A method and system for detecting faults throughout the entire life cycle of a safety valve

Abstract

This application relates to the field of safety valve fault monitoring technology, specifically to a method and system for detecting faults throughout the entire lifecycle of safety valves. It aims to address the problems of traditional methods relying on leakage flow monitoring, which cannot detect latent faults such as valve core jamming and spring fatigue, leading to frequent sudden safety accidents. Furthermore, it addresses the issue that power supply systems often employ a single power generation mode, resulting in energy storage module depletion and monitoring interruptions in low-leakage scenarios. This application effectively solves the problems of missing latent fault monitoring and unstable power supply in safety valves. Multi-dimensional data acquisition can detect spring force decay and abnormal valve core movement in advance, preventing faults from escalating into serious leaks. The composite power supply system maintains basic monitoring during the non-leakage phase and rapidly replenishes energy during the leakage phase, ensuring continuous operation throughout the entire lifecycle. Model analysis results can guide maintenance personnel to accurately locate faulty components, reducing unnecessary downtime for maintenance.

Description

Technical Field

[0001] This application relates to the field of safety valve fault monitoring technology, and more specifically, to a method and system for detecting faults throughout the entire life cycle of a safety valve. Background Technology

[0002] Traditional safety valve monitoring technology has significant drawbacks: In terms of fault monitoring, existing solutions such as patent CN119900857A can only identify overt faults such as leakage through flow monitoring, and lack effective detection methods for latent faults such as valve core jamming and spring fatigue. Valve core jamming is caused by hydrogen embrittlement leading to adhesion of the sealing surface and obstruction by tiny impurities, while spring fatigue is manifested as the decrease in elasticity after long-term pressure cycling.

[0003] These types of latent faults often do not initially involve obvious leaks, but gradually evolve into overt faults. For example, spring fatigue can cause the set pressure to drift, initially triggering minor leaks, which eventually develop into catastrophic large-flow discharge accidents. In terms of power supply reliability, existing variable reluctance power generation schemes heavily rely on large-flow discharge conditions. In scenarios without discharge, such as hydrogen storage system pressure maintenance or low-load long-distance pipelines, the energy storage module's power will continuously deplete, causing monitoring interruptions.

[0004] This monitoring blind spot is particularly dangerous in critical scenarios such as high-pressure hydrogen storage and chemical pipelines. Current technologies cannot achieve full-parameter integrated diagnosis, nor can they guarantee continuous power supply, severely restricting the full lifecycle health management of safety valves. To address these issues, existing technologies urgently need improvement. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the aforementioned issues, this invention proposes a method and system for detecting faults throughout the entire lifecycle of a safety valve. This aims to solve the problems of traditional methods relying on leakage flow monitoring, which cannot detect hidden faults such as valve core jamming and spring fatigue, leading to frequent sudden safety accidents, and the fact that power supply systems often adopt a single power generation mode, resulting in the depletion of energy storage modules and monitoring interruptions in low-leakage scenarios.

[0007] (II) Technical Solution

[0008] The present invention discloses a method for detecting faults throughout the entire lifecycle of a safety valve, the technical solution of which is as follows: Multiple sensors installed on the safety valve collect operational data, including valve core force data, valve core displacement data, and safety valve leakage flow data; a multi-source power generation system is used to power the sensors and edge computing module, the multi-source power generation system including a vibration power generation unit and a variable reluctance power generation unit; the operational data is input into a pre-trained LSTM neural network model for fusion analysis to identify the fault type, severity level, and remaining service life; an early warning is triggered based on the identification results; the multi-source power generation system power supply includes: continuous micro-energy replenishment using pipeline vibration through the vibration power generation unit; rapid charging during high-flow discharge using the variable reluctance power generation unit; and priority control logic to ensure that the energy storage module's charge is never lower than 50%.

[0009] Furthermore, this application also proposes that the collected operational data include: collecting the force data of the spring on the valve core through a piezoelectric force sensor installed on the top of the valve core; collecting the displacement-time curve data of the valve core through a laser micro-displacement sensor installed on the side wall of the push rod; and collecting the leakage flow data of the safety valve through a float flow meter and a variable reluctance generator.

[0010] Furthermore, this application also proposes that the fault type identification includes: if a sluggish segment appears in the displacement-time curve, it is determined to be valve core jamming; if the force data is continuously lower than the design threshold, it is determined to be spring fatigue; if the leakage flow exceeds the preset threshold, it is determined to be a leakage fault.

[0011] Furthermore, this application proposes that the input features of the LSTM neural network model include: the mean and fluctuation amplitude of the force data; the start time and smoothness of the displacement data; and the mean and growth rate of the leakage flow data.

[0012] Furthermore, this application also proposes that the early warning includes: local early warning, which uses LED indicator lights and buzzers to provide audible and visual alarms according to the severity level; and cloud early warning, which uses a LoRa wireless communication module to upload fault information to a cloud platform and push it to a mobile terminal.

[0013] Furthermore, this application also proposes to include: filtering and normalizing the collected raw data in order to remove vibration interference.

[0014] Furthermore, this application also proposes an intelligent fault detection system for the entire life cycle of a safety valve, comprising: a sensing module for collecting data on valve core force, displacement, and leakage flow; a power supply module, including a vibration power generation unit and a variable reluctance power generation unit, for supplying power to the system; a processing module for running an LSTM neural network model to perform fusion analysis on the data collected by the sensing module; and an early warning module for issuing local and cloud-based early warnings based on the processing results.

[0015] Furthermore, this application also proposes that the sensing module includes: a piezoelectric force sensor mounted on the top of the valve core; a laser micro-displacement sensor mounted on the side wall of the push rod; a float flowmeter; and a variable reluctance generator.

