Safety valve on-line check and performance evaluation system

Abstract

The present application relates to the technical field of safety valve testing and monitoring, and discloses a safety valve online calibration and performance evaluation system. The system comprises a field data acquisition module, an edge computing and feature extraction module, a cloud performance evaluation and decision module, and a man-machine interaction module. By synchronously collecting pressure, acoustic emission, visual and process data, multi-dimensional features are extracted, and comprehensive evaluation and decision are made based on dynamic performance baseline, multi-parameter fusion degradation model and system synergy analysis. The present application realizes the leap from single parameter calibration to system level performance prediction and evaluation, and provides a solution for predictive maintenance and full life cycle management of safety valves.

Description

Technical Field

[0001] This invention belongs to the field of safety valve testing and monitoring technology, specifically relating to an online calibration and performance evaluation system for safety valves. Background Technology

[0002] In the field of industrial safety, safety instrumented systems (SAS) are the core line of defense for ensuring the safe operation of critical process facilities (such as petrochemical and energy production). As a key final actuator in a SIS, the reliability of safety valves directly affects the safety integrity level of the entire system. Regular and accurate calibration and evaluation of safety valves are essential technical means to ensure their effective function and prevent catastrophic accidents.

[0003] Online calibration technology for safety valves aims to test and verify the valve's key performance parameters without interrupting the process. Traditional online calibration methods primarily focus on measuring the safety valve's opening and reseating pressures, i.e., verifying its accurate operation at the set pressure points. The core objective of this method is to confirm the immediate accuracy of the valve's mechanical action.

[0004] Current technologies generally treat safety valves as independent mechanical components for verification, which leads to the following problems in their verification logic and evaluation system: The verification process only collects and focuses on single actuation pressure data, completely ignoring the overall performance attributes that the safety valve should possess as an integral part of the safety instrumented system loop. This results in the inability to assess the valve's performance degradation trend over long-term operation, the stability of its response time, and the system reliability of its collaboration with upstream sensors and logic controllers.

[0005] Traditional methods provide a binary "pass / fail" conclusion, lacking quantitative analysis of valve remaining life, potential failure modes, and contributions to the reliability of the entire safety loop. This results in a lack of data support for preventative maintenance strategies in factories, hindering risk-based verification. Therefore, a solution capable of systematic online verification and comprehensive performance evaluation of safety valves is desired. Summary of the Invention

[0006] The purpose of this invention is to provide an online calibration and performance evaluation system for safety valves, in order to resolve the contradiction in the prior art that can only calibrate the opening and reseating pressures of safety valves, but cannot quantitatively evaluate their overall performance, long-term reliability and system synergy as a component of a safety instrumented system loop.

[0007] This invention provides an online calibration and performance evaluation system for safety valves, comprising:

[0008] The field data acquisition module is used to simultaneously acquire real-time process data from multiple heterogeneous sources during the online calibration of safety valves;

[0009] The edge computing and feature extraction module is deployed in the field-side industrial gateway device and is used to perform real-time preprocessing and primary feature extraction on the raw data acquired by the field data acquisition module.

[0010] The cloud-based performance evaluation and decision-making module, as the core analysis engine of the system, receives the preprocessed multi-dimensional feature dataset uploaded by the edge computing and feature extraction module and performs in-depth performance modeling and evaluation.

[0011] The human-computer interaction and report generation module is used to display evaluation results to users and provide interactive interfaces.

[0012] Preferably, the field data acquisition module includes a high dynamic pressure sensor array, a non-contact acoustic emission sensor, a high frame rate industrial vision unit, and a process interface unit.

[0013] The high dynamic pressure sensor array continuously acquires pressure pulsation signals from the inlet pipe of the safety valve.

[0014] The non-contact acoustic emission sensor is installed at a specific location on the outside of the valve body to collect high-frequency acoustic emission signals generated throughout the valve's operation.

[0015] The high frame rate industrial vision unit captures images of the valve stem's displacement trajectory.

[0016] The process interface unit acquires in real time the upstream pressure transmitter readings associated with the safety valve, the output status signals of the logic controller, and the temperature and flow data of the process medium from the plant's process control system via the industrial bus protocol.

[0017] Preferably, the edge computing and feature extraction module includes a signal synchronization and time-scale alignment unit, a time-domain and frequency-domain feature extraction unit, and a visual feature parsing unit;

[0018] The signal synchronization and time-stamp alignment unit receives data streams from the high dynamic pressure sensor, acoustic emission sensor, industrial vision unit, and process interface unit, and unifies all data streams to the same millisecond-level precision time base based on hardware timestamps and software interpolation algorithms.

