A pressure gauge performance data monitoring and management system and method based on the Internet of Things

By constructing a forward-oriented evidence storage kernel using the improved MTAD-GAT, the problems of anomaly identification and degradation location in pressure gauge performance management are solved, enabling accurate pressure gauge degradation identification and management, and improving identification accuracy and management efficiency.

CN123115657APending Publication Date: 2026-11-17SHANXI HUANENG ANKE MINING TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611269138.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing pressure gauge performance management methods struggle to identify and trace the turning point of pressure changes during industrial operations, leading to inaccurate anomaly assessments, a lack of continuous data processing mechanisms, and an inability to effectively identify and address pressure gauge degradation.

Method used

An improved MTAD-GAT is used to construct a steering evidence core. Through performance deficit write-off, calibration edge weight transfer and minimum degradation evidence core extraction, pressure gauge degradation location and disposal simulation are realized, and IoT data is used for precise monitoring and management.

Benefits of technology

It enables accurate identification and precise management of pressure gauge degradation, avoids the fragmentation effect of fixed time windows, improves the accuracy of identifying the location and range of pressure gauge degradation, and realizes closed-loop management.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123115657A_ABST
    Figure CN123115657A_ABST
Patent Text Reader

Abstract

This invention discloses a pressure gauge performance data monitoring and management system and method based on the Internet of Things (IoT), comprising the following steps: segmenting the operating range to obtain fragments, forming a turning point evidence core, and extracting the attention edge weight state of the calibration map; restoring the turning point evidence core and continuing propagation during reverse overwrite; rearranging the propagation results according to the actual time sequence and the metered pressure coordinates, and deducting to form a performance deficit; accessing subsequent operating range fragments and transplanting the attention edge weight state of the calibration map to cancel the performance deficit; withdrawing the transplanted states level by level and extracting the minimum degradation evidence core; connecting operating range fragments with the same evidence core to form a performance degradation chain and generating a management task; replaying the propagation after the task is completed, and determining the handling result based on the performance deficit state. This invention relates to the field of pressure instrument monitoring technology, realizing accurate monitoring of pressure gauge performance degradation and improving the efficiency of fault identification and handling.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of pressure instrument monitoring technology, and in particular to a pressure gauge performance data monitoring and management system and method based on the Internet of Things. Background Technology

[0002] Pressure gauges are widely used for pressure measurement in industrial pipelines, pressure vessels, and production equipment. Current pressure gauge performance management primarily employs periodic calibration, manual inspection, remote data acquisition, and threshold alarms. The Internet of Things (IoT) can continuously acquire pressure gauge operating data, and time-series analysis models can identify sudden changes, drifts, and abnormal fluctuations in readings. Existing analysis methods often process continuous data within fixed time windows, determining the pressure gauge's status based on prediction errors, reconstruction errors, or set thresholds.

[0003] Pressure changes during industrial operations are affected by production cycle time, load adjustments, and process switching, making it difficult for single-run data to cover the complete pressure increase and decrease process. Back-range range data scattered across different time periods lack model state connections, failing to retain historical propagation relationships formed by pressure shift positions, and making it difficult to continuously assess retracement deviations, zero-point recovery anomalies, and local range degradation. Existing methods directly convert unclosed anomalies into alarm results, lacking a mechanism for subsequent operational data to verify and cancel historical anomalies; calibration data is mostly used for numerical correction, not for locating model relationships that cause changes in anomaly closure states; after re-inspection, calibration, and maintenance, the effectiveness of the measures is mainly judged based on newly acquired data, making it difficult to identify anomaly retention and migration.

[0004] Therefore, how to provide a pressure gauge performance data monitoring and management system and method based on the Internet of Things is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a pressure gauge performance data monitoring and management system and method based on the Internet of Things. This invention uses an improved MTAD-GAT to construct a steering evidence core, and through performance deficit write-off, calibration edge weight transfer and minimum degradation evidence core extraction, it realizes pressure gauge degradation location and disposal simulation, which has the advantages of accurate identification, clear traceability and strong maintenance targeting.

[0006] A pressure gauge performance data monitoring and management method based on the Internet of Things according to an embodiment of the present invention includes the following steps: The pressure gauge operation sequence collected by the Internet of Things is divided into operation range segments according to the direction of pressure change. The attention edge weight state of the MTAD-GAT graph at the frozen turning position and the GRU hidden state are used to form a turning evidence kernel. The attention edge weight state of the calibration graph is extracted from the pressure gauge calibration sequence. When the pressure range of the subsequent operating range fragment is reversed and covered by the historical operating range fragment, the turning evidence core corresponding to the historical operating range fragment is restored, and the subsequent operating range fragment is propagated based on the restored graph attention edge weight state and GRU hidden state. The propagation results are rearranged according to the actual time sequence and the metered pressure coordinates. The two rearrangement results are then matched, and the unclosed part is retained as a performance deficit. The subsequent running range fragments are connected to the corresponding turning evidence storage core of the performance deficit, the performance deficit is written off, and the calibration graph attention edge weight state is transferred to the graph attention edge associated with the unwritten performance deficit until the performance deficit is closed. By progressively withdrawing the transplanted calibration graph attention edge weight states, the minimum graph attention edge set that causes performance deficit and loss of closure is determined as the minimum degradation kernel; Connect the operating range fragments with the same minimum degradation certificate to form a performance degradation chain, and generate pressure gauge management tasks based on the minimum degradation certificate and the performance degradation chain; After the task is completed, the runtime fragments after the task completion will be connected to the corresponding redirection and evidence storage core for continued propagation of the performance deficit. The handling result will be determined based on the elimination status, retention status, or migration status of the performance deficit.

[0007] Optionally, the formation of the transfer evidence storage core specifically includes: By comparing the pressure change direction of adjacent sampling positions in the pressure gauge operation sequence, the position where the pressure change direction reverses is marked as the turning position, and the pressure gauge operation sequence is divided into operation range segments with the turning position as the boundary. When MTAD-GAT propagates to the turning position, freeze the graph attention edge weight state and GRU hidden state corresponding to the turning position; The graph attention edge weight state, GRU hidden state, turning position and the running range fragment with the turning position as the termination position are bound together to form a turning evidence core.

[0008] Optionally, the propagation of subsequent operational range fragmentation continuation specifically includes: Calculate the intersection of the pressure intervals of the subsequent operating range segment and the historical operating range segment. When the pressure change direction of the subsequent operating range segment and the historical operating range segment are opposite and the pressure interval intersection is the same as the pressure interval of the historical operating range segment, the pressure interval intersection is determined as the overlapping pressure interval. Recover the graph attention edge weight state and GRU hidden state from the steering evidence storage kernel corresponding to the historical operation range fragments, and take the first sampling position of the subsequent operation range fragments entering the overlapping pressure range as the access position. Starting from the access location, MTAD-GAT propagation is performed on subsequent running range fragmentation continuation based on the recovered graph attention edge weight state and GRU hidden state to obtain the propagation result.

[0009] Optionally, the rearrangement of the propagation results according to the actual time sequence and the metered pressure coordinates specifically includes: The propagation results are arranged according to the sampling time corresponding to each sampling location to form the actual time rearrangement results; The propagation results are arranged according to the pressure values ​​corresponding to each sampling location. Sampling locations with opposite pressure change directions and corresponding to the same metering pressure coordinate are aligned with each other to form a rearranged metering pressure coordinate result. The actual time rearrangement results are linked with the metered pressure coordinate rearrangement results using the sampling location as an index.