[0016] Furthermore, this application also proposes a computing device, including at least one processor and a memory, wherein the memory stores instructions that, when executed by the processor, implement the aforementioned safety valve full life cycle fault detection method.

[0017] Furthermore, this application also proposes a non-transitory machine-readable storage medium storing executable instructions that, when executed, cause the machine to perform the aforementioned safety valve lifecycle fault detection method.

[0018] (III) Beneficial Effects

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0020] This invention collects operational data through multiple sensors installed on the safety valve, including valve core force data, valve core displacement data, and safety valve leakage flow data. A multi-source power generation system, comprising a vibration power generation unit and a variable reluctance power generation unit, powers the sensors and edge computing module. The operational data is input into a pre-trained LSTM neural network model for fusion analysis to identify fault types, severity levels, and remaining service life. An early warning is triggered based on the identification results. The multi-source power generation system provides power through: continuous micro-energy replenishment via pipeline vibration using the vibration power generation unit; rapid charging during high-flow discharge using the variable reluctance power generation unit; and priority control logic to ensure the energy storage module's charge is never lower than 50%. This invention can simultaneously detect latent faults such as valve core jamming and spring fatigue, as well as overt leakage faults, and ensures continuous power supply through the multi-source power generation system. This approach solves the monitoring blind spots and power interruption problems inherent in existing technologies. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A schematic diagram of the logical framework structure for a fault detection method throughout the entire life cycle of a safety valve;

[0023] Figure 2 This is a schematic diagram of the normal displacement-time curve;

[0024] Figure 3 This is a schematic diagram of the displacement-time curve of the carabiner;

[0025] Figure 4 This is a schematic diagram of the normal force-time curve.

[0026] Figure 5 This is a schematic diagram of the fatigue force-time curve.

[0027] Figure 6 A schematic diagram of the micro-leakage curve;

[0028] Figure 7 This is a schematic diagram of the large-volume discharge curve. Detailed Implementation

[0029] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application. It should be noted that similar reference numerals and letters in the following drawings indicate similar items; therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0030] In existing technologies, safety valve monitoring systems have long suffered from the industry pain points of limited monitoring dimensions and insufficient power supply stability. Traditional methods rely on leakage flow monitoring, which cannot detect hidden faults such as valve core jamming and spring fatigue, leading to frequent sudden safety accidents. Power supply systems often adopt a single power generation mode, and in low-discharge scenarios, the energy storage module's power is depleted, causing monitoring interruptions. These problems are particularly prominent in the pressure-holding phase of high-pressure hydrogen storage systems, where conventional monitoring methods cannot operate continuously under non-discharge conditions, creating safety blind spots.

[0031] Therefore, as Figures 1-7 As shown, this application proposes a method for fault detection throughout the entire life cycle of a safety valve, including:

[0032] S100. Acquire operational data: Collect operational data through multiple sensors installed on the safety valve. The operational data includes valve core force data, valve core displacement data, and safety valve leakage flow data. Power the sensors and edge computing module using a multi-source power generation system, which includes a vibration power generation unit and a variable reluctance power generation unit.

[0033] S200, Analyze operational data: Input operational data into a pre-trained LSTM neural network model for fusion analysis to identify fault type, severity level, and remaining service life;

[0034] S300, Trigger Warning: Trigger a warning based on the identification result.

[0035] The multi-source power generation system includes continuous micro-energy replenishment through pipeline vibration via a vibration power generation unit; rapid charging during high-flow discharge via a variable reluctance power generation unit; and ensuring that the energy storage module's charge is never lower than 50% via priority control logic.

[0036] Among them, valve core force data refers to the physical quantity reflecting the pressure exerted by the spring on the valve core, which can be collected using a piezoelectric force sensor to detect the decrease in elastic force caused by spring fatigue. Valve core displacement data refers to the positional change information of the valve core during its movement, which can be acquired using a laser micro-displacement sensor to identify abnormal actions caused by valve core jamming.

[0037] Safety valve leakage flow data refers to the flow rate of the medium through the sealing surface. Specifically, it can be measured using a float flowmeter combined with a variable reluctance generator to determine obvious leakage faults. A vibration power generation unit is a device that converts mechanical vibration into electrical energy. It can employ piezoelectric ceramics or electromagnetic induction structures to maintain basic power supply under low-discharge conditions.

[0038] A variable reluctance power generation unit refers to a power generation device that utilizes airflow to drive changes in magnetic reluctance. Specifically, it can employ a combination of rotating blades and coils for rapid energy replenishment during large-volume discharges. Priority control logic refers to the strategy for managing the charging and discharging of the energy storage module. Specifically, it can use a threshold triggering circuit to prioritize vibration-based power generation to maintain a minimum charge level.

[0039] Specifically, this solution constructs a fault evolution monitoring chain using three types of sensors. When the spring fatigues, the force sensor captures abnormal data that is consistently below the design value; valve core jamming is identified through the sluggish section in the displacement curve; and leakage flow data shows a significant increase in the later stages of the fault.

[0040] The LSTM model performs time-series analysis on characteristics such as mean force, displacement smoothness, and flow rate growth rate to establish the correlation between early faults and later leakage. The power supply system automatically switches energy supply modes according to operating conditions: during normal pressure maintenance, vibration-generated power sustains sensor operation; during large-flow discharge, magnetoresistive power provides rapid charging. Priority control ensures that the energy storage module's charge does not fall below a critical value, preventing monitoring interruptions.

[0041] Traditional solutions only monitor leakage flow and rely on a single power supply method, failing to identify early mechanical faults and enabling power outages in low-leakage scenarios. This solution achieves early warning of hidden faults through multi-dimensional data correlation analysis, with a composite power supply system covering energy needs across all operating conditions. Existing threshold alarm mechanisms cannot determine the root cause of faults, while the LSTM model in this solution can simultaneously output fault type and remaining lifetime prediction.

