A gas flowmeter abnormality early warning and remote adjustment control method and system

Abstract

The present application relates to the technical field of gas flow meter, and more particularly to a gas flow meter abnormal early warning and remote adjustment control method and system, which obtains multi-dimensional time sequence data of a resonant sensing unit, analyzes features and reconstructs a logic level, and constructs a structure health state perception atlas; calculates local offset amplitude and mines damage disturbance features to generate a fatigue damage disturbance spectrum graph; performs deep analysis to identify mutation nodes and damage evolution types; traces potential damage nodes based on full-cycle time sequence drift, predicts potential damage nodes, identifies expansion and propagation ports, and constructs a multi-dimensional fatigue damage expansion trajectory vector field through multi-time point evolution simulation and diffusion coupling; and performs multi-strategy excitation intervention deduction and decision optimization on the vector field to form an excitation frequency remote adjustment strategy. The method realizes early damage accurate identification, evolution trend prediction and active closed-loop regulation, effectively suppresses structure fatigue expansion, and guarantees measurement accuracy and equipment safety.

Description

Technical Field

[0001] This invention relates to the field of gas flow meter technology, and in particular to a method and system for abnormal early warning and remote adjustment control of gas flow meters. Background Technology

[0002] With the continuous advancement of intelligent construction of urban gas pipeline networks, gas flow meters, as key metering and monitoring terminals, directly impact energy management efficiency and pipeline safety through their operational stability and measurement accuracy. High-precision gas metering equipment, such as Coriolis mass flow meters and ultrasonic flow meters, generally rely on the stable resonant characteristics of the sensor pipeline to achieve accurate mass flow measurement. However, during long-term operation, the sensor pipeline is susceptible to progressive damage such as microcracks, stress concentration, and structural fatigue due to factors such as alternating fluid loads, external vibration interference, temperature cycling stress, and material aging, leading to a drift in its inherent resonant frequency. This abnormal shift in resonant frequency not only introduces significant metering errors but can also, in severe cases, cause sensor structural fractures, resulting in equipment failure or even safety accidents.

[0003] Currently, traditional gas flow meter health monitoring mainly relies on periodic manual inspections or simple threshold alarm mechanisms, making it difficult to sensitively identify early structural fatigue damage and predict its evolution trend. Some existing technologies assess equipment status by collecting vibration signals and performing spectral analysis, but these are mostly limited to single-dimensional signal processing and lack the ability to fuse and analyze multi-source heterogeneous data, thus failing to construct a complete structural health evolution map. Furthermore, existing systems generally lack closed-loop control capabilities, making it difficult to proactively intervene even when anomalies are detected, thus failing to achieve the transformation from passive alarm to proactive control.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this invention is to provide a method and system for abnormal early warning and remote adjustment and control of gas flow meters, aiming to solve the technical problems of existing gas flow meter structural health monitoring technologies, such as difficulty in accurately identifying early fatigue damage evolution trends, lack of multi-source data fusion and analysis capabilities, and inability to achieve the transformation from passive alarm to active remote control.

[0006] To achieve the above objectives, the present invention provides a method for abnormal early warning and remote adjustment control of a gas flow meter, the method comprising: Acquire multi-dimensional time-series data of the resonant sensing unit of the gas flow meter, perform resonant characteristic information analysis and structural state logic hierarchy reconstruction, and construct a structural health status perception map of the flow meter sensor pipeline. The local offset amplitude of multi-resonance features is calculated from the structural health status perception map, and structural damage disturbance features are mined to construct a fatigue damage disturbance spectrum map. Deep feature analysis and node damage mutation feature analysis are performed on the fatigue damage disturbance spectrum to generate the damage evolution dynamic type of mutation nodes. The fatigue damage disturbance spectrum is traced back to its full-cycle temporal drift, and potential damage nodes are predicted based on the damage evolution dynamic type to identify other potential damage propagation ports. Multi-time point damage evolution simulation and damage perturbation diffusion coupling are performed on other potential damage propagation ports to construct a multidimensional fatigue damage propagation trajectory vector field; A multi-strategy excitation intervention was simulated for the multidimensional fatigue damage propagation trajectory vector field, and then the excitation control decision was optimized to construct a remote adjustment optimization strategy for the excitation frequency.

[0007] Optionally, the step of acquiring multi-dimensional time-series data of the gas flow meter resonant sensing unit, performing resonant characteristic information analysis and structural state logic hierarchy reconstruction, and constructing a structural health status perception map of the flow meter sensor pipeline includes: Multi-dimensional resonant timing data stream is obtained based on the excitation drive unit and sensing detection unit of the gas flow meter; Heterogeneous feature information is parsed from the multi-dimensional resonant timing data stream to generate heterogeneous resonant feature information, which includes the resonant frequency of the pipeline, the driving phase difference, the feedback voltage of the excitation coil, and the resonant quality factor Q value. The heterogeneous resonance characteristic information is time-stamp aligned and numerically standardized to construct multidimensional structural state standardized features. Multi-feature correlation analysis was performed on the standardized features of multidimensional structural states to extract the causal relationships between different features. Structural dimension mining is performed on the standardized features of the multidimensional structural state, and logical hierarchy is reconstructed based on the causal relationship to construct a structural health status perception map of the flow meter sensor pipeline.

[0008] Optionally, the step of calculating the local offset amplitude of multi-resonance features in the structural health status perception map and mining structural damage disturbance features to construct a fatigue damage disturbance spectrum map includes: The sensor pipeline structure health status perception map is decomposed into time-series states to generate multiple continuous resonance detection windows; The local offset amplitude of the multi-resonance characteristics is calculated for multiple continuous resonance detection windows to obtain the resonance offset amplitude value of each window; Based on the resonant offset amplitude value, transient large offset distortion is identified, and structural state distortion points are marked. Based on a preset fatigue damage sensitivity threshold, damage risk sensitivity assessment is performed on structural state distortion points to extract potential damage perception factors. Dynamic damage perturbation features are mined from potential damage sensing factors to construct a fatigue damage perturbation spectrum.

[0009] Optionally, the step of performing dynamic damage perturbation feature mining on potential damage sensing factors to construct a fatigue damage perturbation spectrum includes: Periodic damage perturbation characteristics are analyzed for potential damage sensing factors, and periodic resonance offset characteristics are extracted. Based on the periodic resonance offset characteristics, multi-time-point offset fitting is performed to construct the damage disturbance waveform curve; Calculate the drift frequency intensity and trend slope of the damage disturbance waveform curve; Based on the drift frequency intensity and trend slope, drift frequency continuity is mined to construct a fatigue damage disturbance spectrum.