[0010] Optionally, performing the subtraction on the two rearrangement results specifically includes: Using the sampling location as an index, the current status message is extracted from the actual time rearrangement result, and the paired status message corresponding to the same metering pressure coordinate and with the opposite pressure change direction is extracted from the metering pressure coordinate rearrangement result. The current state message and the paired state message are deducted step by step along the attention edge of the graph, and the remaining state message after deduction is determined as the unclosed state message; Binding unclosed state messages with corresponding sampling locations, metering pressure coordinates, graph attention edges that generated unclosed state messages, and the steering evidence cores recovered by continued propagation creates a performance deficit.

[0011] Optionally, the write-off of performance deficits specifically includes: Restore the redirection evidence core bound to the performance deficit, and continue the MTAD-GAT propagation of subsequent operational range fragments from the sampling position corresponding to the metering pressure coordinate bound to the performance deficit. Along the graph attention edge bound to the performance debt, the state messages formed by the continued propagation are deducted in reverse from the unclosed state messages in the performance debt. The remaining unclosed state messages after deduction, along with the corresponding binding relationships, are retained as unwritten performance debts. Based on the correspondence between the metering pressure coordinates and the graph attention edges, the weight states of the calibration graph attention edges are progressively transferred to the graph attention edges bound to the unresolved performance debts. After each transfer, MTAD-GAT propagation continues based on the weight states of the transferred graph attention edges until the unclosed state messages in the unresolved performance debts are deducted.

[0012] Optionally, the process of determining the minimum degradation certificate specifically includes: Select the transplanted graph attention edges in reverse order according to the transplantation order of the calibration graph attention edge weight states, withdraw the calibration graph attention edge weight states corresponding to the selected graph attention edges, and restore the graph attention edge weight states of the selected graph attention edges before transplantation. After each withdrawal, restore the GRU hidden state in the redirection storage core bound to the performance deficit, continue MTAD-GAT propagation based on the graph attention edge weight state after withdrawal, and perform reverse deduction on the state messages formed by the continued propagation and the unclosed state messages in the performance deficit. The graph attention edges that cause performance deficits to reappear after withdrawal are reset to their corresponding calibration graph attention edge weight states, and the corresponding graph attention edges are retained. Graph attention edges that remain closed after withdrawal due to performance deficits are excluded, and the minimum set of retained graph attention edges is determined as the minimum degradation kernel.

[0013] Optionally, the generation of the pressure gauge management task specifically includes: Connect the operating range segments with the same minimum degradation certificate and corresponding continuous pressure range in the order of their generation to form a performance degradation chain; A retest task is generated when the performance degradation chain is fixed within a single pressure range; a calibration task is generated when the performance degradation chain spans consecutive pressure ranges; a maintenance task is generated when the operating range fragment continues to be connected to the performance degradation chain after the calibration task is performed; and a replacement task is generated when the operating range fragment continues to be connected to the performance degradation chain after the maintenance task is performed.

[0014] Optionally, determining the disposal result specifically includes: Before performing the pressure gauge management task, retain the binding relationship between the redirection evidence and the performance deficit; After the pressure gauge management task is completed, the running range fragment after the task is completed will be connected to the reserved steering evidence core to continue the MTAD-GAT propagation, and the status messages formed by the continued propagation and the unclosed status messages in the performance deficit will be deducted in reverse. When an unclosed state message is deducted and no new unclosed state message is generated, the performance debt is determined to be in the elimination state. When an unclosed state message is still only bound to the metering pressure coordinate and graph attention edge that the performance debt was originally bound to, the performance debt is determined to be in the retention state. When an unclosed state message is bound to other metering pressure coordinates or other graph attention edges, the performance debt is determined to be in the migration state.

[0015] According to an embodiment of the present invention, a pressure gauge performance data monitoring and management system based on the Internet of Things includes the following modules: The fragment evidence storage module is used to divide the pressure gauge operation sequence collected by the Internet of Things into operation range fragments, freeze the attention edge weight state of the MTAD-GAT graph of the turning position and the GRU hidden state to form the turning evidence storage kernel, and extract the attention edge weight state of the calibration graph from the pressure gauge calibration sequence. The status continuation module is used to restore the turnaround certificate core corresponding to the historical operation range fragment and propagate MTAD-GAT for subsequent operation range fragment continuation. The dual-sequence locking module is used to rearrange the propagation results according to the actual time sequence and the metering pressure coordinates, and perform locking to form a performance deficit; The debt write-off module is used to connect subsequent operational range fragments to the corresponding redirection and evidence storage core for performance debts, and to migrate the attention edge weight state of the calibration graph to write off performance debts. The proof core extraction module is used to progressively withdraw the attention edge weight states of the calibration graph and determine the minimum degenerate proof core; The degradation management module is used to connect operating range fragments with the same minimum degradation certificate to form a performance degradation chain and generate pressure gauge management tasks. The replay module is used to connect the operational range fragments after the task is completed to the corresponding redirection evidence storage and continuation propagation for performance deficits, and to determine the handling result.

[0016] The beneficial effects of this invention are: (1) The present invention freezes the attention edge weight state and GRU hidden state of the MTAD-GAT diagram at the pressure change turning position, and connects the subsequent reverse running range fragments to the turning evidence core for continued propagation, so that the pressure data scattered in different time periods can be connected in the historical model state. It can identify abnormal return performance in the working condition where there is no complete pressure rise and fall cycle, and avoids the pressure change process being interrupted by a fixed time window.

[0017] (2) The present invention retains the unclosed state message formed by the double sequence engagement as a performance deficit, performs the write-off using the subsequent running range fragment, and determines the minimum degradation verification by the transplantation and step-by-step withdrawal of the calibration graph attention edge weight state. It can lock the graph attention edge that continuously causes the abnormality from the model propagation relationship, which is different from simply alarming based on the indication error or abnormal score, and improves the identification accuracy of the degradation position and degradation range of the pressure gauge.