[0042] Specifically, latent faults refer to potential defects in safety valves that do not directly manifest as obvious abnormalities such as leakage or pipe bursts during operation, but will gradually deteriorate and lead to obvious faults. Their core characteristic is "no immediate visible harm, but there is a risk of progressive failure". In contrast, obvious faults, such as leakage, manifest as direct functional failure or safety hazards.

[0043] In this solution, valve core jamming, such as hydrogen embrittlement leading to sealing surface adhesion, pipeline impurities obstructing valve core movement, and spring fatigue, are all latent faults. Initially, there is no leakage, but it can lead to a fault evolution chain from set pressure drift to minor leakage to large-flow discharge. The set pressure specifically refers to the preset pressure at which the safety valve begins to open and discharge under specified operating conditions. For example, the set pressure of the safety valve in a hydrogen storage system is usually 30MPa, which is a core parameter to ensure system safety. Set pressure drift refers to the phenomenon that the actual opening pressure of the safety valve deviates from the design value due to factors such as spring fatigue and valve core jamming. For example, if the design is 30MPa, but after drifting, it opens at 28MPa, which can lead to premature discharge wasting media or delayed discharge causing overpressure and pipe rupture.

[0044] This application effectively solves the problems of missing monitoring of latent faults and unstable power supply in safety valves. Multi-dimensional data acquisition can detect spring force decay and abnormal valve core movement in advance, preventing faults from developing into serious leaks. The composite power supply system maintains basic monitoring during the non-discharge phase and quickly replenishes energy during the discharge phase, ensuring continuous operation throughout its entire life cycle. Model analysis results can guide maintenance personnel to accurately locate faulty components, reducing unnecessary downtime for maintenance.

[0045] This application further proposes to collect the force data of the spring on the valve core by using a piezoelectric force sensor installed on the top of the valve core, to collect the displacement-time curve data of the valve core by using a laser micro-displacement sensor installed on the side wall of the push rod, and to collect the leakage flow data of the safety valve by using a float flowmeter and a variable reluctance generator.

[0046] Among them, piezoelectric force sensors refer to devices that measure dynamic or quasi-static forces based on the piezoelectric effect. Specifically, they can be implemented using highly sensitive piezoelectric crystal sensors, which directly contact the top of the valve core to obtain the actual force applied by the spring. Laser micro-displacement sensors refer to devices that use the principle of laser triangular reflection for non-contact displacement measurement. Specifically, they can be implemented using high-sampling-frequency laser sensors, which are laterally mounted on the outer wall of the push rod to avoid interfering with the movement trajectory of the valve core.

[0047] A float flow meter is a device that measures fluid flow rate by changing the position of a float. Specifically, it can be implemented using a low-resolution, vertically mounted flow meter connected to the outlet pipe of a safety valve to monitor minute leaks. A variable reluctance generator is a device that generates electricity by using airflow to drive a rotor that cuts magnetic field lines. Specifically, it can be implemented using an impeller generator with a variable reluctance structure, integrated into the discharge pipe to simultaneously perform flow monitoring and energy harvesting functions.

[0048] Specifically, a piezoelectric force sensor is positioned on the top of the valve core, in contact with the spring, directly capturing the force changes during spring compression or release, thereby eliminating errors caused by lever transmission in traditional indirect measurement methods. A laser micro-displacement sensor is aligned at an angle with the sidewall of the push rod, calculating the valve core's movement trajectory by measuring the displacement of the reflected light spot; this non-contact measurement method avoids the influence of mechanical wear on the displacement data.

[0049] The float flowmeter and the variable reluctance generator form a complementary mechanism for flow monitoring. The former detects minute leaks by raising and lowering the float under low flow conditions, while the latter reflects the flow intensity by changing the rotor speed during high flow discharge. The data from both are fused to achieve full-range leak monitoring.

[0050] Traditional solutions rely on a single type of sensor to monitor flow parameters, failing to capture latent fault characteristics such as spring force attenuation and abnormal valve core movement. This solution, through the coordinated arrangement of multi-dimensional sensing devices, simultaneously collects mechanical forces, kinematic parameters, and fluid flow data while ensuring the integrity of the valve body structure, forming a complete monitoring system covering both latent and overt faults in the safety valve.

[0051] This application can directly identify spring fatigue-induced elasticity decay, valve core jamming-induced displacement abnormalities, and leakage flow changes, solving the problem of missed detection of hidden faults caused by insufficient data acquisition dimensions in traditional methods.

[0052] The combined design of piezoelectric force sensor and laser micro-displacement sensor enables cross-verification between spring force data and valve core displacement curve, avoiding misjudgment based on a single parameter; the adaptable arrangement of float flowmeter and variable reluctance generator ensures continuous monitoring throughout the entire process from minor leakage to large flow discharge, providing a reliable data foundation for fault diagnosis of safety valves throughout their entire life cycle.

[0053] This application further proposes a fault identification method that determines valve core jamming by the stagnation section of the displacement-time curve, determines spring fatigue by the force data being continuously lower than the design threshold, and determines leakage fault by the leakage flow exceeding a preset threshold.

[0054] The stagnation segment of the displacement-time curve refers to the abnormal stagnation phenomenon that occurs during the movement of the valve core. This can be achieved by collecting displacement data using a laser micro-displacement sensor and analyzing sudden drops in displacement change per unit time, thus detecting early signs of obstructed mechanical movement. The stagnation segment refers to abnormal stagnation or sudden speed drop during the valve core's movement. The quantitative judgment criteria are: in the displacement data collected by the laser micro-displacement sensor, a displacement stagnation ≥0.2s or a sudden drop in displacement change per unit time ≥50% occurs. For example, if the normal upward speed of the valve core is 0.5mm / s, and it suddenly drops to ≤0.25mm / s, it corresponds to obstructed movement of the valve core due to impurities or hydrogen embrittlement.