[0010] Optionally, the step of performing deep feature analysis and nodal damage mutation feature analysis on the fatigue damage disturbance spectrum to generate the damage evolution dynamic type of mutation nodes includes: Deep feature analysis is performed on the fatigue damage disturbance spectrum, and damage feature annotation and encoding are performed to obtain the damage disturbance feature code; Based on the damage disturbance feature encoding, the sensor pipeline structure health status perception map is correlated and mapped and the topology is located to obtain the map location information of the damage factor; Based on the location information of the map, damage propagation path mining is performed to extract the damage propagation path; A path abrupt change key node analysis was performed on the damage propagation path to extract microcrack initiation nodes, fatigue accumulation nodes, and low-frequency activation nodes of stress concentration in the path. The node state abrupt change characteristics of the microcrack initiation node, fatigue accumulation node, and low-frequency activation node of stress concentration are analyzed to generate the damage evolution dynamic type of the abrupt node.

[0011] Optionally, the step of performing full-cycle time-series drift tracing on the fatigue damage disturbance spectrum and predicting potential damage nodes based on the damage evolution dynamics type, and identifying other potential damage propagation ports, includes: Multi-feature correlation mining is performed on the fatigue damage disturbance spectrum to identify associated damage feature groups; Deeply deconstruct the damage propagation logic of the associated damage feature group to generate damage extension propagation logic; Based on the damage propagation logic, the damage propagation path is traced throughout the entire time cycle to obtain the full-cycle damage propagation chain. Based on the aforementioned dynamic damage evolution type and full-cycle damage propagation chain, potential damage nodes are predicted, and other potential damage propagation ports are identified.

[0012] Optionally, the step of performing multi-time-point damage evolution simulation and damage perturbation diffusion coupling on other potential damage propagation ports to construct a multidimensional fatigue damage propagation trajectory vector field includes: Identify the current structural health status of sensor pipelines based on multi-dimensional resonant timing data streams; Real-time dynamic stress distribution analysis is performed on the current structural health status of the sensor pipeline to extract key stress characteristics; Based on the key stress characteristics, multi-time-point damage evolution simulations were performed on other potential damage propagation ports to generate damage evolution simulation data; Damage trajectory evolution analysis is performed on damage evolution simulation data to generate damage evolution trajectories; Calculate the direction of expansion, rate of expansion, strain potential energy, and probability of fracture failure of the damage evolution trajectory; Based on the aforementioned expansion direction, expansion rate, strain potential energy, and fracture failure probability, damage perturbation diffusion coupling is performed to construct a multidimensional fatigue damage expansion trajectory vector field.

[0013] Optionally, the step of performing multi-strategy excitation intervention deduction on the multidimensional fatigue damage propagation trajectory vector field, followed by excitation control decision optimization, to construct a remote adjustment optimization strategy for the excitation frequency includes: Based on the flow meter parameter adjustment space, identify the excitation parameter resources that can be remotely called by the gas flow meter; according to the excitation parameter resources that can be remotely called, perform multi-strategy anti-damage intervention simulation on the multidimensional fatigue damage propagation trajectory vector field, and generate damage control simulation data of multiple excitation adjustment schemes; Calculate the recovery speed and residual damage propagation rate of the sensor pipeline resonance state after the excitation intervention of the damage control simulation data; predict the secondary fracture risk probability based on the recovery speed and residual damage propagation rate to generate a secondary risk probability; and generate an adaptive anomaly warning signal based on the secondary risk probability. The secondary residual damage density is calculated from the damage control simulation data to generate a secondary residual damage density; the damage diffusion trend is analyzed from the secondary residual damage density to obtain a secondary damage diffusion trend map. Based on the secondary damage diffusion trend diagram, the excitation control decision is optimized, and a remote adjustment optimization strategy for the excitation frequency is constructed. Based on the remote adjustment optimization strategy for the excitation frequency and the adaptive damage anomaly early warning signal, the fatigue damage anomaly early warning and remote adjustment control operation of the gas flow meter structure is carried out.

[0014] Optionally, the construction of the remote adjustment and optimization strategy for the excitation frequency includes: A multi-objective weighted algorithm is used to solve the objective function in the risk-adverse decision optimization strategy. The objective function includes the target of resonance amplitude stability after excitation frequency adjustment, the target of minimizing measurement error, and the target of suppressing pipeline stress fatigue. The optimal excitation drive frequency parameters and the optimal drive voltage gain parameters are generated based on the solution results. The optimal excitation drive frequency parameters and optimal drive voltage gain parameters are sent to the excitation control module of the gas flow meter via a remote communication protocol. The excitation control module receives parameters and performs closed-loop feedback control based on the phase-locked loop circuit, adjusting the excitation frequency of the sensor circuit to the preset optimal resonant operating point in real time. Record the resonant frequency drift and measurement deviation data before and after adjustment, and perform online iterative updates of model parameters to generate a dynamically corrected risk-resistant decision optimization strategy.

[0015] Furthermore, to achieve the above objectives, the present invention also provides a gas flow meter abnormality early warning and remote adjustment control system, the system comprising: The graph construction module is used to acquire multi-dimensional time-series data of the resonant sensing unit of the gas flow meter, perform resonant characteristic information analysis and structural state logic hierarchy reconstruction, and construct a structural health status perception graph of the flow meter sensor pipeline. The spectrum mining module is used to calculate the local offset amplitude of multi-resonance features in the structural health status perception map and to mine structural damage disturbance features in order to construct a fatigue damage disturbance spectrum map. The type generation module is used to perform deep feature analysis and node damage mutation feature analysis on the fatigue damage disturbance spectrum map, and generate the damage evolution dynamic type of the mutation node. The port identification module is used to trace the full-cycle temporal drift of the fatigue damage disturbance spectrum and predict potential damage nodes based on the damage evolution dynamic type, and identify other potential damage propagation ports. The vector modeling module is used to simulate damage evolution at multiple time points and couple damage disturbance diffusion for other potential damage propagation ports, and to construct a multidimensional fatigue damage propagation trajectory vector field. The strategy optimization module is used to perform multi-strategy excitation intervention simulation on the multidimensional fatigue damage propagation trajectory vector field, and then optimize the excitation control decision to construct an optimized strategy for remote adjustment of excitation frequency.

[0016] This invention provides a method for abnormal early warning and remote adjustment control of gas flow meters. By constructing a structural health status perception map and a fatigue damage disturbance spectrum map, this method achieves highly sensitive identification and dynamic evolution feature extraction of early-stage minor damage in the gas flow meter sensor pipeline, overcoming the problems of missed or false alarms caused by traditional monitoring methods relying on single threshold alarms. Through full-cycle time-series drift tracing and damage propagation path mining, it can accurately predict potential damage propagation ports, improving the forward-looking prediction capability of structural fatigue failure trends. Furthermore, by constructing a multi-dimensional fatigue damage propagation trajectory vector field and performing multi-strategy excitation intervention simulation, it achieves a closed-loop transformation from passive early warning to active control. The generated remote adjustment optimization strategy for excitation frequency can dynamically adjust the sensor operating point, suppress damage propagation, restore resonance stability, effectively extend equipment service life, and ensure the long-term operational accuracy and intrinsic safety of the gas metering system. It is particularly suitable for unattended, remotely deployed intelligent gas pipeline network operation and maintenance scenarios. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating an embodiment of the gas flow meter abnormality early warning and remote adjustment control method of the present invention; Figure 2 This is a structural block diagram of an embodiment of the gas flow meter abnormality early warning and remote adjustment control system of the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] Reference Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the gas flow meter abnormality early warning and remote adjustment control method of the present invention, which presents an embodiment of the gas flow meter abnormality early warning and remote adjustment control method of the present invention.