[0019] The time-domain and frequency-domain feature extraction unit processes the synchronized pressure signal and acoustic emission signal. For the pressure signal, it extracts the instantaneous value of the starting pressure, the maximum value of the pressure rise rate, the average value of the steady-state emission pressure, the instantaneous value of the reseating pressure, and the pressure drop rate. It also performs a fast Fourier transform on the pressure signal to extract its main frequency energy and the amplitude of specific harmonic components. For the acoustic emission signal, it extracts the event count rate, absolute energy, ringing count, amplitude distribution, and the energy entropy of each sub-band obtained by wavelet packet decomposition.

[0020] The visual feature analysis unit performs subpixel-level edge detection and tracking on the valve stem displacement image sequence, calculates the displacement-time curve of the valve stem from rest to full opening, and extracts the average speed, maximum acceleration and hysteresis time of the valve stem from it.

[0021] Preferably, the cloud-based performance evaluation and decision-making module includes a dynamic performance baseline library, a multi-parameter fusion degradation model, a system synergy analysis unit, and a risk evaluation and decision-making unit;

[0022] The dynamic performance baseline library stores the baseline performance feature vector of each controlled safety valve established by this system after initial commissioning or the last offline overhaul.

[0023] The multi-parameter fusion degradation model is used to quantitatively evaluate the performance degradation degree of the safety valve. The multi-parameter fusion degradation model first calculates the Mahalanobis distance between the feature vector extracted in the current verification cycle and the corresponding benchmark vector in the dynamic performance baseline library as a comprehensive deviation index. Then, time series analysis is introduced to arrange the comprehensive deviation index of each verification in chronological order. The exponential weighted moving average algorithm is used to fit its changing trend to calculate the degradation slope that characterizes the performance degradation rate.

[0024] The system coordination analysis unit is used to evaluate the response coordination of the safety valve in the entire safety instrument loop. The system coordination analysis unit analyzes the time difference between the upstream pressure transmitter reading obtained from the process interface unit being greater than the set threshold and the visual feature analysis unit detecting that the valve stem has started to move as the system response delay. At the same time, it analyzes the time difference between the logic controller outputting the action command and the valve actually starting to open as the valve execution delay, and evaluates the stability and trend of these delay times by comparing them with historical data.

[0025] The risk assessment and decision-making unit makes a comprehensive decision based on the comprehensive deviation and degradation slope output by the multi-parameter fusion degradation model and the delay time stability index output by the system synergy analysis unit. The risk assessment and decision-making unit presets multiple decision thresholds. When the comprehensive deviation is greater than the first threshold but the degradation slope is not greater than the second threshold, an early warning signal is generated and it is recommended to shorten the verification cycle. When the comprehensive deviation and degradation slope are both greater than their respective thresholds, a performance degradation alarm is generated and it is recommended to arrange preventive maintenance. When the fluctuation coefficient of the system response delay or valve execution delay is greater than the third threshold, a system synergy alarm is generated.

[0026] Preferably, the human-computer interaction and report generation module includes a multi-dimensional data visualization interface, a structured report generator, and a maintenance work order interface;

[0027] The multi-dimensional data visualization interface displays the key parameters of the current verification, the comprehensive deviation index, the degradation trend curve, and the system delay analysis graph in real time in the form of a dashboard.

[0028] The structured report generator automatically integrates the raw data, extracted features, model evaluation results, and decision recommendations from each verification to generate a verification and evaluation report that conforms to industry standards.

[0029] The maintenance work order interface automatically pushes equipment information, fault descriptions, and suggested measures to the factory's computerized maintenance management system when the risk assessment and decision-making unit generates maintenance recommendations, thereby triggering the work order creation process.

[0030] Preferably, the calculation process of Mahalanobis distance in the multi-parameter fusion degradation model is as follows:

[0031] First, retrieve the baseline eigenvector and covariance matrix of the current safety valve from the dynamic performance baseline library; then, obtain the eigenvector for this verification.

[0032] Then calculate the difference between the current eigenvector and the reference eigenvector; finally, multiply the difference vector between the eigenvector and the reference eigenvector by the inverse of the reference covariance matrix, and then multiply by the transpose of the difference vector. The square root of the resulting scalar value is the Mahalanobis distance.

[0033] Preferably, the update mechanism of the dynamic performance baseline library is as follows:

[0034] After the safety valve has undergone a confirmed offline disassembly and repair or replacement of key components, a complete online verification process is performed.

[0035] The mean of the feature vectors extracted in this verification is used as the new baseline feature vector, and the covariance of the feature vectors measured repeatedly in this verification is used as the new baseline covariance matrix, thereby updating the dynamic performance baseline of the valve.

[0036] Preferably, the high dynamic pressure sensor array in the field data acquisition module uses multiple pressure sensors, which are distributed and installed at specific intervals along the axis of the safety valve inlet pipe.

[0037] The signal synchronization and time-scale alignment unit performs weighted averaging processing on the pressure signals from multiple pressure sensors to eliminate the influence of local flow field disturbances, and obtains a composite signal representing the average pressure of the valve inlet section.