[0055] The design threshold refers to the minimum allowable value of the spring force. This is achieved by continuously monitoring the force data using a piezoelectric force sensor and comparing it with a preset threshold, used to identify progressive fatigue degradation of the spring material. The preset threshold refers to the graded alarm limit for leakage flow. This is achieved by monitoring minute leaks and large-flow discharges using a float flowmeter and a variable reluctance generator, respectively, used to distinguish different stages of fault development.

[0056] Specifically, when the laser micro-displacement sensor detects that the valve core displacement stagnates or suddenly drops in speed within a specific time period, the system determines that the valve core is stuck. At this point, even if no leakage has occurred, the root cause of the fault can be identified. The piezoelectric force sensor continuously collects spring force data. If the force is lower than the design threshold for several consecutive monitoring cycles, spring fatigue is determined, avoiding misjudgments caused by instantaneous interference.

[0057] A float flowmeter monitors minute leaks, while a variable reluctance generator monitors large leaks. The combination of these two technologies enables tiered early warning systems for leaks. The aforementioned criteria are cross-validated through a collaborative verification mechanism. For example, when a minute leak is detected, force data is used to differentiate the cause of the fault as valve core jamming or spring fatigue, thereby improving identification accuracy.

[0058] Existing technologies rely solely on leakage flow rate as a single criterion for fault identification, failing to distinguish between latent faults such as valve core jamming or spring fatigue, and lacking early warning capabilities during the leak-free phase. This solution directly identifies latent faults by using displacement stagnation segments and force duration thresholds, while combining leakage flow rate-based criterion to cover overt faults, forming a comprehensive monitoring system covering all stages from fault initiation to outbreak.

[0059] This application can identify latent faults such as valve core jamming and spring fatigue in the leak-free stage, avoiding the deterioration of faults and the occurrence of safety accidents; it can accurately distinguish fault types through multi-criteria collaborative verification, reducing false alarms and missed alarms; and it can realize graded early warning for minor leaks and large flow discharges, providing accurate basis for operation and maintenance decisions.

[0060] This application further proposes a method for detecting faults throughout the entire life cycle of a safety valve. By using the mean and fluctuation amplitude of the force data, the opening time and smoothness of the displacement data, and the mean and growth rate of the leakage flow data as input features of the LSTM neural network model, the method can accurately identify the fault type.

[0061] The mean of the force data refers to the arithmetic mean of the data collected by the piezoelectric force sensor over one consecutive hour, specifically using a FUTEK LCM300 sensor, to capture the overall attenuation trend of the spring force. The fluctuation range of the force data refers to the standard deviation of the data over one hour, reflecting dynamic anomalies caused by spring creep or breakage by calculating the data dispersion. The opening time of the displacement data refers to the time it takes for the valve core to move from its initial position to its fully open position, measured using a Keyence IL-300 laser micro-displacement sensor, to identify response lag caused by valve core jamming. The smoothness of the displacement data is calculated using the variance of the first derivative of the displacement-time curve, to detect discontinuities in movement caused by sealing surface adhesion. The mean of the leakage flow data is calculated by averaging the data collected by the float flowmeter over 10 consecutive minutes, to monitor minute leaks. The growth rate of the leakage flow data is calculated by the ratio of the difference between the mean flow rates of adjacent periods to the previous period, to identify potential leaks in advance.

[0062] Specifically, this method first uses wavelet filtering to remove high-frequency vibration noise from the raw data, and then normalizes the data to eliminate dimensional differences. After inputting the preprocessed data into the LSTM model, a decreasing trend in the mean force combined with abnormal fluctuation amplitude can identify spring fatigue; a combination of prolonged opening time and decreased smoothness can indicate valve core jamming; and an abnormal mean leakage flow rate coupled with a sudden change in the growth rate can warn of leak expansion. The time-series processing capability of LSTM can capture the cumulative changes of features over time. For example, a slow decrease in the mean force and a gradual increase in fluctuation amplitude over several consecutive days can detect the trend of spring performance degradation in advance. A multi-feature cross-validation mechanism further eliminates misjudgments based on a single feature. For example, when the force is abnormal but the opening time is normal, false alarms caused by external interference can be ruled out.

[0063] Traditional methods rely solely on a single flow parameter as input feature, failing to capture latent faults without significant leakage, such as valve core jamming and spring fatigue. This approach constructs a multi-dimensional feature system, deeply integrating the temporal dynamic features of force, displacement, and flow parameters with an LSTM model, thus overcoming the blind spots in fault identification caused by the single feature in existing technologies. For example, current techniques cannot distinguish between minute leaks caused by spring fatigue and valve core jamming, while this approach, through combined analysis of force fluctuation amplitude and displacement smoothness, can accurately trace the root cause of the fault.

[0064] This application effectively identifies early-stage, latent faults in safety valves, providing early warnings of spring fatigue or valve core jamming even before significant leakage occurs. Time-series correlation analysis of multi-dimensional features can detect fault evolution trends in advance, preventing sudden safety accidents. A multi-feature cross-validation mechanism significantly improves the model's anti-interference capability and reduces false alarms caused by pipeline vibration or instantaneous flow fluctuations. This solution achieves comprehensive monitoring of spring performance degradation, abnormal valve core movement, and leakage changes, providing accurate fault type judgment criteria for operation and maintenance decisions.

[0065] This application further proposes that the early warning system includes local early warning and cloud-based early warning. Local early warning uses LED indicators and buzzers to provide audible and visual alarms according to the severity level; cloud-based early warning uses a LoRa wireless communication module to upload fault information to a cloud platform and push it to mobile terminals.