[0021] In one embodiment, the gas flow meter abnormality early warning and remote adjustment control method includes: Step S100: Obtain multi-dimensional time-series data of the gas flow meter resonant sensing unit, perform resonant characteristic information analysis and structural state logic level reconstruction, and construct a structural health status perception map of the flow meter sensor pipeline.

[0022] The resonant sensing unit of the gas flow meter is a core sensing component that senses the mass flow rate of fluid and operates based on stable resonance characteristics. It reflects the mass flow rate through changes in its inherent resonant frequency, and its structural integrity directly affects measurement accuracy and equipment safety. In this embodiment, the resonant sensing unit can generate a stable resonant response under excitation based on physical mechanisms such as the Coriolis effect or ultrasonic time difference. Multi-dimensional time-series data can be a collection of time-series signals continuously acquired from the unresonant sensing unit across multiple physical dimensions. This data can provide multi-angle dynamic input for structural health assessment, enhancing sensitivity to minor damage. Furthermore, the multi-dimensional time-series data can be simultaneously acquired using embedded accelerometers, temperature sensors, strain gauges, and a resonant frequency monitoring module, collecting signals such as vibration, temperature, strain, and resonant frequency. Resonance characteristic information can be a set of characteristic parameters directly related to resonance behavior extracted from the multi-dimensional time-series data. This information can characterize the current resonance state of the sensor, serving as a basis for structural health assessment. In one specific embodiment, resonance characteristic information can be obtained by performing spectral analysis, envelope demodulation, or modal identification on the original time-series signal to obtain parameters such as dominant frequency, damping ratio, and phase difference. The structural state logic hierarchy can be a logical model that abstracts the health status of the sensor pipeline structure in layers. It can be used to map continuous physical states into a computable logic hierarchy, supporting state reconstruction and anomaly identification. Furthermore, the structural state logic hierarchy can be based on material mechanics models and historical damage data to construct a multi-level state mapping relationship from microcracks to macroscopic failure. The structural health status perception map can be a visualized or computable health status characterization model constructed by integrating multi-dimensional resonance characteristics and the structural state logic hierarchy. It can be used to achieve highly sensitive identification and dynamic evolution tracking of progressive damage such as early microcracks and stress concentration. In this embodiment, the structural health status perception map can receive multi-dimensional time-series data and resonance characteristic information as input, and output a state basis for constructing a fatigue damage perturbation spectrum map.

[0023] Step S200: Calculate the local offset amplitude of the multi-resonance feature of the structural health status perception map and mine the structural damage disturbance feature to construct the fatigue damage disturbance spectrum map.

[0024] The local offset amplitude of multi-resonance features can be a measure of the local deviation of each resonance feature in the structural health status perception map relative to the baseline state. It can be used to quantify the degree of abnormality of resonance characteristics at different locations or modes and locate potential damage areas. In a specific embodiment, the local offset amplitude of multi-resonance features can be compared with the current feature and the historical health baseline through a sliding window to calculate the Euclidean distance or correlation coefficient deviation. The structural damage disturbance feature can be a unique manifestation mode of the resonant system disturbance caused by microcracks or stress concentration in the frequency domain or time-frequency domain. It can be used as a fingerprint feature of fatigue damage to distinguish between normal aging and abnormal degradation. Furthermore, the structural damage disturbance feature can extract non-stationary disturbance components through wavelet packet decomposition, Hilbert-Huang transform, or deep autoencoder. The fatigue damage disturbance spectrum can be a frequency domain damage characterization map formed by integrating the local offset of multi-resonance features and the structural damage disturbance feature. It can be used to provide spectral visualization of the spatial distribution and intensity of damage, supporting subsequent evolution analysis and prediction. In this embodiment, the fatigue damage disturbance spectrum can be derived from the structural health status perception map and serve as the input basis for deep feature analysis and full-cycle tracing.

[0025] Step S300: Perform deep feature analysis and node damage mutation feature analysis on the fatigue damage disturbance spectrum map to generate the damage evolution dynamic type of the mutation node.

[0026] Among these features, deep features can be high-order abstract features extracted from fatigue damage perturbation spectrum maps through nonlinear mapping. These features can be used to capture the implicit patterns of damage evolution and improve classification and prediction accuracy. In a specific embodiment, deep features can be extracted hierarchically from the spectrum map using convolutional neural networks, graph neural networks, or self-attention mechanisms. Nodal damage mutation features can be feature identifiers of abrupt changes in damage state at specific structural nodes. These can be used to identify critical damage points that are about to rapidly expand, triggering early warning mechanisms. Furthermore, nodal damage mutation features can be identified by detecting the time derivative or second-order difference mutation points of deep features. The damage evolution dynamic type of mutation nodes can be a classification description of the damage development path of mutation nodes. This can be used to guide the selection of subsequent intervention strategies and distinguish between reversible drift and irreversible crack propagation.

[0027] Step S400: Perform full-cycle time-series drift tracing on the fatigue damage disturbance spectrum map, and predict potential damage nodes based on the dynamic type of damage evolution to identify other potential damage propagation ports.

[0028] The full-cycle time-series drift can be the complete time evolution trajectory of all resonant frequencies or other key parameters from equipment commissioning to the present. It can be used to reveal long-term patterns in damage accumulation, supporting trend extrapolation and failure point prediction. In one specific embodiment, the full-cycle time-series drift can reconstruct the time series from long-term stored historical monitoring data, followed by denoising and alignment. Potential damage nodes can be structural locations that have not yet shown obvious anomalies but are at high risk of evolving into damage. They can be used to identify areas requiring key monitoring or intervention in advance, enabling proactive maintenance. Furthermore, potential damage nodes can be probabilistically predicted based on the dynamic type of damage evolution and stress field simulation results. Other potential damage propagation ports can be structural connections or boundary locations, other than identified abrupt change nodes, that may become paths for further damage propagation. These can be used to clarify the possible direction and exit of damage propagation, providing boundary conditions for vector field modeling.

[0029] Step S500: Perform multi-time point damage evolution simulation and damage disturbance diffusion coupling on other potential damage propagation ports to construct a multidimensional fatigue damage propagation trajectory vector field.

