclusion is that the self-sealing intermittent seal has failed or there is a stable irreversible leak, the measurement and control system actively sends a safety intervention command to the field data acquisition and monitoring control system or distributed control system of the hydrogen refueling station via the industrial communication bus. This intervention command includes automatically generating a maintenance and cleaning work order or a replacement work order for the valve under test, and triggering an emergency safety isolation procedure, such as interlocking and closing the main control safety shut-off valve upstream of the valve under test and forcibly opening the pipeline venting branch.

[0068] Specifically, this step directly transforms the complex edge computing results into a safety barrier for the high-pressure hydrogen pipeline system. The complete packaging and encapsulation of leak detection and audit characteristic variables ensures extremely high traceability and verification capability for every fault alarm throughout its entire lifecycle; while automatically triggered safety isolation actions eliminate the time delays caused by manual intervention, fundamentally preventing the potential leakage of 70 MPa high-pressure flammable fluid in subsequent operation due to valve self-sealing failure.

[0069] Specifically, after executing the above-mentioned logical fusion judgment and system intervention steps, step five finally outputs a clear qualitative classification conclusion on the valve's health status, and based on this conclusion, generates a complete test audit record of the data chain, while triggering security isolation and operation and maintenance closed loop. Thus, this solution achieves fully automated and high-precision detection and verification of the reversible intermittent self-closing phenomenon of high-pressure fluid valves, from boundary construction, controlled excitation, feature extraction, dimensionless judgment to closed-loop handling.

[0070] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An automatic detection and judgment method for valve sealing failure, characterized in that, include: Establish a closed volume cavity under shutdown conditions and obtain the equivalent volume, and simultaneously acquire pressure measurement sequences and temperature measurement sequences; A micro-oscillation excitation is applied using an instruction-level symmetric strategy, and the segmented interval timestamps and steady-state reset extraction times are extracted. The dynamic compressibility factor sequence is obtained, and the pressure measurement sequence and temperature measurement sequence are transformed into an equivalent mass sequence based on the real gas equation of state, combined with the equivalent volume and the dynamic compressibility factor sequence. The rate of change of mass during the upper half of the swing and the rate of change of mass during the lower half of the swing are calculated based on the segmented interval timestamps and the equivalent mass sequence. The equivalent reset mass is extracted based on the steady-state reset extraction time and the equivalent mass sequence. Based on the mass change rate of the upper half-stroke and the mass change rate of the lower half-stroke, a dimensionless self-blocking reversible directional characteristic parameter is constructed, and a first-level sign consistency determination based on polarity statistics is performed to output the logical state flag bit of the directional system deviation. Perform a second-level reset consistency determination based on data clustering on the equivalent reset quality to output the logical status flag bits of the two types of residual reset states; The logical state flags of the integrated directional system deviation and the two types of residual states (reset) are subjected to a logical AND operation to output a qualitative diagnostic conclusion.

2. The automatic detection and judgment method for valve sealing failure according to claim 1, characterized in that, include: After establishing a closed volume cavity under shutdown conditions and obtaining the equivalent volume, the absolute value of the first derivative of the temperature measurement sequence is monitored. When the absolute value of the first derivative of the temperature measurement sequence is continuously lower than the temperature fluctuation threshold, the closed volume cavity is determined to have entered a thermodynamic quasi-steady state. The moment corresponding to the thermodynamic quasi-steady state is taken as the effective starting point for applying the micro-oscillation excitation.

3. The automatic detection and judgment method for valve sealing failure according to claim 1, characterized in that, A micro-oscillation excitation is applied using an instruction-level symmetric strategy, including: The pressure regulating control step size in the pressure increasing direction is continuously sent to the pressure regulating actuator until the number of continuous applications in the single half stroke is reached, thus completing the upper half stroke excitation; The pressure control step size in the pressure reduction direction is continuously sent to the pressure regulating actuator until it reaches twice the number of consecutive single half stroke applications, thus completing the lower half stroke excitation; The pressure regulating control step size in the pressure increasing direction is continuously sent to the pressure regulating actuator until the number of consecutive applications in a single half-stroke is reached, thus completing the reset command excitation.