[0066] Local early warning refers to a mechanism that directly issues alarms on-site through visual and auditory signals. This can be achieved using a combination of multi-color LEDs and an adjustable-volume buzzer, with different colors and sound patterns distinguishing the severity of the fault. Cloud-based early warning refers to a mechanism that remotely transmits fault data to a management platform. This can be achieved using a low-power wide-area communication module, enabling cross-regional information synchronization via wireless network.

[0067] Specifically, local alerts dynamically adjust the alarm mode based on preset fault severity classification standards. For example, minor faults trigger a low-frequency flashing green indicator light, while severe faults trigger a continuously bright red indicator light accompanied by a high-decibel buzzer. Cloud-based alerts automatically send a data packet containing the fault type, severity, and key parameters to the cloud upon detecting a fault, allowing administrators to receive alert information in real time via mobile devices. Local and cloud-based alerts employ a parallel triggering mechanism; local alerts remain active during network interruptions and automatically retransmit data once the network is restored, providing dual protection.

[0068] Existing early warning systems typically rely on single local audible and visual alarms or independent cloud data transmission, failing to achieve coordinated on-site response and remote monitoring. Conventional local alarms lack fault level matching mechanisms, easily leading to misjudgments or response delays; traditional cloud communication depends on high-power modules, making it difficult to adapt to the power supply requirements of long-term low-load scenarios.

[0069] This application enables synchronized early warning for both on-site operators and remote management personnel, solving the problem of insufficient coverage from a single early warning method. Local tiered alarms ensure differentiated responses to faults of varying urgency, avoiding excessive disruption to normal production; low-power remote communication expands the monitoring range while maintaining power supply reliability, enabling fault information in areas without network access to be transmitted promptly through local mechanisms, and automatically synchronized to the cloud to form a complete record when network access is available.

[0070] This application further proposes to perform filtering and normalization preprocessing on the collected raw data to remove vibration interference.

[0071] Filtering refers to eliminating noise interference in the data through signal processing algorithms. Specifically, wavelet filtering can be used. This algorithm is based on the difference between the frequency range of pipeline vibration and the characteristics of fault signals, filtering out high-frequency mechanical vibration interference while retaining the sluggish signal of valve core jamming. Normalization preprocessing refers to mapping sensor data with different dimensions to a unified numerical range. Specifically, the Min-Max normalization method can be used to convert the force, displacement, and flow data to the 0-1 range according to the sensor range, eliminating the influence of dimensional differences on model analysis.

[0072] Specifically, during the operation of the safety valve, the raw data collected by the piezoelectric force sensor and the laser micro-displacement sensor are subject to high-frequency noise interference due to pipeline vibration. A wavelet filtering algorithm is used to decompose and reconstruct the data, attenuating the 10-100Hz frequency band where vibration interference occurs, while retaining the 50Hz frequency band where valve core jamming characteristics are present. The filtered data is further processed using Min-Max normalization; for example, force data of 0-200N is linearly mapped to the 0-1 interval, displacement data of 0-10mm is mapped to the 0-1 interval, and flow rate data of 0-500L / min is mapped to the 0-1 interval, forming standardized input data for analysis by the LSTM neural network model. Thus, the preprocessed data eliminates instantaneous outliers caused by vibration and balances the weight distribution of data from different sensors.

[0073] Existing technologies typically employ general low-pass filtering algorithms and normalization methods with fixed numerical ranges, which risk filtering out effective signals or compressing feature differences. This solution, however, uses wavelet filtering algorithms to specifically separate vibration interference from fault signals, combined with a normalization parameter design based on the sensor's physical range, thus removing noise while fully preserving the effective information of fault characteristics.

[0074] This application effectively solves the problem of feature distortion in sensor data caused by vibration interference, improves the recognition accuracy of LSTM models for latent faults such as spring fatigue and valve core jamming, avoids model weight shift caused by differences in data scale, and provides a highly reliable data preprocessing foundation for fault detection throughout the entire life cycle of safety valves.

[0075] To more intuitively demonstrate the monitoring effectiveness of this solution, Figures 2 to 7 Simulated curves of key parameters are provided.

[0076] In terms of fault identification, such as Figure 2 As shown, the displacement-time curve of a normal valve core is smooth and without abnormalities. When valve core jamming occurs, as... Figure 3 As shown, the displacement curve will show a distinct "sluggish section," that is, a fluctuating part, which manifests as a sudden drop in motion speed. Figure 4 This demonstrates the normal fluctuation of the spring force around the design value (100N), while Figure 5 This shows the trend that the force applied when the spring is fatigued is consistently lower than the design value and gradually decreases. Figure 6 and Figure 7 The evolution process from small leakage (flow rate <5L / min) to large flow discharge is shown, and the flow rate growth trend can be correlated with changes in force and displacement.

[0077] This application further proposes an intelligent fault detection system for the entire life cycle of a safety valve, including a sensing module, a power supply module, a processing module, and an early warning module.

[0078] The sensing module refers to the device used to collect data on valve core force, displacement, and leakage flow. Specifically, it can be implemented by combining piezoelectric force sensors, laser micro-displacement sensors, float flow meters, and variable reluctance generators. Through multi-dimensional data acquisition, it covers mechanical performance parameters and dynamic behavior characteristics.

[0079] The power supply module refers to an energy supply device that includes a vibration power generation unit and a variable reluctance power generation unit. Specifically, it can capture the vibration energy of the pipeline through the vibration power generation unit composed of piezoelectric ceramic sheets, and combine it with the variable reluctance coil to generate electricity by cutting magnetic field lines during large-flow discharge, forming a dual-mode power supply mechanism of continuous micro-energy replenishment and fast charging. The processing module refers to the computing unit that runs the LSTM neural network model. Specifically, it can use an embedded processor to load a pre-trained model to perform feature extraction and correlation analysis on multi-source time series data.