[0018] (3) The present invention connects the operating range fragments with the same minimum degradation certificate core into a performance degradation chain, and before the task is restored after the pressure gauge management task is completed, it turns to the certificate core for continued propagation. It can distinguish the elimination state, retention state and migration state of performance deficit, avoid judging the maintenance effectiveness only based on the decrease in value after the treatment, and realize the closed-loop management of pressure gauge degradation identification, task generation and treatment verification. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a pressure gauge performance data monitoring and management method based on the Internet of Things proposed in this invention; Figure 2 This is a flowchart of the performance degradation identification process based on the redirection evidence storage kernel, dual-order delimitation, and edge weight withdrawal in this invention; Figure 3 This is a module structure diagram of a pressure gauge performance data monitoring and management system based on the Internet of Things proposed in this invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0021] refer to Figures 1-3 A method for monitoring and managing pressure gauge performance data based on the Internet of Things includes the following steps: The pressure gauge operation sequence collected by the Internet of Things is divided into operation range segments according to the direction of pressure change. The attention edge weight state of the MTAD-GAT graph at the frozen turning position and the GRU hidden state are used to form a turning evidence kernel. The attention edge weight state of the calibration graph is extracted from the pressure gauge calibration sequence. MTAD-GAT uses each sampling position in the running range fragment as a graph node, establishes graph attention edges between adjacent sampling positions according to the sampling time sequence, and uses the pressure value of the corresponding sampling position as the graph node input; MTAD-GAT calculates the attention weight corresponding to the graph attention edge based on the input of the connected graph nodes, and GRU recursively calculates the GRU hidden state according to the sampling time sequence. This implementation uses a pre-trained MTAD-GAT. MTAD-GAT receives graph node inputs and graph attention edge connections corresponding to the running range fragments, and outputs the feature vector set, graph attention edge weight state, and GRU hidden state corresponding to each sampling position. The model parameters of MTAD-GAT remain fixed during the pressure gauge performance monitoring process, and the graph attention edge weight state and GRU hidden state are frozen, restored, and transferred in accordance with the method described in this implementation. When the pressure range of the subsequent operating range fragment is reversed and covered by the historical operating range fragment, the turning evidence core corresponding to the historical operating range fragment is restored, and the subsequent operating range fragment is propagated based on the restored graph attention edge weight state and GRU hidden state. The pressure range of the subsequent operating range segment overlaps the pressure range of the historical operating range segment in the opposite direction, meaning that the pressure change direction of the subsequent operating range segment is opposite to that of the historical operating range segment, and the pressure range of the subsequent operating range segment includes the entire pressure range of the historical operating range segment. The propagation results are rearranged according to the actual time sequence and the metered pressure coordinates. The two rearrangement results are then matched, and the unclosed part is retained as a performance deficit. The subsequent running range fragments are connected to the corresponding turning evidence storage core of the performance deficit, the performance deficit is written off, and the calibration graph attention edge weight state is transferred to the graph attention edge associated with the unwritten performance deficit until the performance deficit is closed. By progressively withdrawing the transplanted calibration graph attention edge weight states, the minimum graph attention edge set that causes performance deficit and loss of closure is determined as the minimum degradation kernel; Connect the operating range fragments with the same minimum degradation certificate to form a performance degradation chain, and generate pressure gauge management tasks based on the minimum degradation certificate and the performance degradation chain; After the task is completed, the runtime fragments after the task completion will be connected to the corresponding redirection and evidence storage core for continued propagation of the performance deficit. The handling result will be determined based on the elimination status, retention status, or migration status of the performance deficit.

[0022] In this embodiment, the formation of the transfer evidence storage core specifically includes: By comparing the pressure change direction of adjacent sampling positions in the pressure gauge operation sequence, the position where the pressure change direction reverses is marked as the turning position, and the pressure gauge operation sequence is divided into operation range segments with the turning position as the boundary. The pressure gauge operation sequence is a data sequence formed by arranging the continuously collected pressure values ​​according to the sampling time when the pressure gauge is in operation. The Internet of Things (IoT) acquisition terminal reads the pressure values ​​output by the pressure gauge, records the corresponding sampling time for each pressure value, and arranges them in the order of sampling time to obtain the pressure gauge operation sequence. The direction of pressure change is determined based on the pressure difference between adjacent sampling locations; the pressure gauge resolution is the smallest resolvable pressure increment recorded in the pressure gauge's technical parameters or calibration record; when the absolute value of the pressure difference is not less than the pressure gauge resolution and the pressure difference is positive, it is determined as the direction of pressure increase; when the absolute value of the pressure difference is not less than the pressure gauge resolution and the pressure difference is negative, it is determined as the direction of pressure decrease; when the absolute value of the pressure difference is less than the pressure gauge resolution, the most recently determined direction of pressure change is maintained; when the direction of pressure change has not yet been determined, the corresponding sampling location is not marked; when the adjacent pressure change direction changes from the direction of pressure increase to the direction of pressure decrease or from the direction of pressure decrease to the direction of pressure increase, the direction of pressure change is determined to have reversed; When MTAD-GAT propagates to the turning position, freeze the graph attention edge weight state and GRU hidden state corresponding to the turning position; The graph attention edge weight state is the attention weight corresponding to each graph attention edge when MTAD-GAT propagates to the turning position; the GRU hidden state is the hidden vector output by GRU after processing the sampled data corresponding to the turning position; the freeze operation is to save the graph attention edge weight state and GRU hidden state corresponding to the turning position, and the saved state is not updated with the subsequent propagation of MTAD-GAT. The graph attention edge weight state, GRU hidden state, turning position and the running range fragment with the turning position as the termination position are bound together to form a turning evidence core.

[0023] When forming the steering evidence core, the graph attention edge weight state, GRU hidden state, and running range fragment with the steering position as the termination position corresponding to the same steering position are saved as the same set of data.

[0024] In this embodiment, the subsequent propagation of the interrupted range specifically includes: Calculate the intersection of the pressure intervals of the subsequent operating range segment and the historical operating range segment. When the pressure change direction of the subsequent operating range segment and the historical operating range segment are opposite and the pressure interval intersection is the same as the pressure interval of the historical operating range segment, the pressure interval intersection is determined as the overlapping pressure interval. Historical operating range fragments are those where the sampling time at the initial sampling position is earlier than that of subsequent operating range fragments, and where corresponding turning evidence cores have already been formed; subsequent operating range fragments are those where the sampling time at the initial sampling position is later than that of historical operating range fragments. The pressure range of the operating range segment is defined by the minimum and maximum pressure values ​​in the operating range segment; the intersection of pressure ranges is the pressure range jointly covered by the pressure range of subsequent operating range segments and the pressure range of historical operating range segments; when the intersection of pressure ranges has the same lower limit pressure value and upper limit pressure value as the pressure range of historical operating range segments, it is determined that the intersection of pressure ranges is the same as the pressure range of historical operating range segments. Recover the graph attention edge weight state and GRU hidden state from the steering evidence storage kernel corresponding to the historical operation range fragments, and take the first sampling position of the subsequent operation range fragments entering the overlapping pressure range as the access position. When restoring the graph attention edge weight state and GRU hidden state, for each graph attention edge in the subsequent operating range segment, the corresponding measurement pressure coordinate combination and connection order are matched with the measurement pressure coordinate combination and connection order corresponding to each graph attention edge in the historical operating range segment. If there are historical graph attention edges with the same measurement pressure coordinate combination and connection order, the corresponding graph attention edge weight state saved in the redirection evidence kernel is loaded into the graph attention edge in the subsequent operating range segment. If there are no historical graph attention edges with the same measurement pressure coordinate combination and connection order, the sum of the absolute values ​​of the corresponding coordinate differences in the measurement pressure coordinate combinations is calculated. The graph attention edge weight state corresponding to the historical graph attention edge with the smallest sum of absolute values ​​and the smallest absolute value of the connection order difference is selected. If the values ​​are still the same, the graph attention edge weight state corresponding to the historical graph attention edge with the smaller connection order is selected. The GRU hidden state saved in the redirection evidence kernel is loaded into the GRU as the initial hidden state for continuing MTAD-GAT propagation. According to the sampling time sequence, find the sampling position in the subsequent operating range segment, and determine the first sampling position whose pressure value is not less than the lower limit pressure value of the overlapping pressure range and not greater than the upper limit pressure value of the overlapping pressure range as the access position. Starting from the access location, MTAD-GAT propagation is performed on subsequent running range fragmentation continuation based on the recovered graph attention edge weight state and GRU hidden state to obtain the propagation result.

[0025] During MTAD-GAT propagation, the sampled data starting from the access location is input into MTAD-GAT according to the sampling time sequence, and propagation is performed according to the recovered graph attention edge weight state, and recursion is performed based on the recovered GRU hidden state.