[0030] Multi-time point damage evolution simulation can be used to numerically simulate potential damage nodes across multiple time slices to predict their development path. This can be used to generate a temporal evolution sequence of damage, supporting the construction of the trajectory vector field. In a specific embodiment, multi-time point damage evolution simulation can employ the finite element method combined with the Paris crack propagation law or cohesion model for time-series simulation. Damage perturbation diffusion coupling can model the process by which different damage sources influence each other through the structural medium and jointly change the overall resonance characteristics. This can be used to reflect the nonlinear interaction effects under multiple damage coexistence, improving the realism of the vector field. Furthermore, damage perturbation diffusion coupling can solve the spatial transfer function of perturbation energy by establishing the coupling equation between the damage field and the vibration field. The multidimensional fatigue damage propagation trajectory vector field can be a vector distribution model describing the direction, rate, and intensity of damage propagation in space. This can be used to provide physical guidance for excitation intervention, clarifying the location and direction of the required reverse perturbation. In this embodiment, the multidimensional fatigue damage propagation trajectory vector field can be jointly constructed by multi-time point damage evolution simulation and damage perturbation diffusion coupling, serving as input for multi-strategy excitation intervention deduction.

[0031] Step S600: Perform multi-strategy excitation intervention simulation on the multidimensional fatigue damage propagation trajectory vector field, then optimize the excitation control decision, and construct an optimization strategy for remote adjustment of excitation frequency.

[0032] The multi-strategy excitation intervention simulation can simulate the suppression effect of various excitation schemes on damage propagation based on a multidimensional fatigue damage propagation trajectory vector field. This can be used to screen the optimal intervention strategy and avoid secondary damage caused by blind adjustment. In a specific embodiment, the multi-strategy excitation intervention simulation can inject excitation signals of different frequencies, amplitudes, and phases into a digital twin model to observe the change response of the damage vector field. Excitation control decision optimization can be the process of selecting the optimal combination of excitation parameters based on the results of the multi-strategy excitation intervention simulation. This can be used to ensure that the remote adjustment strategy achieves a balance between safety, effectiveness, and energy consumption. Furthermore, the excitation control decision optimization can use multi-objective optimization algorithms (such as NSGA-II) or reinforcement learning strategies to evaluate the comprehensive score of each scheme. The excitation frequency remote adjustment optimization strategy can be the final generated, remotely executable excitation frequency adjustment instruction set, which can be used to dynamically adjust the sensor operating point, suppress damage propagation, and restore resonance stability. In this embodiment, the excitation frequency remote adjustment optimization strategy can be output by the excitation control decision optimization and applied to the gas flow meter resonant sensing unit to achieve closed-loop control.

[0033] Taking an unmanned urban gas pressure regulating station as an example, the gas flow meter abnormality early warning and remote regulation control method in this embodiment can be to deploy an intelligent flow meter equipped with this method in a gas pressure regulating station in a remote area. The system continuously collects multi-dimensional time-series data of the resonant sensing unit and automatically constructs a structural health status perception map. When a slight frequency shift is detected in a weld area and the disturbance spectrum shows energy leakage characteristics, the system determines that a microcrack has started and confirms that the shift is accelerating through full-cycle tracing. Subsequently, it predicts that the downstream flange connection is a potential expansion port and constructs a damage expansion vector field to show that the crack will extend along the axial direction. After the system deduces various excitation strategies, it selects to apply a small-amplitude frequency sweep excitation near 125Hz. After optimization, this strategy is remotely sent to the field controller. The excitation causes a slight closing effect on the crack surface, and the resonant frequency gradually returns to the reference value, successfully avoiding the risk of sudden fracture and ensuring continuous gas supply to the pipeline network.

[0034] In one embodiment, multi-dimensional time-series data of the gas flow meter resonant sensing unit are acquired, resonant characteristic information is analyzed, and structural state logic hierarchy is reconstructed to construct a structural health status perception map of the flow meter sensor pipeline, including: Multi-dimensional resonant timing data stream is obtained based on the excitation drive unit and sensing detection unit of the gas flow meter; The excitation drive unit can be an actuator in the gas flow meter used to apply an excitation signal to the resonant sensing unit to maintain its stable resonant state. It can provide controllable electromagnetic or piezoelectric excitation, causing the sensor pipeline to vibrate continuously near its natural frequency. In an exemplary embodiment, the excitation drive unit can output a current with a specific frequency and amplitude to drive the coil or piezoelectric element through a closed-loop control circuit. The sensing detection unit can be a sensing module used to monitor the vibration state of the resonant sensing unit in real time and output electrical signals. It can be used to collect vibration displacement, velocity, or acceleration information and feed it back to the control system to maintain resonance or for state assessment. Furthermore, the sensing detection unit can convert mechanical vibration into a voltage signal based on the principles of electromagnetic induction, capacitance change, or optical interference. The multi-dimensional resonant timing data stream can be multi-channel time-series data reflecting the dynamic behavior of the resonant system, synchronously output by the excitation drive unit and the sensing detection unit. It can be used to provide a complete description of the resonant process, including the coupling relationship between drive and response, enhancing the observability of the internal state of the system. In this embodiment, the multi-dimensional resonant timing data stream can be synchronously sampled by a high-precision ADC, including drive current, feedback voltage, phase difference, and frequency tracking signal.

[0035] Heterogeneous feature information is analyzed on the multi-dimensional resonant timing data stream to generate heterogeneous resonant feature information, which includes the resonant frequency of the pipeline, the driving phase difference, the feedback voltage of the excitation coil, and the resonant quality factor Q value. Heterogeneous feature information analysis can be a process of uniformly extracting features from resonant signals from different physical mechanisms, dimensions, and sampling characteristics, transforming raw heterogeneous data into a set of physically meaningful features that can be jointly analyzed. In a specific embodiment, heterogeneous feature information analysis can use signal processing and system identification methods to analyze the key parameters of each channel signal separately. Heterogeneous resonance feature information can be a set of features extracted from multi-dimensional resonance time-series data streams that have different physical meanings but jointly characterize the resonance state. It can comprehensively depict the dynamic characteristics of the resonant system and improve the sensitivity to micro-damage. The pipeline resonant frequency can be the dominant resonant frequency of the sensor pipeline under the current structural state. It can be used to directly reflect changes in structural stiffness and mass distribution and is one of the most sensitive indicators of damage.