[0080] The early warning module refers to a combination module that includes a local sound and light alarm device and a wireless communication unit. Specifically, it can provide on-site warnings through a buzzer and LED indicator, and transmit fault information to the cloud platform through a LoRa module.

[0081] Specifically, the piezoelectric force sensor in the sensing module monitors the spring force at the top of the valve core in real time. When the spring's elasticity decreases due to fatigue, the average force data remains consistently below the design threshold. Specifically, "consistently below the design threshold" means that the spring force data collected by the piezoelectric force sensor is less than 10% of the design value for three consecutive monitoring cycles. For example, if the design value is 100N, the measured value is consistently ≤90N, eliminating transient fluctuations caused by pipeline vibration and ensuring the accuracy of spring fatigue assessment.

[0082] The laser micro-displacement sensor captures the displacement curve of the push rod. When the valve core becomes sluggish due to hydrogen embrittlement or impurities, a non-smooth segment appears in the displacement-time curve. The float flowmeter and the variable reluctance generator synchronously detect the leakage flow. When the sealing surface fails, the flow data exceeds the preset range.

[0083] In the power supply module, the vibration power generation unit continuously generates microwatts of electricity under normal pipeline vibration conditions to maintain the system's basic power consumption requirements; the variable reluctance power generation unit utilizes high-speed airflow to drive the impeller and change the reluctance during large-flow discharge from the safety valve, achieving rapid charging at the milliwatt level. The processing module uses an LSTM neural network to perform time-series correlation analysis on features such as the mean force, displacement smoothness, and leakage growth rate to identify early signs of spring fatigue and potential risks of valve core jamming. The early warning module activates local audible and visual alarms based on the fault level and pushes the diagnostic results to the mobile terminals of maintenance personnel via wireless communication.

[0084] Existing solutions monitor leakage faults using only a single flow sensor, failing to acquire valve core mechanical status data, leading to missed detections of latent faults such as spring fatigue and valve core jamming. Furthermore, relying on a single airflow power generation method makes monitoring prone to interruption due to energy storage depletion in scenarios without leakage. This solution, however, integrates multiple sensors to collect mechanical parameters and dynamic behavior data, combined with dual-mode power generation to achieve continuous power supply under all operating conditions, thus resolving the blind spots in latent fault identification and monitoring interruption issues.

[0085] In the pressure-holding phase of a high-pressure hydrogen storage system, this application utilizes vibration-generated power to maintain continuous sensor operation, monitor the spring force decay trend in real time, and provide early warning of spring force failure risk more than 7 days in advance. In long-distance pipeline pressure fluctuation scenarios, it identifies valve core jamming through displacement curve hysteresis characteristics, preventing sudden sealing failures. During large-flow releases, it simultaneously performs fault diagnosis and rapid energy replenishment, ensuring continuous monitoring. The processing module uses an LSTM model to correlate and analyze multi-source data, improving the accuracy of latent fault identification to over 95% and reducing the false alarm rate to below 3%.

[0086] This application further proposes a sensing module including a piezoelectric force sensor mounted on the top of the valve core, a laser micro-displacement sensor mounted on the side wall of the push rod, a float flowmeter, and a variable reluctance generator.

[0087] Among them, piezoelectric force sensors are devices that measure dynamic or static forces based on the piezoelectric effect. They can be made of quartz crystal or ceramic materials and achieve real-time monitoring of spring force by detecting changes in the charge generated by the force applied to the top of the valve core, thus capturing changes in elastic force caused by the decay of spring stiffness. Laser micro-displacement sensors are non-contact displacement detection devices that utilize the principle of laser triangulation. They can employ a combination of a semiconductor laser and a CCD array, calculating the valve core displacement by measuring the positional offset of the reflected light spot on the side wall of the top rod, avoiding measurement errors caused by mechanical contact.

[0088] A float flowmeter is a device that measures fluid flow by the change in the vertical position of a float within a tapered tube. Specifically, it can employ a high-pressure resistant metal float and a magnetically coupled pointer structure for directly measuring the leakage flow of safety valves. A variable reluctance generator is a device that uses the released airflow to drive a rotor to cut magnetic field lines and generate electricity. Specifically, it can employ a multi-pole permanent magnet and a silicon steel sheet laminated stator structure to convert the kinetic energy of the airflow into electrical energy during the release process.

[0089] Specifically, a piezoelectric force sensor is positioned on the spring contact surface at the top of the valve core, directly measuring the axial force applied by the spring. When the spring fatigues and its elasticity decreases, the amplitude of the voltage signal output by the sensor will show a continuous decreasing trend. A laser micro-displacement sensor is aligned with the reflection area on the side wall of the push rod. By recording the abrupt displacement characteristics in the valve core's movement trajectory, it can detect valve core jamming caused by hydrogen embrittlement or impurities. A float flowmeter is installed in the safety valve's discharge outlet pipe, reflecting the leakage flow rate through changes in the float's position. The impeller of the variable reluctance generator is coaxially arranged with the discharge airflow, generating electricity synchronously when a leak occurs.

[0090] Existing solutions typically employ only a single type of sensor for fault detection, such as monitoring for leaks using only a flow meter, without acquiring data on valve core motion and spring mechanical properties. This solution combines a piezoelectric force sensor with a laser micro-displacement sensor to simultaneously monitor two types of latent faults: spring force decay and valve core jamming. Furthermore, by integrating a variable reluctance generator with a float flow meter, it simultaneously performs flow detection and energy recovery when a leak occurs.