[0026] In this embodiment, the propagation results are rearranged according to the actual time sequence and the metered pressure coordinates, specifically including: The propagation results are arranged according to the sampling time corresponding to each sampling location to form the actual time rearrangement results; The propagation results are saved together with the corresponding sampling location, sampling time, pressure value, and direction of pressure change; when generating the actual time rearrangement results, the propagation results are arranged from earliest to latest according to the sampling time, and the propagation results with the same sampling time are arranged according to the order of the sampling locations; The propagation results are arranged according to the pressure values ​​corresponding to each sampling location. Sampling locations with opposite pressure change directions and corresponding to the same metering pressure coordinate are aligned with each other to form a rearranged metering pressure coordinate result. The pressure coordinates are discrete pressure positions divided according to the pressure gauge resolution; the lower limit of the pressure gauge range is the lower limit value of the range recorded in the pressure gauge technical parameters or calibration record; starting from the lower limit of the pressure gauge range, the pressure coordinates are divided sequentially according to the pressure gauge resolution, and the pressure values ​​of each sampling position are assigned to the nearest pressure coordinate, and if the distance is the same, they are assigned to the pressure coordinate with the smaller value. The two metering pressure coordinates at the two ends of the graph attention edge are arranged in ascending order of value to form the metering pressure coordinate combination corresponding to the graph attention edge; when the direction of pressure change changes, the order of the two metering pressure coordinates in the metering pressure coordinate combination remains unchanged. The pressure gauge calibration sequence is formed by arranging the standard pressure values ​​recorded in the calibration record, the pressure values ​​output by the pressure gauge, and the corresponding acquisition times in the calibration order. The pressure gauge calibration sequence is input into MTAD-GAT according to the same graph node and graph attention edge construction method as the pressure gauge operation sequence. The pressure values ​​output by the pressure gauge are used as graph node inputs, and each standard pressure value is assigned to the corresponding metering pressure coordinate according to the division rules of the metering pressure coordinate. For each graph attention edge in the calibration propagation path, the metering pressure coordinate combination is obtained according to the standard pressure values ​​corresponding to the sampling positions at both ends of the graph attention edge. The metering pressure coordinate combination, the direction of pressure change, the connection order of the graph attention edge in the calibration propagation path, and the corresponding attention weight are recorded to form the calibration graph attention edge weight state. When generating the rearrangement results of the metering pressure coordinates, the propagation results are arranged in ascending order of the metering pressure coordinates. When there are multiple propagation results under the same metering pressure coordinate, the propagation results corresponding to the pressure increase direction and the propagation results corresponding to the pressure decrease direction are arranged according to the sampling time, and aligned one by one according to the arrangement order. After the propagation results in any direction are aligned, the remaining propagation results in the other direction are retained under the corresponding metering pressure coordinate. The actual time rearrangement results are linked with the metered pressure coordinate rearrangement results using the sampling location as an index.

[0027] In this embodiment, the two rearrangement results are coupled together, specifically including: Using the sampling location as an index, the current status message is extracted from the actual time rearrangement result, and the paired status message corresponding to the same metering pressure coordinate and with the opposite pressure change direction is extracted from the metering pressure coordinate rearrangement result. The status message is a set of feature vectors saved by MTAD-GAT after propagation at the corresponding sampling position. The set of feature vectors consists of feature vectors formed by propagation along each graph attention edge within the propagation path. The current status message is the status message associated with the current sampling position in the actual time rearrangement result. The paired status message is the status message in the metering pressure coordinate rearrangement result that is located at the same metering pressure coordinate as the current status message, has the opposite pressure change direction, and is already aligned. When there is no paired status message, the feature vectors containing non-zero feature values ​​in the current status message are combined according to the connection order of the corresponding graph attention edges to form an unclosed status message, and the corresponding graph attention edge is determined as the graph attention edge that generates the unclosed status message. When all feature values ​​in the current status message are equal to zero, no unclosed status message is formed. The current state message and the paired state message are deducted step by step along the attention edge of the graph, and the remaining state message after deduction is determined as the unclosed state message; The propagation path corresponding to the current state message is composed of the graph attention edges that the current state message passes through sequentially in MTAD-GAT, obtained according to the connection relationship between each graph attention edge; the connection order of graph attention edges along the propagation path is deducted step by step; in the propagation path of the current state message and the paired state message, the corresponding graph attention edges are determined according to the combination of the measurement pressure coordinates assigned to the sampling positions at both ends of each graph attention edge and the connection order of the graph attention edges in the propagation path; the feature vectors formed by propagation along the corresponding graph attention edges are extracted, and the feature values ​​are subtracted according to the same feature positions; when all feature values ​​in each subtraction result are equal to zero, the closure is determined; when there are non-zero subtraction results, the non-zero subtraction results are combined according to the connection order of the graph attention edges to form an unclosed state message; In this embodiment, when the absolute value of the difference of the eigenvalues ​​is not greater than the model calculation tolerance, it is treated as zero; the model calculation tolerance is taken as the maximum value of the absolute value of the eigenvalues ​​obtained after deducting the corresponding status messages in the pressure gauge calibration sequence; when all eigenvalues ​​are processed to zero according to the above rules, the closure is determined to be complete. Binding unclosed state messages with corresponding sampling locations, metering pressure coordinates, graph attention edges that generated unclosed state messages, and the steering evidence cores recovered by continued propagation creates a performance deficit.

[0028] The graph attention edge that generates the unclosed state message is the graph attention edge that has a non-zero difference result after step-by-step subtraction; when a performance deficit is formed, the unclosed state message, the corresponding sampling position, the metering pressure coordinate, the graph attention edge that generates the unclosed state message, and the turning evidence core recovered by the continued propagation are saved as the same set of data.

[0029] In this implementation, the write-off of performance deficits specifically includes: Restore the redirection evidence core bound to the performance deficit, and continue the MTAD-GAT propagation of subsequent operational range fragments from the sampling position corresponding to the metering pressure coordinate bound to the performance deficit. The sampling position corresponding to the metering pressure coordinate bound to the performance deficit is the sampling position that is assigned to the same metering pressure coordinate and has the earliest sampling time in the subsequent operating range segments; starting from the corresponding sampling position, the sampling data in the subsequent operating range segments are input into MTAD-GAT in the order of sampling time. If there are no sampling positions of the metering pressure coordinates that are bound to the performance deficit in the subsequent operation range fragments, the performance deficit and the corresponding binding relationship are retained, and the cancellation is performed after the corresponding sampling positions are formed in the subsequent operation range fragments. Along the graph attention edge bound to the performance debt, the state messages formed by the continued propagation are deducted in reverse from the unclosed state messages in the performance debt. The remaining unclosed state messages after deduction, along with the corresponding binding relationships, are retained as unwritten performance debts. Reverse deduction is performed in the reverse order of the graph attention edges bound to the performance debt during the step-by-step deduction process. For each graph attention edge, the feature value of the corresponding feature position in the unclosed state message is subtracted from the feature value of the same feature position in the state message formed by the continuation propagation. If there is a non-zero feature value after deduction, the non-zero deduction result, together with the originally bound sampling position, metering pressure coordinate, graph attention edge, and turning evidence core, is retained as an unwritten performance debt. When all feature values ​​are equal to zero after reverse deduction, the performance debt is determined to be closed, and no more unwritten performance debts are formed. Based on the correspondence between the metering pressure coordinates and the graph attention edges, the weight states of the calibration graph attention edges are progressively transferred to the graph attention edges bound to the unresolved performance debts. After each transfer, MTAD-GAT propagation continues based on the weight states of the transferred graph attention edges until the unclosed state messages in the unresolved performance debts are deducted.