[0036] The driving phase difference, which can be the phase offset between the excitation driving signal and the sensing detection signal, can be used to characterize the system's damping characteristics and energy transfer efficiency, and is sensitive to the crack closure / opening state. The excitation coil feedback voltage, which can be the back electromotive force voltage signal generated by the electromagnetic induction of the excitation coil during vibration, can be used to indirectly reflect the vibration amplitude and velocity, and can be used for non-contact state monitoring. The resonance quality factor Q value, a dimensionless parameter measuring the ratio of energy loss to energy storage in a resonant system, is highly sensitive to minute dissipation mechanisms such as internal material friction and crack interface friction. Heterogeneous feature information analysis of the multi-dimensional resonant time-series data stream can generate heterogeneous resonance feature information by applying targeted algorithms to extract physical feature parameters for each channel signal. Furthermore, this operation can be achieved by extracting the phase difference through phase-locked amplification of the driving current and feedback voltage, and outputting the real-time resonant frequency through a frequency tracking loop, thereby transforming the original voltage / current signal into a resonant state index with clear physical meaning.

[0037] The heterogeneous resonance characteristic information is time-stamp aligned and numerically standardized to construct multidimensional structural state standardized features. The timestamp alignment process can be a time synchronization calibration operation for resonant feature data acquired asynchronously from multiple sources. This can be used to eliminate feature misalignment caused by sampling clock deviations, ensuring that multidimensional features are correlated under a unified time reference. In this embodiment, timestamp alignment can be achieved using hardware-triggered synchronization or software interpolation alignment (such as linear interpolation or spline interpolation). Numerical standardization can be a preprocessing operation that converts resonant features of different dimensions and magnitudes into a unified numerical range. This can be used to eliminate the interference of dimensional differences on subsequent correlation analysis and model training. In an exemplary embodiment, numerical standardization can apply methods such as Z-score standardization, Min-Max normalization, or RobustScaling. The multidimensional structural state standardized features can be a set of multidimensional resonant features after time alignment and numerical standardization, constituting a unified input space for structural health assessment. This can be used to provide a homogeneous and comparable data foundation for causal analysis and logical reconstruction. In this embodiment, the multidimensional structural state standardized features are generated from heterogeneous resonant feature information through timestamp alignment and numerical standardization, serving as direct input for multi-feature correlation analysis.

[0038] Multi-feature correlation analysis was performed on the standardized features of multidimensional structural states to extract the causal relationships between different features. Multi-feature correlation analysis can be an analytical process that mines statistical dependencies or physical causal relationships between standardized features of multidimensional structural states. It can be used to reveal the intrinsic coupling mechanisms between different resonant parameters and identify damage-sensitive feature combinations. In a specific embodiment, multi-feature correlation analysis can employ Granger causality tests, transfer entropy, structural equation modeling, or graph neural networks to learn causal graphs. Causal relationships can be directional physical or statistical dependencies between multidimensional resonant features, which can be used to provide a basis for logical hierarchy reconstruction and distinguish between principal and response features.

[0039] We perform structural dimension mining on the standardized features of multidimensional structural states and reconstruct the logical hierarchy based on causal relationships to construct a structural health status perception map of the flow meter sensor pipeline.

[0040] Structural dimension mining can identify the feature subspaces or key dimensions most sensitive to structural damage from a multidimensional feature space. This can be used to reduce redundancy and focus on core variables that truly reflect the structural state. In an exemplary embodiment, structural dimension mining can assess feature importance through principal component analysis, mutual information filtering, or attention mechanisms. Logical hierarchy reconstruction can organize multidimensional features into a hierarchical state representation model based on causal relationships. This can be used to construct interpretable health state mappings, supporting reasoning from phenomena to mechanisms. In this embodiment, logical hierarchy reconstruction can construct a tree-like or graph-like logical structure based on a causal graph, aggregating lower-level features into higher-level state nodes.

[0041] Structural dimension mining of standardized features of multidimensional structural states can assess the contribution of each feature to changes in structural state and screen key dimensions. Furthermore, this operation can be achieved by ranking feature importance based on SHAP values ​​and selecting the optimal feature subset through maximizing mutual information, thereby focusing on damage-sensitive features and improving model efficiency and robustness. Logical hierarchical reconstruction based on causal relationships can aggregate lower-level features into higher-level state nodes according to causal direction, forming a hierarchical representation. Further, this operation can be achieved by constructing a Bayesian network as the logical hierarchy and organizing state nodes using a knowledge graph, thus enabling a leap from data features to structural health semantics. Constructing a structural health state perception map of flowmeter sensor pipelines can be achieved by visualizing or encoding the logical hierarchical reconstruction results into a computable graph structure. Further, this operation can be achieved by generating a node-edge graph database record and outputting a spatial state distribution in the form of a heatmap, thereby providing a highly sensitive and interpretable representation of early damage, laying the foundation for subsequent early warning and control.

[0042] Taking a high-precision gas metering station for trade settlement as an example, the gas flow meter abnormality early warning and remote adjustment control method in this embodiment can be to deploy the intelligent flow meter of this solution at the natural gas gate station. The system synchronously collects the current, feedback voltage and displacement signal of the excitation drive coil and the sensing coil, and analyzes the heterogeneous features such as resonant frequency, Q value and phase difference in real time. After timestamp alignment and standardization, it is found that the Q value slowly decreases while the frequency remains stable. Multi-feature correlation analysis reveals that there is a strong causal relationship between the change in Q value and the nonlinear distortion of the phase difference. Structural dimension mining confirms that this combination is a sensitive indicator of weld microcracks. Logical level reconstruction maps this state to the microcrack initiation layer. The constructed structural health status perception map shows that the risk level of local areas has increased but has not reached the alarm threshold in the background. Based on this, the system arranges preventive maintenance in advance to avoid trade metering disputes and transmission losses caused by sudden fractures.

[0043] In one embodiment, the local offset amplitude of multi-resonance features is calculated from the structural health status perception map, and structural damage disturbance features are mined to construct a fatigue damage disturbance spectrum map, including: The sensor pipeline structure health status perception spectrum is decomposed into time-series states to generate multiple continuous resonance detection windows.

[0044] The sensor pipeline structural health status perception map can be a time-space joint map characterizing the structural health status of a gas flow meter sensor pipeline. It can serve as the input basis for time-series state decomposition, supporting fine-grained analysis of the resonant drift dynamic process. Time-series state decomposition involves dividing the structural health status perception map into several continuous segments along the time dimension. This can be used to achieve segmented modeling of the long-term resonant state evolution process, improving the ability to detect local anomalies. In an exemplary embodiment, time-series state decomposition can be described in context using sliding windows, change point detection, or adaptive segmentation algorithms to segment the time-series health status. Multiple continuous resonant detection windows can be a series of adjacent time sub-intervals covering the entire monitoring period obtained after time-series state decomposition. These can be used to provide independent analysis units for calculating local offset amplitudes, supporting dynamic drift trajectory reconstruction.

[0045] The local offset amplitude of the multi-resonance characteristics is calculated for multiple continuous resonance detection windows to obtain the resonance offset amplitude value of each window.