[0038] Preferably, the system is deployed on an industrial internet platform using a microservices architecture;

[0039] The on-site data acquisition module and the edge computing and feature extraction module constitute an edge microservice;

[0040] The cloud-based performance evaluation and decision-making module and the human-computer interaction and report generation module constitute cloud-based microservices.

[0041] The edge microservices and cloud microservices communicate asynchronously via message queues encrypted with transport layer security protocols.

[0042] Preferably, the visual feature parsing unit uses the Canny edge detection algorithm combined with morphological closing operation to extract the valve stem contour, and uses a phase-correlation-based subpixel registration algorithm to calculate the valve stem displacement-time curve.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0044] 1. This invention overcomes the limitations of traditional online verification, which only focuses on two isolated parameters: opening and reseating pressure. By constructing a hardware system that integrates high-dynamic pressure, acoustic emission, visual images, and synchronous process data acquisition, and designing corresponding multi-dimensional feature extraction algorithms, it achieves a comprehensive digital characterization of the mechanical action characteristics, internal wear state, vibration modes, and system response timing of the safety valve. This provides a data foundation for a deeper understanding of valve performance.

[0045] 2. This invention upgrades the binary judgment of "pass / fail" for a single verification to a quantitative tracking of the continuous performance degradation trend by establishing a dynamic performance baseline library and a multi-parameter fusion degradation model. The Mahalanobis distance comprehensive deviation index can sensitively capture the coordinated anomalies of multiple feature parameters, while the degradation slope calculation based on time series can provide early warning of accelerated performance degradation. Thus, the maintenance strategy shifts from post-correction and periodic prevention to predictive maintenance based on real-time performance status.

[0046] 3. The innovative system coordination analysis unit of this invention evaluates the safety valve within its respective safety instrumented system loop. By accurately measuring and analyzing system response delays and valve execution delays, potential problems in the coordinated operation between the valve and upstream sensors and controllers can be identified, such as signal transmission attenuation, controller logic delays, or valve mechanical jamming tendencies. This represents a fundamental leap from "component verification" to "system-level performance evaluation," and is of critical value in ensuring the safety integrity level of the entire safety loop.

[0047] 4. This invention significantly improves the efficiency and objectivity of verification and evaluation through a fully automated design across the entire chain, from data acquisition, edge processing, cloud-based intelligent analysis to decision output. The automatic generation of structured reports and seamless integration with the maintenance management system form a closed-loop management system from performance monitoring to maintenance execution, greatly reducing the cost of manual intervention and the risk of misjudgment, providing solid technical support for the full lifecycle digital management of industrial assets. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention;

[0049] Figure 2 This is a schematic diagram of the core principle framework of the multi-parameter fusion degradation model and system synergy analysis in this invention;

[0050] Figure 3 This is a flowchart illustrating the main stages of on-site data acquisition and edge feature extraction in this invention.

[0051] Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow between the edge side and the cloud in this invention. Detailed Implementation

[0052] Example 1: Reference Figures 1 to 4 The safety valve online verification and performance evaluation system of the present invention consists of four main parts: a field data acquisition module, an edge computing and feature extraction module, a cloud performance evaluation and decision-making module, and a human-computer interaction and report generation module. The modules achieve seamless data flow and functional collaboration through industrial-grade communication links.

[0053] The entire system is deployed on an industrial internet platform using a microservice architecture. The field data acquisition module, along with the edge computing and feature extraction module, constitutes an edge microservice, deployed in an industrial gateway device close to the safety valve being tested. The cloud performance evaluation and decision-making module, along with the human-machine interaction and report generation module, constitutes a cloud microservice, deployed on an enterprise private or public cloud platform. Asynchronous and reliable data communication between the edge and cloud microservices is achieved through message queues encrypted with transport layer security protocols, ensuring the complete upload of critical feature data and the reliable issuance of commands even under weak network conditions or intermittent network outages.

[0054] The field data acquisition module is the foundation of the entire system's sensing capability; its structural layout and signal acquisition logic are shown in the attached figure. Figure 3 As shown, the field data acquisition module comprises four core sub-units: a high dynamic pressure sensor array, a non-contact acoustic emission sensor, a high frame rate industrial vision unit, and a process interface unit.

[0055] The high-dynamic pressure sensor array consists of three piezoelectric pressure sensors, evenly spaced at 50 mm intervals along the axis of the safety valve inlet pipe, to capture pressure pulsation characteristics at different locations on the pipe cross-section. Each sensor has a sampling frequency greater than 1 kHz, a measurement range covering 0 to 10 MPa, and an accuracy class of 0.25. The output signal is connected to the analog input channel of the edge gateway via a shielded twisted-pair cable. The non-contact acoustic emission sensor uses a resonant piezoelectric ceramic probe, installed directly above the spring cavity on the outer wall of the valve body. This location is most sensitive to high-frequency elastic waves generated by friction of the valve disc sealing surface, spring vibration, and micro-leakage.