[0091] This application can accurately identify spring force attenuation caused by spring fatigue and displacement abnormalities caused by valve core jamming, avoiding missed fault detection due to missing data from a single sensor. The piezoelectric force sensor directly measures the spring force, overcoming the errors caused by indirect calculation; the non-contact measurement method of the laser micro-displacement sensor eliminates the influence of mechanical wear on displacement data; the combined design of the float flowmeter and the variable reluctance generator generates electricity using the venting airflow while detecting leakage flow, ensuring continuous power supply to the monitoring system under venting conditions.

[0092] This application further proposes a computing device, including at least one processor and a memory. The memory stores instructions, which, when executed by the processor, enable a method for detecting faults throughout the entire lifecycle of a safety valve. The method includes: collecting operational data through multiple sensors installed on the safety valve, including valve core force data, valve core displacement data, and safety valve leakage flow data; powering the sensors and edge computing module using a multi-source power generation system, including a vibration power generation unit and a variable reluctance power generation unit; inputting the operational data into a pre-trained LSTM neural network model for fusion analysis to identify the fault type, severity level, and remaining service life; and triggering an early warning based on the identification results.

[0093] A multi-source power generation system refers to a composite energy device that supplies power to equipment through two or more energy sources. In this scheme, it specifically refers to the combination of a vibration power generation unit and a variable reluctance power generation unit.

[0094] The vibration power generation unit, based on electromagnetic induction or piezoelectric effect, converts the mechanical vibration of industrial pipelines during daily operation into electrical energy, adapting to low-load scenarios such as pressure holding without venting, and achieving continuous micro-energy replenishment;

[0095] The variable reluctance power generation unit, based on the principle of reluctance change, uses the airflow during large-flow discharge to drive the rotor to cut magnetic field lines, converting the kinetic energy of the airflow into electrical energy. It is suitable for safety valve opening and discharge scenarios, and can achieve fast charging.

[0096] The two work together through priority control logic. For example, vibration power generation is prioritized to replenish energy, and variable reluctance power generation is activated when the energy is insufficient, ensuring that the energy storage module has ≥50% power and solving the limitations of a single power generation method in the scenario.

[0097] The processor refers to the hardware unit that executes computational logic, specifically a multi-core embedded processor, used to call upon the operational data collected by the sensors and perform inference calculations for the LSTM neural network model. The memory refers to the physical medium that stores executable instructions and model parameters, specifically flash memory chips or solid-state drives, used to embed fault detection algorithms and power supply control strategies. The multi-source power generation system refers to a power supply unit composed of various energy conversion devices, specifically a combination of piezoelectric vibration energy harvesters and reluctance generators, used to supplement energy in pipeline vibration and large-flow discharge scenarios. The LSTM neural network model refers to a deep learning model with temporal feature processing capabilities, specifically a recurrent neural network containing forget gates, input gates, and output gates, used to analyze the dynamic changes in force, displacement, and leakage flow.

[0098] Specifically, a piezoelectric force sensor installed on the safety valve collects real-time force data from the top of the valve core, a laser micro-displacement sensor records the displacement curve of the push rod sidewall, and a float flowmeter monitors leakage flow. After the processor acquires the above data through the bus interface, it calls the preprocessing program stored in memory for filtering and normalization. The processed time-series data is input into an LSTM neural network model. The model analyzes the mean force, displacement opening time, and leakage growth rate to identify fault modes such as spring fatigue, valve core jamming, or seal failure.

[0099] When an anomaly is detected, the processor drives LED indicators and a buzzer via the GPIO interface to issue a local alarm, while simultaneously uploading fault information to the cloud via the wireless communication module. The vibration power generation unit of the multi-source power generation system utilizes the mechanical vibration of the pipeline to generate microwatts of electrical energy, maintaining the basic operation of the sensors and processor. During large-flow discharges, the variable reluctance power generation unit generates milliwatts of electrical energy to rapidly charge the energy storage module. A priority control algorithm monitors the energy storage module's power level in real time, automatically cutting off unnecessary loads when the power level falls below a set threshold, ensuring the continuous operation of core monitoring functions.

[0100] Traditional computing devices rely on a single power source, which can easily lead to monitoring interruptions due to energy depletion under non-discharge conditions. This solution utilizes the collaborative operation of multiple power generation units to maintain power supply using vibration energy in low-load scenarios and to rapidly replenish energy using magnetoresistive power generation in high-flow scenarios. Combined with a priority control strategy, it ensures the operation of critical functions. Existing data processing units typically only support single-type fault diagnosis, while this solution uses an LSTM neural network to fuse and analyze multi-dimensional time-series data, enabling it to simultaneously capture complex fault characteristics such as spring force decay, abnormal valve core movement, and minute leaks, achieving early warning of hidden faults.

[0101] This application addresses the issue of data link disruption caused by power outages in safety valve monitoring systems, ensuring continuous monitoring throughout the entire lifecycle through multi-source energy acquisition and intelligent power consumption management. Simultaneously, based on multi-dimensional sensor data fusion and deep learning algorithms, it accurately identifies latent faults such as valve core jamming and spring fatigue, avoiding the risk of missed detections caused by single flow monitoring. This computing device integrates fault diagnosis algorithms and energy management strategies into an embedded hardware platform, making it suitable for industrial scenarios with low power consumption and high reliability requirements, such as hydrogen storage system pressure maintenance and long-distance pipelines.

[0102] This application further proposes a non-transitory machine-readable storage medium storing executable instructions. When executed, the instructions cause the machine to perform a method for detecting faults throughout the lifecycle of a safety valve. The method includes: powering sensors and edge computing modules through a multi-source power generation system; collecting data on valve core force, displacement, and leakage flow; performing a fusion analysis of fault type, severity level, and remaining lifespan using an LSTM neural network model; and triggering local and cloud-based early warnings based on the analysis results.