[0046] The resonance offset amplitude value can be the comprehensive offset of multiple resonance features calculated within each resonance detection window relative to the baseline state. This can be used to quantify the degree of structural anomaly within that time window, serving as a direct basis for distortion identification. Calculating the local offset amplitude of multiple resonance features for multiple consecutive resonance detection windows can involve calculating the offset of features such as resonant frequency and damping relative to a healthy baseline within each window. Furthermore, this operation can be achieved by using Euclidean distance to measure the multidimensional feature offset and Mahalanobis distance to consider the feature covariance structure, thereby obtaining structural anomaly intensity indices for each time period, supporting distortion identification. Obtaining the resonance offset amplitude value for each window can be achieved by outputting the comprehensive offset value calculated for each window. Further, this operation can be achieved by outputting a single dominant frequency offset value or an output of the comprehensive offset index after weighted fusion of multiple features, thereby forming a scalar sequence that can be used for time-series anomaly detection.

[0047] Transient large displacement distortion is identified based on the resonance offset amplitude value, and structural distortion points are marked.

[0048] Transient large-amplitude distortion can be a significant and non-stationary resonance offset anomaly occurring within a single or a few consecutive windows. It can be used to indicate early structural damage signals that may be caused by microcrack initiation or stress concentration. Structural state distortion points can be specific time locations or state nodes marked after transient large-amplitude distortion identification. These can be used as candidate points for subsequent damage risk assessment, narrowing the analysis scope and improving efficiency. Transient large-amplitude distortion identification based on resonance offset amplitude values ​​can detect significant non-stationary abrupt changes in the offset amplitude value sequence. Furthermore, this operation can be achieved by using the 3σ principle to identify outliers and applying wavelet coefficient maxima to detect transient events, thus effectively separating transient anomalies caused by actual damage from normal operating condition fluctuations. Marking structural state distortion points can be done by recording the timestamps or state indices corresponding to the identified transient distortions. Furthermore, this operation can be achieved by marking on the time axis and associating it with the original data, and setting distortion state labels on the graph nodes, thus providing a clear set of candidate points for subsequent risk assessment.

[0049] Based on a preset fatigue damage sensitivity threshold, the damage risk sensitivity of structural state distortion points is assessed, and potential damage perception factors are extracted.

[0050] The preset fatigue damage sensitivity threshold can be a criterion parameter pre-set based on material properties, operating history, and failure models to assess the risk level of distortion points. This threshold allows for quantitative screening of damage probabilities at different distortion points, avoiding treating all fluctuations the same. In an exemplary embodiment, the preset fatigue damage sensitivity threshold can be derived from laboratory accelerated fatigue test data or field failure statistical fitting, and can be dynamically adjusted according to operating conditions. Damage risk sensitivity assessment is a process of determining the risk level of structural distortion points based on the fatigue damage sensitivity threshold. This can be used to distinguish between high-risk real damage and low-risk operating condition disturbances, improving the accuracy of potential damage factor extraction. For example, damage risk sensitivity assessment can use fuzzy membership functions to map risk levels and Bayesian posterior probability to calculate damage probability. Potential damage perception factors can be feature vectors corresponding to structural distortion points with high damage probability confirmed after risk assessment. These can be used as core inputs for dynamic damage disturbance feature mining, representing candidate signal sources for early damage. In a specific embodiment, potential damage perception factors can be extracted from structural distortion points after damage risk sensitivity assessment and used to construct a fatigue damage disturbance spectrum.

[0051] Dynamic damage perturbation features are mined from potential damage sensing factors to construct a fatigue damage perturbation spectrum.

[0052] Among them, dynamic damage perturbation features can be non-stationary perturbation modes that reflect the evolution of damage over time and are further mined from potential damage sensing factors. They can be used to reveal the dynamic fingerprints of processes such as microcrack propagation or stress relaxation, enhancing the physical interpretability of the spectrogram. In an exemplary embodiment, the dynamic damage perturbation features can extract the perturbation evolution law through time-varying spectrum analysis, recursive graph reconstruction, or dynamic mode decomposition.

[0053] For example, in the scenario of a remote monitoring station for long-distance gas pipelines in high-altitude and cold regions, the gas flow meter anomaly early warning and remote adjustment control method of this embodiment can be as follows: In an extremely cold winter environment, a smart gas flow meter is continuously running. The system performs time-series state decomposition on its structural health status perception spectrum to generate a continuous resonance detection window every 10 minutes. After calculating the resonance offset amplitude value of each window, it is found that a transient large offset distortion occurs in a certain window between 3:00 and 4:00 AM, with an amplitude exceeding the normal thermal expansion and contraction range. The system marks this period as a structural state distortion point and evaluates it in conjunction with a preset low-temperature fatigue sensitivity threshold (considering the material embrittlement effect), confirming it as a high-risk damage signal. Subsequently, the multidimensional features of this period are extracted as potential damage perception factors, and its disturbance exhibits dynamic characteristics of frequency modulation and energy leakage coexisting. Finally, the constructed fatigue damage disturbance spectrum clearly shows that there are signs of microcrack initiation in this area, triggering subsequent prediction and intervention processes, thus avoiding metering failure caused by low-temperature brittle fracture.

[0054] In one embodiment, dynamic damage perturbation feature mining is performed on potential damage sensing factors to construct a fatigue damage perturbation spectrum, including: Periodic damage perturbation characteristics are analyzed for potential damage sensing factors, and periodic resonance offset characteristics are extracted.

[0055] The periodic damage disturbance feature analysis can be a specialized analysis process to determine whether there are resonant disturbance modes related to the fatigue load cycle in the potential damage sensing factors. It can be used to identify structural response anomalies induced by periodic conditions such as alternating fluid loads or temperature cycling, and to eliminate non-periodic random disturbances. In this embodiment, the periodic damage disturbance feature analysis can determine whether the migration sequence has significant periodicity through autocorrelation function, spectral peak detection, or periodogram method. The periodic resonant migration feature can be a resonant frequency migration mode synchronized with the external excitation cycle extracted from the periodic damage disturbance feature analysis. It can be used as key evidence of real fatigue damage, reflecting the cumulative deterioration behavior of the structure under repeated loads. Furthermore, the periodic resonant migration feature can be extracted from the potential damage sensing factors after periodic damage disturbance feature analysis and used for multi-time-point migration fitting.

[0056] Analyzing the periodic damage disturbance characteristics of potential damage sensing factors can help examine whether their time series exhibits a repetitive shift pattern related to the operating cycle. This can be further achieved by calculating the autocorrelation function and detecting significant peaks, and by performing a harmonic component significance test on the spectrum. This effectively filters out non-periodic noise and focuses on the structural response induced by the true fatigue mechanism. Extracting periodic resonant shift features can be achieved by separating the shift component synchronized with the main excitation cycle from the sensing factors that pass the periodicity test. In a specific embodiment, this can be achieved by using a bandpass filter to extract the fundamental frequency component and using lock-in amplification technology to extract in-phase / quadrature shift components. This yields physically interpretable damage features and improves the reliability of subsequent modeling.