[0056] The sensor's effective frequency band covers 20 kHz to 1 MHz, and the preamplifier gain is adjustable from 40 dB to 60 dB. The signal is transmitted via coaxial cable to a dedicated acoustic emission acquisition card at the edge gateway. The high frame rate industrial vision unit consists of a high-speed CMOS image sensor, a telecentric optical lens, and an LED strobe light source. It is mounted on the side of the valve stem guide sleeve to ensure that the field of view completely covers the entire stroke of the valve stem from fully closed to fully open. The image acquisition frame rate is set to 200 frames per second, the resolution is greater than 1280×1024 pixels, and the exposure time is fixed at 2 milliseconds to freeze the blurring effect under high-speed motion.

[0057] The process interface unit establishes a real-time communication connection with the plant's process control system through standard industrial bus protocols (such as PROFIBUS-DP or ModbusTCP), periodically reading the instantaneous readings of the upstream pressure transmitter, the digital output status of the safety instrument system logic controller, the temperature value of the process medium (in degrees Celsius), and the mass flow rate value (in kilograms per second). The sampling period is synchronized with the pressure sensor, providing contextual data for subsequent system coordination analysis.

[0058] The edge computing and feature extraction module is deployed inside the industrial gateway, and its internal processing flow is as shown in the attached figure. Figure 3 As shown, it includes three functional stages: a signal synchronization and time-stamp alignment unit, a time-domain and frequency-domain feature extraction unit, and a visual feature parsing unit. The signal synchronization and time-stamp alignment unit first receives three raw pressure signals from the high dynamic pressure sensor array, a single broadband signal from the acoustic emission sensor, image frame sequence metadata (including hardware trigger timestamps) from the industrial vision unit, and periodic process variable packets from the process interface unit.

[0059] All data streams carry nanosecond-level timestamps generated by a hardware clock. The signal synchronization and time-stamp alignment unit first performs a weighted average of the three pressure signals: due to local turbulence disturbances in the pipe flow field, the pressure at a single measuring point may be distorted. Therefore, a synthesis algorithm based on spatial location weights is used, assigning a weight of 0.6 to the middle sensor and 0.2 weights to the sensors at both ends, synthesizing a single signal representing the average pressure at the valve inlet section, denoted as... .

[0060] Subsequently, the signal is synchronized with the time-scale alignment unit. Using the time axis as a reference, the acoustic emission signal, process variables and visual frame timestamps are resampled using a cubic spline interpolation algorithm. All data streams are unified to the same millisecond-level precision time grid, and the time alignment error is controlled within ±0.5 milliseconds to ensure strict correspondence between multi-source data and physical events.

[0061] The time-domain and frequency-domain feature extraction units after synchronization Signals and acoustic emission signals They are processed in parallel, respectively. First, the pressure change rate is calculated using the first-order difference. And locate its peak point to determine the take-off time. and the time to return to one's seat Based on this, the following time-domain features are extracted: instantaneous value of takeoff pressure. Defined as Moment Value; maximum rate of increase in pressure ,Right now Maximum positive value within the 100-millisecond window before takeoff; mean steady-state emission pressure. ,calculate milliseconds to Within milliseconds Value; Instantaneous value of reseating pressure Defined as Moment Value; Pressure drop rate ,Right now The smallest negative value (absolute value) within a 100-millisecond window after the return to the seat.

[0062] Simultaneously, a Fast Fourier Transform was performed on the signal with 2048 sampling points, and a Hanning window was selected as the window function to extract the main frequency energy. (i.e., the energy corresponding to the maximum peak value in the spectrum) and the amplitudes H3, H5, and H7 of the 3rd, 5th, and 7th harmonic components are used as frequency domain features to characterize the nonlinear vibration characteristics during valve operation. For acoustic emission signals... First, low-frequency mechanical noise is removed using a fourth-order Butterworth high-pass filter (cutoff frequency of 20 kHz). Then, an event detection algorithm is used to identify valid acoustic emission events. Based on this, the event count rate is extracted. (Total number of events per unit time), absolute energy (Integral of the square of the signal over the entire time period), ring count (Number of times the signal crosses a preset threshold), amplitude distribution histogram (divided into 10 amplitude intervals at 10 dB intervals, and the proportion of events in each interval is statistically analyzed). In addition, the acoustic emission signal is decomposed into 8-level wavelet packets using the db4 wavelet basis, resulting in 256 sub-band signals. The energy of each sub-band signal is calculated. ( (ranging from 1 to 256), then calculate the total energy. This allows us to obtain the energy percentage of each subband. Finally, calculate the Shannon entropy. As a characteristic of the energy entropy of acoustic emission signals, the lower the value, the more concentrated the energy is in a few frequency bands, which usually corresponds to a specific type of mechanical fault.