4. The automatic detection and determination method for valve sealing failure according to claim 3, characterized in that, Extract the steady-state reset extraction time, including: After the reset command excitation is completed, monitor the first derivative of the pressure measurement sequence; When the first derivative of the pressure measurement sequence falls back to the threshold range of small background noise and remains there for a preset time, the corresponding moment is extracted as the steady-state reset extraction moment.

5. The automatic detection and judgment method for valve sealing failure according to claim 1, characterized in that, Transforming pressure and temperature measurement sequences into equivalent mass sequences includes: The equivalent mass sequence is calculated point by point by multiplying the pressure measurement sequence by the equivalent volume and dividing by the combined product of the dynamic compressibility factor sequence, the hydrogen-to-gas constant, and the temperature measurement sequence.

6. The automatic detection and determination method for valve sealing failure according to claim 5, characterized in that, The calculation of the mass change rate of the upper half-stroke and the mass change rate of the lower half-stroke based on the segmented interval timestamps and equivalent mass sequences includes: Calculate the difference between the equivalent mass corresponding to the end time of the swing command and the equivalent mass corresponding to the start time of the swing command in the segmented interval timestamp, and divide it by the swing duration to obtain the half-swing mass change rate. Calculate the difference between the equivalent mass corresponding to the end time of the swing command and the equivalent mass corresponding to the start time of the swing command in the segmented interval timestamp, and divide it by the swing duration to obtain the half-stroke mass change rate of the swing. Extracting the equivalent reset quality includes: taking the steady-state reset extraction time as the starting point, continuously extracting the equivalent quality corresponding to the preset smooth sampling points on the time axis and calculating the arithmetic mean, and defining the arithmetic mean as the equivalent reset quality.

7. The automatic detection and determination method for valve sealing failure according to claim 6, characterized in that, Construct dimensionless directional characteristic parameters for self-blocking reversibility, including: The difference between the absolute value of the rate of change of mass during the upper half of the swing and the absolute value of the rate of change of mass during the lower half of the swing is used as the numerator. The absolute value of the rate of change of mass during the upper half of the swing, the absolute value of the rate of change of mass during the lower half of the swing, and the sum of the numerical stability term are used as the denominator. Dividing the numerator by the denominator yields the directional characteristic parameter of self-blocking reversibility.

8. The automatic detection and determination method for valve sealing failure according to claim 7, characterized in that, Performing a first-level sign consistency determination based on polarity statistics includes: The absolute value of the directional characteristic parameter of self-blocking reversibility is compared with the minimum significant dead zone threshold. When the absolute value of the self-blocking reversibility directional feature parameter is greater than or equal to the minimum significance dead zone threshold, the mathematical symbol of the self-blocking reversibility directional feature parameter is extracted and included in the effective polarity sample. The proportion of a single, absolutely dominant symbol in a statistically valid polarity sample; When the proportion of a single absolute dominant symbol exceeds the first-level directional determination threshold, output the logic status flag of the directional system deviation.

9. The automatic detection and determination method for valve sealing failure according to claim 6, characterized in that, Perform a second-level reset consistency assessment based on data clustering on the equivalent reset quality, including: A one-dimensional data clustering algorithm is used to extract the center point of the equivalent reset quality; When the one-dimensional data clustering algorithm identifies two cluster centers and the absolute difference between the values ​​of the two cluster centers is greater than the preset second-level cluster separation degree, the logic state flags of the two residual states are reset.

10. The automatic detection and determination method for valve sealing failure according to claim 1, characterized in that, After performing a logical AND operation on the logic state flags for merging directional system bias and the logic state flags for the two residual states of reset, it also includes: When the result of the logical AND operation is false, calculate the absolute value of the difference between the equivalent mass corresponding to the steady-state reset extraction time of the last micro-oscillation and the equivalent mass corresponding to the start time of the upward swing command of the first micro-oscillation, and divide it by the corresponding time span to obtain the total equivalent mass loss rate. When the total equivalent mass loss rate is greater than the preset safety warning threshold, it is determined to be a stable irreversible leak; when the total equivalent mass loss rate is less than the preset safety warning threshold, it is determined to be in a healthy and intact state. After outputting the qualitative diagnostic conclusion, a leak detection audit record containing the qualitative diagnostic conclusion is generated, and a safety intervention instruction is issued to the distributed control system.

Images