[0030] The hierarchical migration is performed according to the order in which the graph attention edges bound to the unresolved performance debts are arranged during the hierarchical deduction process. Each time, a graph attention edge is selected, and the corresponding metering pressure coordinate combination, pressure change direction, and connection order are obtained. The attention weights with the same three items are extracted from the calibration graph attention edge weight state. If there are no attention weights with the same three items, the sum of the absolute values ​​of the corresponding coordinate differences in the metering pressure coordinate combinations is calculated among the calibration graph attention edges with the same pressure change direction. The attention weight corresponding to the calibration graph attention edge with the smallest sum of absolute values ​​and the smallest absolute value of the connection order difference is selected. If there are no calibration graph attention edges with the same pressure change direction, the attention weights are selected in the same way among all calibration graph attention edges. If the values ​​are still the same, the attention weight corresponding to the calibration graph attention edge with the smaller connection order is selected. The extracted attention weights replace the current attention weights of the selected graph attention edges, the GRU hidden state in the redirection evidence core is restored, and MTAD-GAT propagation is continued. The state messages formed by the continued propagation and the unclosed state messages in the unresolved performance debts are deducted in reverse.

[0031] When all feature values ​​of an unclosed state message in an unresolved performance debt are reduced to zero, the performance debt is considered closed and the migration stops; if non-zero feature values ​​still exist, the migration continues to the calibration graph attention edge weight state corresponding to the next graph attention edge.

[0032] In this embodiment, the process of determining the minimum degradation verification specifically includes: Select the transplanted graph attention edges in reverse order according to the transplantation order of the calibration graph attention edge weight states, withdraw the calibration graph attention edge weight states corresponding to the selected graph attention edges, and restore the graph attention edge weight states of the selected graph attention edges before transplantation. Before transplanting the calibration graph attention edge weight state, the current graph attention edge weight state of the corresponding graph attention edge is saved; the undo operation is performed edge by edge in the reverse order of the transplantation order, and the calibration graph attention edge weight state in the selected graph attention edge is replaced with the saved graph attention edge weight state before transplantation. After each withdrawal, restore the GRU hidden state in the redirection storage core bound to the performance deficit, continue MTAD-GAT propagation based on the graph attention edge weight state after withdrawal, and perform reverse deduction on the state messages formed by the continued propagation and the unclosed state messages in the performance deficit. After each withdrawal, the hidden state of the GRU stored in the evidence storage core is loaded into the GRU, and the MTAD-GAT propagation continues from the sampling position corresponding to the metering pressure coordinate bound to the performance debt. The state messages formed by the continued propagation and the unclosed state messages in the performance debt are processed according to the reverse deduction method described above. When a non-zero feature value reappears in the deduction result, it is determined that the performance debt has reappeared. When all feature values ​​in the deduction result are equal to zero, it is determined that the performance debt remains closed. The graph attention edges that cause performance deficits to reappear after withdrawal are reset to their corresponding calibration graph attention edge weight states, and the corresponding graph attention edges are retained. Graph attention edges that remain closed after withdrawal due to performance deficits are excluded, and the minimum set of retained graph attention edges is determined as the minimum degradation kernel.

[0033] When a performance deficit reappears, the selected graph attention edge is reset to the corresponding calibration graph attention edge weight state and retained. When the performance deficit remains closed, the pre-transfer graph attention edge weight state of the selected graph attention edge is maintained and the selected graph attention edge is excluded. After processing all transferred graph attention edges, each retained graph attention edge is withdrawn in reverse order of the transfer order of the calibration graph attention edge weight state, and the GRU hidden state is restored, MTAD-GAT propagation is continued, and reverse deduction is performed in the aforementioned manner. When the performance deficit remains closed after the retained graph attention edge is withdrawn, the corresponding graph attention edge is excluded. When the performance deficit reappears, the corresponding calibration graph attention edge weight state is reset and retained. The above operations are repeated for the remaining retained graph attention edges. When the set of retained graph attention edges has not changed after one round of operations, the set of retained graph attention edges is determined as the minimum degenerate core.

[0034] In this embodiment, generating a pressure gauge management task specifically includes: Connect the operating range segments with the same minimum degradation certificate and corresponding continuous pressure range in the order of their generation to form a performance degradation chain; The same minimum degradation core means that the minimum degradation cores corresponding to different operating range segments contain the same number of graph attention edges. After arranging them according to the connection order of the propagation path, the combination of measurement pressure coordinates and the connection order corresponding to each graph attention edge are the same one by one. The generation order is determined from the earliest to the latest according to the sampling time of the starting sampling position of each operating range segment. When two pressure ranges overlap, the two pressure ranges are determined to be continuous; when two pressure ranges do not overlap, the larger of the lower limit pressure values ​​of the two pressure ranges is subtracted from the smaller of the upper limit pressure values, and the difference is not greater than the pressure gauge resolution, the two pressure ranges are determined to be continuous; when the pressure ranges corresponding to the operating range fragments with the same minimum degradation certificate are not continuous, different performance degradation chains are formed respectively. When connecting operating range segments, the connection relationship between adjacent operating range segments is saved according to the generation order; when the newly generated operating range segment has the same minimum degradation certificate as the existing operating range segment in the performance degradation chain, and the corresponding pressure range is continuous with the corresponding pressure range of the operating range segment at the end of the performance degradation chain, the newly generated operating range segment is connected to the performance degradation chain. A retest task is generated when the performance degradation chain is fixed within a single pressure range; a calibration task is generated when the performance degradation chain spans consecutive pressure ranges; a maintenance task is generated when the operating range fragment continues to be connected to the performance degradation chain after the calibration task is performed; and a replacement task is generated when the operating range fragment continues to be connected to the performance degradation chain after the maintenance task is performed.

[0035] When the lower and upper pressure values ​​of the pressure range corresponding to each operating range segment in the performance degradation chain are the same, the performance degradation chain is determined to be fixed in a single pressure range; when the performance degradation chain contains at least two different and continuous pressure ranges, the performance degradation chain is determined to span continuous pressure ranges.

[0036] When generating pressure gauge management tasks, determine the re-inspection task, calibration task, maintenance task, or replacement task based on the triggering conditions, and save the determined task type in association with the performance degradation chain that triggered the pressure gauge management task.

[0037] The operating range fragment after the calibration task is the first operating range fragment formed after the calibration task is completed; the operating range fragment after the maintenance task is the first operating range fragment formed after the maintenance task is completed.

[0038] In this embodiment, determining the disposal result specifically includes: Before performing the pressure gauge management task, retain the binding relationship between the redirection evidence and the performance deficit; When retaining the binding relationship between the redirection evidence core and the performance deficit, the redirection evidence core, unclosed state message, sampling location, metering pressure coordinates and graph attention edge are saved as the same set of data, and the saved data is not overwritten during the execution of the pressure gauge management task; After the pressure gauge management task is completed, the running range fragment after the task is completed will be connected to the reserved steering evidence core to continue the MTAD-GAT propagation, and the status messages formed by the continued propagation and the unclosed status messages in the performance deficit will be deducted in reverse. The running range fragment after the task is completed is the first running range fragment formed after the pressure gauge management task is completed; the running range fragment after the task is replaced is the first running range fragment formed after the replaced pressure gauge is put into operation; the access to the steering certificate and the execution of reverse deduction are performed in accordance with the continuing MTAD-GAT propagation method and reverse deduction method described above. In the reverse deduction after the pressure gauge management task is completed, a correspondence is established edge by edge based on the combination and connection order of the metering pressure coordinates corresponding to the graph attention edges in the propagation path corresponding to the original unclosed state message and the continuation propagation path after the task is completed. On each group of corresponding graph attention edges, the feature value of the corresponding feature position in the original unclosed state message is subtracted from the feature value of the same feature position in the state message formed by the continuation propagation. Non-zero feature vectors in the propagation path corresponding to the original unclosed state message that have not established a correspondence retain the binding relationship of the original sampling position, metering pressure coordinates, and graph attention edges. Feature vectors in the continuation propagation path after the task is completed that have not established a correspondence and contain non-zero feature values ​​are bound to the sampling position, metering pressure coordinates, and graph attention edges that generated the corresponding feature vectors. The non-zero deduction result and the non-zero feature vectors that have not established a correspondence are combined to form the unclosed state message after deduction. In the continuation propagation path after the task is completed, graph attention edges with the same metering pressure coordinate combination and connection order as the original bound graph attention edges of performance debt are identified as the original bound graph attention edges of performance debt. Graph attention edges with different metering pressure coordinate combinations or connection orders are identified as other graph attention edges. When an unclosed state message is deducted and no new unclosed state message is generated, the performance debt is determined to be in the elimination state. When an unclosed state message is still only bound to the metering pressure coordinate and graph attention edge that the performance debt was originally bound to, the performance debt is determined to be in the retention state. When an unclosed state message is bound to other metering pressure coordinates or other graph attention edges, the performance debt is determined to be in the migration state.