[0063] The visual feature parsing unit performs sub-pixel-level processing on the synchronized valve stem image sequence. First, the Canny edge detection algorithm combined with morphological closing operations is used to extract the contour line of the valve stem top in each frame. Then, a phase-correlation-based sub-pixel registration algorithm is used to calculate the displacement vector of the valve stem top between adjacent frames, accumulating to obtain the valve stem displacement-time curve. The valve stem displacement-time curve takes the fully closed position of the valve stem as the zero point, and the fully open stroke as... (Unit: millimeters). From Extract the following kinematic features: average velocity of valve stem , for The moment when the first deviation from 0 is greater than 0.1 mm, for The moment when 0.95L is first reached; maximum acceleration Through the The maximum absolute value after performing a second numerical derivative; action lag time. Defined as the time from the start command issued by the process control system (obtained by the process interface unit) to... The time difference between them. All of the above features are packaged into structured data objects and uploaded to the cloud microservice via a message queue.

[0064] The cloud-based performance evaluation and decision-making module serves as the core intelligent engine of the system, and its internal logical framework is shown in the attached figure. Figure 2 As shown, it includes a dynamic performance baseline library, a multi-parameter fusion degradation model, a system synergy analysis unit, and a risk assessment and decision-making unit.

[0065] The dynamic performance baseline library maintains an independent performance profile for each controlled safety valve, which contains a baseline feature vector. and the benchmark covariance matrix . This is achieved by performing a complete online verification process through this system after the valve is initially put into use or after a confirmed offline disassembly and repair. The arithmetic mean of the feature vectors extracted from five consecutive repeated verifications is then calculated. This is the sample covariance matrix of these five eigenvectors, used to characterize the intrinsic correlation between the various feature parameters. The dynamic performance baseline library supports version management; each update records the update time, operator, and maintenance work order number to ensure traceability.

[0066] A multi-parameter fusion degradation model is used to quantify the performance deviation of the safety valve in the current calibration cycle. Let the feature vector extracted in this calibration be... Its dimensions are (n≥15, covering all the aforementioned time-domain, frequency-domain, and visual features). The multi-parameter fusion degradation model first calculates... and Mahalanobis distance between The calculation formula is as follows:

[0067] ;

[0068] In this formula, For the deviation vector, Let it be its inverse covariance matrix. It is a transpose. The advantage of Mahalanobis distance is that it can automatically normalize each feature and take into account the correlation between features, thereby avoiding the evaluation bias caused by Euclidean distance in multidimensional space due to the dimensions and correlation. The larger the value, the more severely the current performance deviates from the baseline.

[0069] Furthermore, this multi-parameter fusion degradation model maintains historical data. Time series of values , The historical verification count is used. An exponentially weighted moving average algorithm is employed to fit the trend of this sequence, with weighting factors... Set the value to 0.3 and calculate the weighted average sequence. Then, regarding Linear regression was performed on the sequence to obtain the degradation slope. The calculation formula is as follows:

[0070] ;

[0071] To verify the average of the serial numbers, for The mean of the sequence. This indicates a continuous trend of performance degradation; the larger the absolute value, the faster the degradation rate.

[0072] The system compatibility analysis unit focuses on evaluating the timing response characteristics of safety valves within the safety instrumented system loop. This unit obtains upstream pressure transmitter readings from the process interface unit. And set a preset action trigger threshold. (Typically 95% of the safety valve's set pressure). When First greater than Record that moment as Simultaneously, the moment when the valve stem begins to move is obtained from the visual feature analysis unit. System response delay Defined as On the other hand, the timing of the logic controller outputting a high level is obtained from the process interface unit. Then the valve execution is delayed. Defined as The system's collaborative analysis unit continuously tracks the most recent 10 verifications. and Data sequences. For each sequence, calculate its standard deviation. with the mean The ratio of the two, i.e., the volatility coefficient .when If the delay time is deemed unstable, it may indicate systemic problems such as abnormal signal transmission, controller logic delay drift, or valve mechanical jamming.

[0073] The risk assessment and decision-making unit integrates all the above analysis results and performs tiered decision-making. This risk assessment and decision-making unit presets three sets of thresholds: the first threshold... Used for overall deviation (set to 2.0), second threshold Used for degradation slope (set to 0.05 / validation), third threshold Used for volatility coefficient (set to 0.15).

[0074] The decision-making logic is as follows: If and If this occurs, a "performance warning" signal will be generated, suggesting that the next verification cycle be shortened by 50%; if and If this occurs, a "performance degradation alert" will be generated, recommending preventative maintenance within 30 days; if or of If this occurs, a "System Coordination Alert" is generated, suggesting a check of the signal links and controller configuration of the safety instrumented loop. All decision results are accompanied by a confidence score, which is calculated based on a weighted average of feature data completeness, signal-to-noise ratio, and historical consistency.