[0057] Based on the periodic resonance offset characteristics, multi-time-point offset fitting is performed to construct the damage disturbance waveform curve.

[0058] Multi-timepoint migration fitting can be a process of continuously modeling periodic resonant migration values ​​observed at multiple time points using mathematical functions. This can be used to reconstruct discrete observation data into a smooth, continuous evolution trajectory, supporting the calculation of dynamic parameters. In this embodiment, multi-timepoint migration fitting can employ spline interpolation, local weighted regression, or Gaussian process regression to fit the migration sequence. The damage perturbation waveform curve can be a function curve generated after multi-timepoint migration fitting, describing the continuous change of resonant frequency drift over time. This can be used to provide a high-fidelity time-domain characterization of damage evolution, serving as the basis for intensity and slope calculations.

[0059] Multi-time-point migration fitting based on periodic resonance migration characteristics can approximate the migration observations over multiple periods using a continuous function. Furthermore, this operation can be achieved by connecting the peak points of each period using cubic spline interpolation and modeling the continuous trajectory under uncertainty using Gaussian process regression, thereby reconstructing the complete drift dynamic process and avoiding information loss due to sparse sampling. Constructing the damage perturbation waveform curve can be achieved by outputting the results of the multi-time-point migration fitting, forming a continuous function representation of time-offset. In a specific embodiment, this operation can be achieved by generating a curve in analytical expression form and outputting a discrete but highly sampled numerical curve, thus providing a smooth and differentiable input basis for intensity and slope calculations.

[0060] Calculate the drift frequency intensity and trend slope of the damage disturbance waveform curve.

[0061] The drift frequency intensity can be a measure of the amplitude of the resonant frequency shift of the damage perturbation waveform curve within a specific time period. It can be used to quantify the energy level or severity of the damage perturbation and reflect the activity of microcrack propagation. In this embodiment, the drift frequency intensity can be obtained by calculating the local extreme value difference, root mean square offset, or envelope amplitude of the curve. The trend slope can be the derivative of the damage perturbation waveform curve in the time dimension or the local linear approximation slope. It can be used to characterize the rate of damage evolution and determine whether it is stable aging or accelerated failure. Furthermore, the trend slope can be obtained by differentiating the fitted curve or by estimating the instantaneous rate of change using sliding window linear regression. Calculating the drift frequency intensity and trend slope of the damage perturbation waveform curve can be achieved by quantifying the amplitude characteristics and rate of change of the curve, respectively. In a specific embodiment, this operation can be achieved by calculating the intensity using envelope analysis and differentiating it using a Savitzky-Golay filter to obtain the slope, using wavelet coefficient energy to measure the intensity and combining it with local linear regression to estimate the slope, thereby obtaining two key dynamic parameters describing the severity and development speed of the damage.

[0062] Drift frequency continuity is mined based on drift frequency intensity and trend slope to construct a fatigue damage disturbance spectrum.

[0063] Drift frequency continuity mining can be used to analyze the duration, acceleration, or intermittent patterns of frequency drift over time based on drift frequency intensity and trend slope. This can reveal the dynamic mechanism of damage development and distinguish between reversible drift and irreversible crack propagation. In this embodiment, drift frequency continuity mining can assess drift behavior through state transition modeling, pattern clustering, or continuity indices (such as the Hurst index). Drift frequency continuity mining based on drift frequency intensity and trend slope can comprehensively consider the time-varying characteristics of intensity and slope to determine the continuity pattern of drift behavior. Furthermore, this operation can be achieved by constructing a two-dimensional state space (intensity-slope) and performing clustering to identify patterns, and using a hidden Markov model to infer continuous state transitions. This can identify whether the damage is in an accelerated propagation stage, providing a basis for intervention timing. Constructing a fatigue damage perturbation spectrum can be achieved by organizing the drift frequency continuity mining results into a spectrum according to frequency-time or frequency-intensity. In one specific embodiment, this operation can be achieved by generating a heatmap with drift frequency as the horizontal axis, intensity as the vertical axis, and slope color-coded, and constructing a three-dimensional spectral cube containing time-frequency-intensity dimensions. This can generate a high-dimensional damage characterization of fusion evolution rate, perturbation intensity, and temporal continuity, supporting accurate prediction and control.

[0064] For example, in the scenario of high-frequency peak shaving at a gas metering station in an industrial area, the abnormal early warning and remote adjustment control method for gas flow meters in this embodiment can be as follows: A gas flow meter experiences severe flow fluctuations during the morning and evening peak hours each day. The system identifies a resonant offset feature that is synchronized with the peak shaving cycle (12 hours) from potential damage sensing factors. Through multi-time-point offset fitting, a damage disturbance waveform curve with a sawtooth-like upward trend is constructed. Calculations show that its drift frequency intensity increases day by day, and the trend slope increases from the initial 0.02Hz / day to 0.08Hz / day. Continuity mining determines that the drift is an accelerated continuous mode, indicating that microcracks are rapidly propagating under alternating stress. Based on this, the fatigue damage disturbance spectrum constructed by the system shows a high-intensity, high-slope continuous spectrum near 240Hz, triggering an advanced early warning and initiating the excitation intervention process, successfully restoring resonant stability before crack penetration.

[0065] In one embodiment, deep feature analysis and nodal damage mutation feature analysis are performed on the fatigue damage disturbance spectrum to generate the damage evolution dynamic type of the mutation node, including: Deep feature analysis is performed on the fatigue damage disturbance spectrum, and damage feature annotation and encoding are performed to obtain the damage disturbance feature code; The damage perturbation feature encoding can be a structured damage representation vector with semantic labels generated from the fatigue damage perturbation spectrum map after deep feature analysis. This vector can be used to achieve semantic mapping from the original spectral perturbation to interpretable damage patterns, supporting subsequent spatial localization and path mining. In this embodiment, the damage perturbation feature encoding can be obtained by extracting features from the spectrum map using a deep neural network and then embedding or classifying them using predefined damage semantic labels. Deep feature analysis of the fatigue damage perturbation spectrum map can be performed by extracting high-order nonlinear features from the spectrum map using a deep learning model. Further, this operation can be achieved by using a convolutional neural network to extract local frequency band features and using a graph attention network to model the dependencies between frequency points, thereby capturing the implicit patterns of damage perturbation and providing a foundation for semantic encoding. Damage feature annotation encoding to obtain the damage perturbation feature encoding can be achieved by mapping deep features to a predefined damage semantic space to generate structured encoding. In an exemplary embodiment, this operation can be achieved by outputting discrete damage type labels using a classification head and generating continuous semantic embedding vectors using an autoencoder, thus giving the spectral perturbation physical interpretability and supporting subsequent spatial correlation.