[0039] After completing the reverse deduction, each non-zero deduction result is determined as an unclosed state message after deduction, and the corresponding metering pressure coordinate and graph attention edge are recorded. When all the original unclosed state messages are deducted to zero and no non-zero deduction result is formed in other metering pressure coordinates or other graph attention edges, the performance deficit is determined to be in the elimination state. When the unclosed state message after deduction corresponds only to the original bound metering pressure coordinate and graph attention edge, the performance deficit is determined to be in the retention state. When the unclosed state message after deduction corresponds to at least one metering pressure coordinate different from the original bound metering pressure coordinate, or corresponds to at least one graph attention edge different from the original bound graph attention edge, the performance deficit is determined to be in the migration state.

[0040] According to an embodiment of the present invention, a pressure gauge performance data monitoring and management system based on the Internet of Things includes the following modules: The fragment evidence storage module is used to divide the pressure gauge operation sequence collected by the Internet of Things into operation range fragments, freeze the attention edge weight state of the MTAD-GAT graph of the turning position and the GRU hidden state to form the turning evidence storage kernel, and extract the attention edge weight state of the calibration graph from the pressure gauge calibration sequence. The status continuation module is used to restore the turnaround certificate core corresponding to the historical operation range fragment and propagate MTAD-GAT for subsequent operation range fragment continuation. The dual-sequence locking module is used to rearrange the propagation results according to the actual time sequence and the metering pressure coordinates, and perform locking to form a performance deficit; The debt write-off module is used to connect subsequent operational range fragments to the corresponding redirection and evidence storage core for performance debts, and to migrate the attention edge weight state of the calibration graph to write off performance debts. The proof core extraction module is used to progressively withdraw the attention edge weight states of the calibration graph and determine the minimum degenerate proof core; The degradation management module is used to connect operating range fragments with the same minimum degradation certificate to form a performance degradation chain and generate pressure gauge management tasks. The replay module is used to connect the operational range fragments after the task is completed to the corresponding redirection evidence storage and continuation propagation for performance deficits, and to determine the handling result.

[0041] Example 1: To verify the feasibility of this invention in practice, it was applied to a pressure gauge performance monitoring and management scenario for a continuous pressure supply device. The controlled medium repeatedly passes through the same pressure range during pressurization, stabilization, depressurization, and recovery. The pressure gauge output sequence contains a complete pressurization segment, a complete depressurization segment, a partial reflux segment, and a short-term fluctuation segment. The existing on-site management method sets a fixed threshold based on the deviation of the indication at a single sampling position, generating an alarm when consecutive sampling values ​​exceed the threshold. During operation, when communication jitter, medium pulsation, and short-term load impacts occur, the fixed threshold method is prone to generating false alarms. When the pressure gauge exhibits zero-point drift, reflux lag, partial jamming, and range nonlinearity, the deviation at a single sampling position may not have reached the threshold, making the fixed threshold method prone to delayed detection or missed detection.

[0042] In this embodiment, ten pressure gauges with an upper limit of 1.60 MPa and a resolution of 0.01 MPa are connected to an IoT data acquisition terminal. These ten pressure gauges cover ten characteristics: zero-point drift, hysteresis, range drift, low-pressure jamming, medium-pressure nonlinearity, high-pressure hysteresis, communication transient jitter, short-term media pulsation, cross-range mechanical wear, and permanent deformation of elastic elements. The IoT data acquisition terminal reads pressure values ​​and records sampling times at fixed sampling intervals, resulting in a total of 568 operating range segments and 28,460 sampling locations across the ten pressure gauges. Each pressure gauge also forms a separate pressure gauge calibration sequence, covering both pressure increase and decrease directions. Standard pressure points are arranged at 0.10 MPa intervals, and the standard pressure values, the pressure values ​​output by the pressure gauges, and the corresponding acquisition times are all saved.

[0043] After the pressure gauge operation sequence enters MTAD-GAT, each sampling position is used as a graph node, and adjacent sampling positions are used to establish graph attention edges according to the sampling time sequence. When the pressure difference between adjacent sampling positions reaches the pressure gauge resolution, the direction of pressure increase or decrease is determined, and the position where the pressure change direction reverses is used as the turning position. After MTAD-GAT propagates to the turning position, it saves the graph attention edge weight state and GRU hidden state. The saved result is bound to the turning position and the corresponding operating range segment to form a turning evidence core. In this embodiment, a total of 214 turning evidence cores are formed. Each turning evidence core retains the propagation state of the corresponding operating range segment before the turning. When subsequent operating range segments enter the same pressure range, the propagation starting point can be directly restored.

[0044] When the pressure change direction of subsequent operating range segments is opposite to that of historical operating range segments, and the pressure range of subsequent operating range segments covers the entire pressure range of historical operating range segments, the steering evidence core corresponding to the historical operating range segments is restored. During the restoration process, attention edges of the graph are matched according to the combination of metered pressure coordinates and the connection order. The matched edges are loaded with the historical graph attention edge weight state, and the GRU loads the hidden state saved by the steering evidence core. Subsequent operating range segments continue propagation from the first sampling position entering the overlapping pressure range, and the propagation results are rearranged according to the sampling time and metered pressure coordinates. The pressure rise status message and the pressure fall status message under the same metered pressure coordinate are subtracted edge by edge. Closure is completed when all feature values ​​are zero, and performance deficit is formed when non-zero subtraction results are retained.

[0045] The calibration sequence for ten pressure gauges uses the pressure values ​​output by the gauges as graph node inputs, and standard pressure values ​​are used to determine the measurement pressure coordinates. Each graph attention edge within the calibration propagation path stores the measurement pressure coordinate combination, pressure change direction, connection order, and attention weight, forming the calibration graph attention edge weight state. If performance deficits still exist after reverse subtraction of range breaks during subsequent operations, the calibration graph attention edge weight state is transferred according to the order of the graph attention edges during the step-by-step subtraction process. When the measurement pressure coordinate combination, pressure change direction, and connection order are completely identical, the corresponding attention weight is directly extracted. If no completely identical results exist, the attention weight with the smallest sum of absolute coordinate differences and the smallest absolute difference in connection order is selected.