[0075] The human-computer interaction and report generation module provides users with an intuitive interface for operation and information display. The multi-dimensional data visualization interface is presented in the form of a web dashboard, containing four core views: a real-time parameter view, displayed using digital gauges and trend charts. Key parameters; a comprehensive deviation view, using a radar chart overlaid with the current feature vector and the baseline vector, displayed with color gradients. Value; Degradation trend view, displayed as a line chart showing historical values. Values ​​and EWMA fitted curves, with current values ​​marked. Value; System delay analysis view, showing the last 10 delays as a box plot. and The distribution of the data is displayed, and data points that exceed the stable range are highlighted.

[0076] The structured report generator is automatically triggered after each verification, integrating raw data summaries, feature extraction results, model evaluation metrics, and decision recommendations to generate a verification and evaluation report compliant with API 576 and ISO 4126 standards. The report includes basic equipment information, verification environment parameters, performance characteristic tables, trend analysis charts, and a maintenance recommendation list. The output format supports both Extensible Markup Language (XML) and portable document formats for easy archiving and auditing.

[0077] The maintenance work order interface is integrated with the factory's computerized maintenance management system using a RESTful API. When the risk assessment and decision-making unit generates maintenance recommendations, this interface automatically constructs a JSON data packet containing a unique equipment identifier, a description of the failure mode (such as "spring stiffness decay" or "valve stem guide wear"), recommended actions (such as "replace the spring assembly" or "clean the guide sleeve"), and a priority level. This data packet is then pushed to the maintenance system, triggering a standardized work order creation process and simultaneously returning a work order number for closed-loop tracking.

[0078] The entire system's workflow begins with a planned or triggered online verification task. The field data acquisition module simultaneously captures multi-source data during valve operation; the edge computing and feature extraction module performs data synchronization, noise reduction, and feature engineering locally, uploading only lightweight feature vectors instead of raw large data streams, greatly reducing bandwidth requirements; the cloud performance evaluation and decision-making module performs in-depth analysis based on dynamic baselines and historical trends, outputting quantitative evaluations and intelligent decisions.

[0079] Ultimately, the human-computer interaction and report generation module transforms complex technical conclusions into actionable business information, forming a complete closed loop from perception and analysis to execution. This embodiment fully demonstrates the fundamental leap of this invention from component-level verification to system-level performance evaluation, providing a solid technical foundation for predictive maintenance and safety integrity assurance of industrial safety valves.

[0080] Example 2: Based on Example 1, this example specifically optimizes the performance evaluation of safety valves under high-risk operating conditions, and is particularly suitable for high-temperature, high-pressure, or highly corrosive media environments. In such scenarios, traditional contact sensors are easily affected by environmental interference and may fail. Therefore, this example reconstructs the sensor selection and installation method of the field data acquisition module and adjusts the feature extraction strategy on the edge side accordingly.

[0081] The high-dynamic pressure sensor array no longer employs direct insertion installation, but instead uses non-invasive ultrasonic pressure measurement technology. Specifically, two pairs of ultrasonic transducers are symmetrically installed on the outer wall of the safety valve inlet pipe, forming a transmitter-receiver channel. By measuring the transit time change of ultrasonic waves propagating in the pipe wall, the stress state of the inner wall of the pipe is inverted, and the internal fluid pressure is then calculated. This method avoids the need for openings in high-temperature and high-pressure pipelines, significantly improving system safety and long-term reliability.

[0082] The sampling frequency remains above 1 kHz, but a temperature compensation algorithm has been added to the signal preprocessing stage. The ultrasonic velocity is corrected in real time using the medium temperature value obtained from the process interface unit to eliminate the influence of thermal expansion on the measurement accuracy.

[0083] The installation location of the non-contact acoustic emission sensor has also been optimized. Considering that heat radiation from the valve body surface may interfere with the piezoelectric probe under high-temperature conditions, this embodiment uses a magnetic heat-insulating bracket to isolate the sensor from the valve body with a 5 mm thick aerogel insulation layer. Simultaneously, the preamplifier for the acoustic emission signal is integrated into the sensor body and packaged with high-temperature resistant electronic components to ensure normal operation even at ambient temperatures up to 200 degrees Celsius.

[0084] In the feature extraction stage, the time-domain and frequency-domain feature extraction units have added an adaptive filtering module for thermal noise. This adaptive filtering module first analyzes the background noise power spectrum of the signal during the inactive period and establishes a dynamic noise threshold; during valve operation, it only retains acoustic emission events with a signal-to-noise ratio greater than 6 dB, effectively suppressing false triggering by high-temperature thermal noise.

[0085] The high-frame-rate industrial vision unit has been upgraded to an infrared thermal imaging vision system. This system employs an uncooled microbolometer focal plane array with a spectral response range of 8 to 14 micrometers and a frame rate increased to 300 frames per second to capture the thermal deformation and motion trajectory of the valve stem at high temperatures. The algorithm of the visual feature analysis unit has been adjusted accordingly: it no longer relies on visible light edge detection but instead performs valve stem contour segmentation based on the temperature gradient field of the thermal image.