[0103] Non-transitory machine-readable storage media refers to physical carriers capable of storing program instructions for a long period, specifically solid-state drives or flash memory chips. These are used to solidify the execution logic of fault detection algorithms, ensuring the stability and repeatability of the detection process. Multi-source power generation systems are composite power supply devices comprising vibration power generation units and variable reluctance power generation units. Specifically, they can be achieved through piezoelectric material vibration power generation and electromagnetic induction power generation. These systems are used to replenish energy in pipeline vibration and high-flow-rate discharge scenarios, maintaining the energy storage module's charge at least 50%. LSTM neural network models are recurrent neural networks with long short-term memory capabilities. They can be trained using time-series data and used to analyze characteristics such as mean force, displacement smoothness, and leakage growth rate to identify latent faults such as spring fatigue and valve core jamming.

[0104] Specifically, when the instructions stored in the storage medium are executed, the vibration power generation unit is first driven to generate micro-energy using pipe vibration. Simultaneously, the variable reluctance power generation unit rapidly charges the module during high-flow discharge. Priority control logic ensures that the energy storage module's charge level is always maintained above half capacity, preventing monitoring failure due to power outages. Subsequently, a piezoelectric force sensor collects spring force data, a laser micro-displacement sensor records the valve core displacement curve, and a float flowmeter monitors the leakage flow. The raw data, after filtering and normalization, is input into an LSTM model for time-series feature analysis. When the model detects continuous force decay, displacement stagnation, or abnormal leakage growth, a local buzzer and LED indicator emit audible and visual alarms according to the fault level. Simultaneously, the diagnostic results are uploaded to a cloud platform via a wireless communication module for remote monitoring.

[0105] Existing technologies rely on a single variable reluctance generator, which cannot maintain power supply under non-discharge conditions, leading to monitoring interruptions. This solution, however, achieves continuous micro-energy replenishment through a vibration-generated unit, combined with energy storage priority control, ensuring power continuity under low-load scenarios. Furthermore, existing technologies do not embed the detection logic into physical storage media, posing a risk of program loss. This solution uses non-volatile storage media to store algorithm instructions, preventing detection logic failure due to system restarts or power outages.

[0106] This application solves the problem of monitoring failure caused by power outage in safety valve monitoring systems under long-term low load or no discharge scenarios. At the same time, by solidifying the detection process through storage medium, it ensures that the fault analysis method is stably executed throughout the entire life cycle, and can continuously identify hidden faults such as valve core jamming and spring fatigue, thus avoiding safety accidents caused by monitoring interruption.

[0107] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A safety valve full life cycle failure detection method, characterized in that, Comprising: Collecting operating data through multiple sensors installed on the safety valve, the operating data including valve core acting force data, valve core displacement data, and safety valve leakage flow data; Powering the sensors and edge computing module using a multi-source power generation system, the multi-source power generation system including a vibration power generation unit and a variable reluctance power generation unit; Inputting the operating data into a pre-trained LSTM neural network model for fusion analysis to identify fault types, severity levels, and remaining service life; Triggering an early warning according to the identification result; The multi-source power generation system power supply includes: Using pipeline vibration to continuously supplement energy through the vibration power generation unit; Using the variable reluctance power generation unit to quickly charge during large flow relief; Using priority control logic to ensure that the energy storage module always has more than 50% of its capacity; The collection of operating data includes: Collecting spring acting force data on the valve core through a piezoelectric force sensor installed on the top of the valve core; Collecting displacement-time curve data of the valve core through a laser micro-displacement sensor installed on the side wall of the top rod; Collecting leakage flow data of the safety valve through a float flowmeter and a variable reluctance power generation device; The identification of fault types includes: If there is a lag section in the displacement-time curve, it is determined that the valve core is stuck; If the acting force data is continuously below the design threshold, it is determined that the spring is fatigued; If the leakage flow exceeds the preset threshold, it is determined that there is a leakage fault.

2. The safety valve full life cycle failure detection method of claim 1, wherein, The input features of the LSTM neural network model include: Mean and fluctuation amplitude of the acting force data; Opening time and smoothness of the displacement data; Mean and growth rate of the leakage flow data.

3. The safety valve full life cycle failure detection method according to claim 1 or 2, characterized in that, The early warning includes: Local early warning, using LED indicator lights and buzzers to sound and light alarms according to the severity level; Cloud early warning, uploading fault information to the cloud platform through the LoRa wireless communication module and pushing it to the mobile terminal.

4. The safety valve full life cycle failure detection method of claim 3, wherein, Also includes: Filtering and normalizing the collected raw data to remove vibration interference.

5. A safety valve full life cycle failure intelligent detection system, characterized in that, The safety valve full life cycle fault intelligent detection system is implemented based on the safety valve full life cycle fault detection method of claim 1, comprising: A perception module for collecting valve core acting force, displacement, and leakage flow data; A power supply module including a vibration power generation unit and a variable reluctance power generation unit for powering the system; A processing module for running an LSTM neural network model to analyze the data collected by the perception module; An early warning module for issuing local and cloud early warnings based on the processing results.

6. The safety valve full life cycle failure intelligent detection system of claim 5, wherein, The perception module includes: A piezoelectric force sensor installed on the top of the valve core; A laser micro-displacement sensor installed on the side wall of the top rod; A float flowmeter and a variable reluctance power generation device.

7. A computing device comprising at least one processor and a memory, the memory storing instructions that, when executed by the processor, implement the safety valve full life cycle fault detection method of any one of claims 1 to 4.

8. A non-transitory machine-readable storage medium storing executable instructions that, when executed, cause a machine to perform the safety valve full life cycle fault detection method of any one of claims 1 to 4.

Images