[0046] After the performance deficit is closed, the calibration graph attention edge weights are withdrawn edge by edge in the reverse order of the migration sequence. Graph attention edges that reappear after withdrawal due to performance deficit are retained, while graph attention edges whose performance deficit remains closed after withdrawal are excluded. The remaining graph attention edges undergo repeated withdrawal, continuation propagation, and reverse deduction. When the retained set no longer decreases, a minimum degradation proof core is formed. In ten stress tables, eight exhibit persistent performance degradation, and the resulting minimum degradation proof core contains one to four graph attention edges. The state messages corresponding to communication transient jitter and short-term media pulsation do not form stable performance deficits and do not generate minimum degradation proof cores.

[0047] Operating range segments with the same minimum degradation criterion and continuous pressure ranges are connected in the order of their generation to form a performance degradation chain. A re-inspection task is generated when the performance degradation chain is fixed within a single pressure range, and a calibration task is generated when it spans consecutive pressure ranges. A maintenance task is generated when the first operating range segment after calibration is connected to the original performance degradation chain, and a replacement task is generated when the first operating range segment after maintenance is connected to the original performance degradation chain. In this embodiment, zero-point drift, hysteresis, range drift, and nonlinearity in the medium-pressure zone generate calibration tasks; low-pressure zone jamming generates maintenance tasks; high-pressure zone hysteresis continues to form the same performance degradation chain after calibration and is converted into a maintenance task; cross-range mechanical wear generates maintenance tasks; and permanent deformation of elastic elements still forms the same performance degradation chain after maintenance and is converted into a replacement task.

[0048] After the pressure gauge management task is completed, the retained steering evidence core is reconnected to the first operational range fragment after the task completion. The propagation path corresponding to the original unclosed state message and the continued propagation path after the task completion are established as edge-by-edge correspondences according to the combination of metering pressure coordinates and the connection order. Non-zero deduction results are re-bound to the sampling position, metering pressure coordinates, and graph attention edges. When all original unclosed state messages are deducted and no new non-zero results are formed, it is determined to be in the elimination state; when the non-zero results are still located at the original metering pressure coordinates and the original graph attention edges, it is determined to be in the retention state; when the non-zero results are transferred to other metering pressure coordinates or other graph attention edges, it is determined to be in the migration state.

[0049] The final actual state of the ten pressure gauges was determined based on the full-range calibration results of the standard pressure source and the disassembly and inspection results of the pressure gauges. The fixed threshold method uses an alarm condition where the maximum indication error of three consecutive operating range segments exceeds 1.00%FS. The control method uses a fixed threshold judgment, generating an alarm when the maximum indication error continuously exceeds 1.00%FS, where FS represents the upper limit of the range. The fixed threshold method identified seven out of eight truly degraded pressure gauges, missing one low-pressure zone stuck pressure gauge. Communication transient jitter and short-term medium pulsation both triggered the fixed threshold alarm, resulting in two false alarms. This invention identifies eight truly degraded pressure gauges after performance degradation, but does not connect the two types of transient disturbances into a performance degradation chain. The eight truly degraded pressure gauges require an average of 3.9 operating range segments to form performance degradation, while the fixed threshold method requires an average of 8.6 operating range segments for the seven already identified truly degraded pressure gauges. Before the treatment, the average maximum indication error of the eight truly degraded pressure gauges was 1.32%FS. After calibration, repair or replacement, the average value dropped to 0.37%FS.

[0050] Table 1: Comparison of Pressure Gauge Degradation Identification, Task Generation, and Handling Results

[0051] Table 1 shows that the fixed threshold method generated nine alarms for ten pressure gauges, seven of which corresponded to actual degradation, and two to communication transient jitter and short-term medium pulsation. Low-pressure area jamming was not identified. Based on the final actual state of the ten pressure gauges, the fixed threshold method correctly identified seven, with an accuracy rate of 70%. This invention identified eight truly degraded pressure gauges, and neither of the two pressure gauges experiencing transient disturbances formed a performance degradation chain. The identification results were consistent with the final re-inspection results, achieving an accuracy rate of 100%. These differences indicate that this invention does not simply amplify the deviation value, but rather uses the propagation state matching results under opposite pressure change directions to determine continuous degradation, thus separating short-term fluctuations from stable degradation.

[0052] The number of operating range segments required to generate a performance deficit for eight truly degraded pressure gauges is three to five, with an average of 3.9. The fixed threshold method requires six to eleven operating range segments to generate an alarm for the seven already identified truly degraded pressure gauges, with an average of 8.6. This invention, before the indication error continuously exceeds the fixed threshold, establishes a performance deficit through the non-closed results of the pressure increase and decrease status messages under the same metering pressure coordinate. The maximum indication error corresponding to low-pressure zone stagnation is only 0.96%FS, and the fixed threshold method does not generate an alarm. This invention generates a stable performance deficit and a maintenance task at the third operating range segment.

[0053] The minimum degradation checklist contains one to four graph attention edges. Zero-point drift, range drift, and high-pressure area return hysteresis all converge to two graph attention edges, while low-pressure area jamming converges to one graph attention edge. Cross-range mechanical wear and permanent deformation of elastic elements extend to four graph attention edges. The number of graph attention edges is consistent with the degradation impact range; a single local jamming is not extended into a full-range fault, and cross-pressure range wear is not compressed into a single-point deviation. The generated re-inspection, calibration, repair, and replacement tasks can be progressively varied according to the performance degradation chain coverage, avoiding the direct generation of high-level handling tasks based on a single alarm.

[0054] The average maximum indication error of the eight real degraded pressure gauges before treatment was 1.32%FS, and the average after treatment was 0.37%FS, with an average reduction of 72.1%. P06 remained connected to the original performance degradation chain in the first operating range interruption after calibration; the management task was changed from calibration to maintenance, and the performance deficit was eliminated after maintenance. P10 retained the same minimum degradation check in the first operating range interruption after maintenance; the management task was changed to replacement, and the maximum indication error decreased to 0.18%FS after replacement. These results demonstrate that the present invention can reproduce the operating state after task completion, identify mechanical hysteresis not eliminated by calibration and permanent deformation not eliminated by maintenance, and the treatment results do not depend on the formal record of task completion.

[0055] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for monitoring and managing pressure gauge performance data based on the Internet of Things, characterized in that, Includes the following steps: The pressure gauge operation sequence collected by the Internet of Things is divided into operation range segments according to the direction of pressure change. The attention edge weight state of the MTAD-GAT graph at the frozen turning position and the GRU hidden state are used to form a turning evidence kernel. The attention edge weight state of the calibration graph is extracted from the pressure gauge calibration sequence. When the pressure range of the subsequent operating range fragment is reversed and covered by the historical operating range fragment, the turning evidence core corresponding to the historical operating range fragment is restored, and the subsequent operating range fragment is propagated based on the restored graph attention edge weight state and GRU hidden state. The propagation results are rearranged according to the actual time sequence and the metered pressure coordinates. The two rearrangement results are then matched, and the unclosed part is retained as a performance deficit. The subsequent running range fragments are connected to the corresponding turning evidence storage core of the performance deficit, the performance deficit is written off, and the calibration graph attention edge weight state is transferred to the graph attention edge associated with the unwritten performance deficit until the performance deficit is closed. By progressively withdrawing the transplanted calibration graph attention edge weight states, the minimum graph attention edge set that causes performance deficit and loss of closure is determined as the minimum degradation kernel; Connect the operating range fragments with the same minimum degradation certificate to form a performance degradation chain, and generate pressure gauge management tasks based on the minimum degradation certificate and the performance degradation chain; After the task is completed, the runtime fragments after the task completion will be connected to the corresponding redirection and evidence storage core for continued propagation of the performance deficit. The handling result will be determined based on the elimination status, retention status, or migration status of the performance deficit.