[0086] By setting a temperature threshold (e.g., a region 50 degrees Celsius above ambient temperature), the hot zone of the valve stem is extracted, and its centroid displacement is calculated as... The alternative is that the kinematic characteristics extracted from this are more likely to reflect the thermodynamic behavior of the valve under real operating conditions, such as the shortening of the stroke caused by thermal expansion or the abnormal acceleration caused by thermal jamming.

[0087] In the cloud-based performance evaluation and decision-making module, the construction of the dynamic performance baseline library also considers the impact of operating temperature. For safety valves operating under high-temperature conditions, its baseline feature vector... It is not a single value, but a value related to the temperature of the medium. function During each verification, the system adjusts the parameters based on the actual measurements. The baseline value is interpolated from the baseline library to obtain the corresponding reference vector, and then the Mahalanobis distance is calculated. This avoids misjudgment caused by fluctuations in operating temperature. In addition, the multi-parameter fusion degradation model introduces a temperature correction factor to normalize the pressure-related features, making them comparable to the ambient temperature reference.

[0088] The system synergy analysis unit also enhances the modeling of temperature delay effects. In high-temperature systems, there is an inherent delay caused by heat conduction and material thermal inertia between the triggering of the upstream pressure transmitter and the actual response of the valve. This embodiment establishes a model by learning from historical data. With medium temperature Empirical relationship between , This is the proportionality coefficient. This is a constant offset. This inherent delay is subtracted from each evaluation, and stability analysis is performed only on the residual delay. This allows cooperative alerts to focus more on genuine system failures rather than normal physical phenomena.

[0089] Through the aforementioned targeted improvements, this embodiment enables the system to maintain high-precision performance evaluation capabilities even under extreme operating conditions, further expanding the application boundaries of the present invention and ensuring its applicability and reliability in key fields such as petrochemicals and nuclear power.

[0090] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0091] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A safety valve online calibration and performance evaluation system, characterized in that, include: The field data acquisition module is used to simultaneously acquire real-time process data from multiple heterogeneous sources during the online calibration of safety valves; The field data acquisition module includes a high dynamic pressure sensor array, a non-contact acoustic emission sensor, a high frame rate industrial vision unit, and a process interface unit. The edge computing and feature extraction module is deployed in the field-side industrial gateway device and is used to perform real-time preprocessing and primary feature extraction on the raw data acquired by the field data acquisition module. The cloud-based performance evaluation and decision-making module, as the core analysis engine of the system, receives the preprocessed multi-dimensional feature dataset uploaded by the edge computing and feature extraction module and performs in-depth performance modeling and evaluation. The human-computer interaction and report generation module is used to display evaluation results to users and provide interactive interfaces; The cloud-based performance evaluation and decision-making module includes a dynamic performance baseline library, a multi-parameter fusion degradation model, a system synergy analysis unit, and a risk assessment and decision-making unit. The dynamic performance baseline library stores the baseline performance feature vector of each controlled safety valve established by this system after initial commissioning or the last offline overhaul. The multi-parameter fusion degradation model is used to quantitatively evaluate the performance degradation degree of the safety valve. The multi-parameter fusion degradation model first calculates the Mahalanobis distance between the feature vector extracted in the current verification cycle and the corresponding benchmark vector in the dynamic performance baseline library as a comprehensive deviation index. Then, time series analysis is introduced to arrange the comprehensive deviation index of each verification in chronological order. The exponential weighted moving average algorithm is used to fit its changing trend to calculate the degradation slope that characterizes the performance degradation rate. The system coordination analysis unit is used to evaluate the response coordination of the safety valve in the entire safety instrument loop. The system coordination analysis unit analyzes the time difference between the upstream pressure transmitter reading obtained from the process interface unit being greater than the set threshold and the visual feature analysis unit detecting that the valve stem has started to move as the system response delay. At the same time, it analyzes the time difference between the logic controller outputting the action command and the valve actually starting to open as the valve execution delay, and evaluates the stability and trend of these delay times by comparing them with historical data. The risk assessment and decision-making unit makes a comprehensive decision based on the comprehensive deviation and degradation slope output by the multi-parameter fusion degradation model and the delay time stability index output by the system synergy analysis unit. This unit presets multiple decision thresholds. When the overall deviation is greater than the first threshold but the degradation slope is not greater than the second threshold, an early warning signal is generated and it is recommended to shorten the verification cycle. When both the overall deviation and the degradation slope are greater than their respective thresholds, a performance degradation alarm is generated and it is recommended to arrange preventive maintenance. When the fluctuation coefficient of the system response delay or valve execution delay is greater than the third threshold, a system coordination alarm is generated.

2. The safety valve online calibration and performance evaluation system according to claim 1, characterized in that, The high dynamic pressure sensor array continuously acquires pressure pulsation signals from the inlet pipe of the safety valve. The non-contact acoustic emission sensor is used to collect high-frequency acoustic emission signals generated throughout the entire valve operation process; The high frame rate industrial vision unit captures images of the valve stem's displacement trajectory. The process interface unit acquires in real time the upstream pressure transmitter readings associated with the safety valve, the output status signals of the logic controller, and the temperature and flow data of the process medium from the plant's process control system via the industrial bus protocol.

3. The safety valve online calibration and performance evaluation system according to claim 2, characterized in that, The edge computing and feature extraction module includes a signal synchronization and time-scale alignment unit, a time-domain and frequency-domain feature extraction unit, and a visual feature parsing unit; The signal synchronization and time-stamp alignment unit receives data streams from the high dynamic pressure sensor, acoustic emission sensor, industrial vision unit, and process interface unit, and unifies all data streams to the same millisecond-level precision time base based on hardware timestamps and software interpolation algorithms. The time-domain and frequency-domain feature extraction unit processes the synchronized pressure signal and acoustic emission signal. For the pressure signal, it extracts the instantaneous value of the starting pressure, the maximum value of the pressure rise rate, the average value of the steady-state emission pressure, the instantaneous value of the reseating pressure, and the pressure drop rate. It also performs a fast Fourier transform on the pressure signal to extract its main frequency energy and the amplitude of the 3rd, 5th, and 7th harmonic components. For the acoustic emission signal, it extracts the event count rate, absolute energy, ringing count, amplitude distribution, and the energy entropy of each sub-band obtained by wavelet packet decomposition. The visual feature analysis unit performs subpixel-level edge detection and tracking on the valve stem displacement image sequence, calculates the displacement-time curve of the valve stem from rest to full opening, and extracts the average speed, maximum acceleration and hysteresis time of the valve stem from it.

4. The safety valve online calibration and performance evaluation system according to claim 3, characterized in that, The human-computer interaction and report generation module includes a multi-dimensional data visualization interface, a structured report generator, and a maintenance work order interface. The multi-dimensional data visualization interface displays the key parameters of the current verification, the comprehensive deviation index, the degradation trend curve, and the system delay analysis graph in real time in the form of a dashboard. The structured report generator automatically integrates the raw data, extracted features, model evaluation results, and decision recommendations from each verification to generate a verification and evaluation report that conforms to industry standards. The maintenance work order interface automatically pushes equipment information, fault descriptions, and suggested measures to the factory's computerized maintenance management system when the risk assessment and decision-making unit generates maintenance recommendations, thereby triggering the work order creation process.

5. The safety valve online calibration and performance evaluation system according to claim 4, characterized in that, The calculation process of Mahalanobis distance in the multi-parameter fusion degradation model is as follows: First, retrieve the baseline eigenvector and covariance matrix of the current safety valve from the dynamic performance baseline library; then, obtain the eigenvector for this verification. Then calculate the difference between the current eigenvector and the reference eigenvector; finally, multiply the difference vector between the eigenvector and the reference eigenvector by the inverse of the reference covariance matrix, and then multiply by the transpose of the difference vector. The square root of the resulting scalar value is the Mahalanobis distance.

6. The safety valve online calibration and performance evaluation system according to claim 5, characterized in that, The update mechanism for the dynamic performance baseline library is as follows: After the safety valve has undergone a confirmed offline disassembly and repair or replacement of key components, a complete online verification process is performed. The mean of the feature vectors extracted in this verification is used as the new baseline feature vector, and the covariance of the feature vectors measured repeatedly in this verification is used as the new baseline covariance matrix, thereby updating the dynamic performance baseline of the valve.

7. The safety valve online calibration and performance evaluation system according to claim 6, characterized in that, The high dynamic pressure sensor array in the field data acquisition module uses multiple pressure sensors, which are distributed and installed at equal intervals along the axis of the safety valve inlet pipe. The signal synchronization and time-scale alignment unit performs weighted averaging processing on the pressure signals from multiple pressure sensors to eliminate the influence of local flow field disturbances, and obtains a composite signal representing the average pressure of the valve inlet section.

8. The safety valve online calibration and performance evaluation system according to claim 7, characterized in that, The system is deployed on an industrial internet platform using a microservices architecture; The on-site data acquisition module and the edge computing and feature extraction module constitute an edge microservice; The cloud-based performance evaluation and decision-making module and the human-computer interaction and report generation module constitute cloud-based microservices. The edge microservices and cloud microservices communicate asynchronously via message queues encrypted with transport layer security protocols.

9. The safety valve online calibration and performance evaluation system according to claim 8, characterized in that, The visual feature analysis unit uses the Canny edge detection algorithm combined with morphological closing operation to extract the valve stem contour, and uses a phase-correlation-based subpixel registration algorithm to calculate the valve stem displacement-time curve.

Images