2. The method for monitoring and managing pressure gauge performance data based on the Internet of Things according to claim 1, characterized in that, The formation of the transfer evidence storage core specifically includes: By comparing the pressure change direction of adjacent sampling positions in the pressure gauge operation sequence, the position where the pressure change direction reverses is marked as the turning position, and the pressure gauge operation sequence is divided into operation range segments with the turning position as the boundary. When MTAD-GAT propagates to the turning position, freeze the graph attention edge weight state and GRU hidden state corresponding to the turning position; The graph attention edge weight state, GRU hidden state, turning position and the running range fragment with the turning position as the termination position are bound together to form a turning evidence core.

3. The method for monitoring and managing pressure gauge performance data based on the Internet of Things according to claim 2, characterized in that, The subsequent propagation of fragmented data ranges includes: Calculate the intersection of the pressure intervals of the subsequent operating range segment and the historical operating range segment. When the pressure change direction of the subsequent operating range segment and the historical operating range segment are opposite and the pressure interval intersection is the same as the pressure interval of the historical operating range segment, the pressure interval intersection is determined as the overlapping pressure interval. Recover the graph attention edge weight state and GRU hidden state from the steering evidence storage kernel corresponding to the historical operation range fragments, and take the first sampling position of the subsequent operation range fragments entering the overlapping pressure range as the access position. Starting from the access location, MTAD-GAT propagation is performed on subsequent running range fragmentation continuation based on the recovered graph attention edge weight state and GRU hidden state to obtain the propagation result.

4. The method for monitoring and managing pressure gauge performance data based on the Internet of Things according to claim 3, characterized in that, The rearrangement of the propagation results according to the actual time sequence and the metered pressure coordinates specifically includes: The propagation results are arranged according to the sampling time corresponding to each sampling location to form the actual time rearrangement results; The propagation results are arranged according to the pressure values ​​corresponding to each sampling location. Sampling locations with opposite pressure change directions and corresponding to the same metering pressure coordinate are aligned with each other to form a rearranged metering pressure coordinate result. The actual time rearrangement results are linked with the metered pressure coordinate rearrangement results using the sampling location as an index.

5. The method for monitoring and managing pressure gauge performance data based on the Internet of Things according to claim 4, characterized in that, The specific steps of performing the subtraction on the two rearrangement results include: Using the sampling location as an index, the current status message is extracted from the actual time rearrangement result, and the paired status message corresponding to the same metering pressure coordinate and with the opposite pressure change direction is extracted from the metering pressure coordinate rearrangement result. The current state message and the paired state message are deducted step by step along the attention edge of the graph, and the remaining state message after deduction is determined as the unclosed state message; Binding unclosed state messages with corresponding sampling locations, metering pressure coordinates, graph attention edges that generated unclosed state messages, and the steering evidence cores recovered by continued propagation creates a performance deficit.

6. The method for monitoring and managing pressure gauge performance data based on the Internet of Things according to claim 5, characterized in that, The process of writing off performance deficits specifically includes: Restore the redirection evidence core bound to the performance deficit, and continue the MTAD-GAT propagation of subsequent operational range fragments from the sampling position corresponding to the metering pressure coordinate bound to the performance deficit. Along the graph attention edge bound to the performance debt, the state messages formed by the continued propagation are deducted in reverse from the unclosed state messages in the performance debt. The remaining unclosed state messages after deduction, along with the corresponding binding relationships, are retained as unwritten performance debts. Based on the correspondence between the metering pressure coordinates and the graph attention edges, the weight states of the calibration graph attention edges are progressively transferred to the graph attention edges bound to the unresolved performance debts. After each transfer, MTAD-GAT propagation continues based on the weight states of the transferred graph attention edges until the unclosed state messages in the unresolved performance debts are deducted.

7. The method for monitoring and managing pressure gauge performance data based on the Internet of Things according to claim 6, characterized in that, The process of determining the minimum degenerate verification code specifically includes: Select the transplanted graph attention edges in reverse order according to the transplantation order of the calibration graph attention edge weight states, withdraw the calibration graph attention edge weight states corresponding to the selected graph attention edges, and restore the graph attention edge weight states of the selected graph attention edges before transplantation. After each withdrawal, restore the GRU hidden state in the redirection storage core bound to the performance deficit, continue MTAD-GAT propagation based on the graph attention edge weight state after withdrawal, and perform reverse deduction on the state messages formed by the continued propagation and the unclosed state messages in the performance deficit. The graph attention edges that cause performance deficits to reappear after withdrawal are reset to their corresponding calibration graph attention edge weight states, and the corresponding graph attention edges are retained. Graph attention edges that remain closed after withdrawal due to performance deficits are excluded, and the minimum set of retained graph attention edges is determined as the minimum degradation kernel.

8. The method for monitoring and managing pressure gauge performance data based on the Internet of Things according to claim 7, characterized in that, The task of generating pressure gauge management specifically includes: Connect the operating range segments with the same minimum degradation certificate and corresponding continuous pressure range in the order of their generation to form a performance degradation chain; A retest task is generated when the performance degradation chain is fixed within a single pressure range; a calibration task is generated when the performance degradation chain spans consecutive pressure ranges; a maintenance task is generated when the operating range fragment continues to be connected to the performance degradation chain after the calibration task is performed; and a replacement task is generated when the operating range fragment continues to be connected to the performance degradation chain after the maintenance task is performed.

9. A method for monitoring and managing pressure gauge performance data based on the Internet of Things according to claim 8, characterized in that, The determination of the disposal result specifically includes: Before performing the pressure gauge management task, retain the binding relationship between the redirection evidence and the performance deficit; After the pressure gauge management task is completed, the running range fragment after the task is completed will be connected to the reserved steering evidence core to continue the MTAD-GAT propagation, and the status messages formed by the continued propagation and the unclosed status messages in the performance deficit will be deducted in reverse. When an unclosed state message is deducted and no new unclosed state message is generated, the performance debt is determined to be in the elimination state. When an unclosed state message is still only bound to the metering pressure coordinate and graph attention edge that the performance debt was originally bound to, the performance debt is determined to be in the retention state. When an unclosed state message is bound to other metering pressure coordinates or other graph attention edges, the performance debt is determined to be in the migration state.

10. A pressure gauge performance data monitoring and management system based on the Internet of Things (IoT), applied to the pressure gauge performance data monitoring and management method based on the IoT as described in any one of claims 1 to 9, characterized in that, Includes the following modules: The fragment evidence storage module is used to divide the pressure gauge operation sequence collected by the Internet of Things into operation range fragments, freeze the attention edge weight state of the MTAD-GAT graph of the turning position and the GRU hidden state to form the turning evidence storage kernel, and extract the attention edge weight state of the calibration graph from the pressure gauge calibration sequence. The status continuation module is used to restore the turnaround certificate core corresponding to the historical operation range fragment and propagate MTAD-GAT for subsequent operation range fragment continuation. The dual-sequence locking module is used to rearrange the propagation results according to the actual time sequence and the metering pressure coordinates, and perform locking to form a performance deficit; The debt write-off module is used to connect subsequent operational range fragments to the corresponding redirection and evidence storage core for performance debts, and to migrate the attention edge weight state of the calibration graph to write off performance debts. The proof core extraction module is used to progressively withdraw the attention edge weight states of the calibration graph and determine the minimum degenerate proof core. The degradation management module is used to connect operating range fragments with the same minimum degradation certificate to form a performance degradation chain and generate pressure gauge management tasks. The replay module is used to connect the operational range fragments after the task is completed to the corresponding redirection evidence storage and continuation propagation for performance deficits, and to determine the handling result.