[0066] Based on the damage disturbance feature encoding, the sensor pipeline structure health status perception map is correlated and mapped and the topology is located to obtain the map location information of the damage factor; The sensor pipeline structure health status perception map can be a health status representation model constructed by fusing multi-dimensional resonance features and logical hierarchy. It can serve as a topological basis for spatial localization of damage factors, providing a joint reference frame for structural geometry and state semantics. In one specific embodiment, the sensor pipeline structure health status perception map can receive damage disturbance feature encoding as query input and output the specific location information of the damage factor in the map. The map location information of the damage factor can be the spatial node or region identifier corresponding to the damage disturbance feature in the structural health status perception map, which can be used to accurately anchor the physical location of the damage in the sensor pipeline, solving the problem of ambiguous anomaly localization. In this embodiment, the map location information of the damage factor can be obtained by comparing the damage disturbance feature encoding with the map node features using a graph matching algorithm to return the best matching position.

[0067] The health status perception map of the sensor pipeline structure is correlated and mapped based on the damage perturbation feature encoding, and the topological structure is located. This can be achieved by searching for the position that best matches the damage encoding in the node feature space of the map. For example, this operation can be implemented by using a graph neural network for cross-domain alignment and using cosine similarity to retrieve the best match in the map node library, thereby achieving a precise mapping of damage from frequency domain representation to physical space. The map location information of the damage factor is obtained, which can be the optimal matching node or region identifier in the output correlation mapping result. In an exemplary embodiment, this operation can be achieved by returning a single high-confidence node ID and outputting the probability distribution of multiple candidate positions, thus clearly defining the specific spatial coordinates of the damage in the sensor pipeline.

[0068] Damage propagation path mining is performed based on map location information to extract damage propagation paths; The damage propagation path can be a sequence of potential conduction channels that may spread along the structural medium from the damage initiation point. This can be used to reveal the spatial evolution trajectory of microcracks from initiation to propagation, providing directional guidance for intervention. In this embodiment, the damage propagation path can be obtained by mining high-probability propagation links based on graph location information, combined with stress transfer models or graph theory shortest path algorithms. Mining damage propagation paths based on graph location information can start from a location node and extrapolate possible propagation paths along the structural topology or stress field direction. Furthermore, this operation can be achieved by tracing paths based on finite element stress gradient directions and calculating the maximum flow path using graph edge weights, thereby revealing the spatial channels through which damage may propagate in the future. Extracting the damage propagation path can be done by organizing the mining results into an ordered node sequence or edge chain. In a specific embodiment, this operation can be achieved by outputting a path node list and generating a path adjacency matrix, thus forming structured path data that can be used for key node analysis.

[0069] A path abrupt change key node analysis was performed on the damage propagation path to extract microcrack initiation nodes, fatigue accumulation nodes, and low-frequency activation nodes of stress concentration in the path. Among them, the key nodes of path mutation can be the set of structural nodes that have a decisive influence on the evolution trend in the damage propagation path. They can be used to focus analysis resources on the most risky locations, improving the efficiency and accuracy of mutation identification. Microcrack initiation nodes can be the structural locations where microcracks first appear in the damage propagation path. They can be used to identify the source of damage, trace the root cause, and prioritize intervention. In this embodiment, microcrack initiation nodes can be identified using a joint criterion of high-frequency perturbation energy concentration and local strain gradient. Fatigue accumulation nodes can be structural regions in the damage path where energy continuously accumulates due to long-term cyclic loading. They can be used to reflect the intermediate stage of damage evolution and characterize the acceleration potential of crack propagation. In an exemplary embodiment, fatigue accumulation nodes can be identified based on the energy dissipation accumulation rate during full-cycle time-series drift.

[0070] Low-frequency activation nodes of stress concentration can be weak points at geometrical discontinuities in a structure, activated by low-frequency external excitation, becoming springboards for damage propagation. They can be used to reveal the risk of secondary damage caused by the coupling of external vibration disturbances and structural weaknesses. In this embodiment, low-frequency activation nodes of stress concentration can be obtained by jointly determining the vibration response amplitude in the low-frequency band (e.g., 0.1–10Hz) and the local stress concentration coefficient. Path abrupt change key node analysis of the damage propagation path can identify three types of nodes that play a crucial role in the evolution of the path. In a specific embodiment, this operation can be achieved by filtering based on the frequency of abrupt changes in the node's historical state and combining it with the material fatigue limit threshold, thereby focusing on high-value analysis objects and improving the targeting of abrupt change identification. Extracting microcrack initiation nodes, fatigue accumulation nodes, and low-frequency activation nodes of stress concentration in the path can be achieved by classifying and extracting three types of key nodes from the path nodes according to various custom criteria. For example, this operation can be achieved by running the three criterion models in parallel for extraction and using multi-task learning for unified identification, enabling semantic decomposition of the damage path and supporting differentiated abrupt change analysis.

[0071] We analyze the abrupt change characteristics of microcrack initiation nodes, fatigue accumulation nodes, and low-frequency activation nodes of stress concentration to generate dynamic damage evolution types of abrupt change nodes.

[0072] The node state mutation features can be quantitative indicators of the abrupt changes in the state of three types of key nodes over time, serving as direct input for generating damage evolution dynamic types and distinguishing different evolution stages. In this embodiment, node state mutation features can be obtained by calculating the time derivative, second-order difference, or mutation index of the node resonance parameters. Analyzing the node state mutation features of microcrack initiation nodes, fatigue accumulation nodes, and low-frequency activation nodes due to stress concentration can be achieved by calculating the state mutation indices for each of the three types of nodes separately. In an exemplary embodiment, this operation can be implemented by using a customized mutation detection algorithm for each type of node and uniformly using a general mutation index for evaluation, thereby quantifying the current evolutionary activity of each type of node. Generating the damage evolution dynamic type of mutated nodes can be achieved by comprehensively considering the mutation features of the three types of nodes and outputting a type label describing the overall evolution stage and trend. For example, this operation can be implemented by combining mutation features based on a rule engine to generate types and using a classification model to predict evolution types end-to-end, thus providing a semantic description of damage evolution supported by physical mechanisms.

[0073] Taking a smart metering station on a high-pressure gas trunk line in a city as an example, the gas flow meter abnormality early warning and remote adjustment control method in this embodiment can be as follows: when the system analyzes the fatigue damage disturbance spectrum of a Coriolis flow meter, it obtains the damage disturbance feature code through deep feature analysis. After the code is associated and mapped with the structural health status perception map, it is accurately located in the weld area at the bottom of the U-shaped measuring pipe. Based on this location, a damage propagation path extending along the pipe wall to the support clamping end is extracted. The path analysis identifies three key nodes: the root of the weld is the microcrack initiation node (high frequency disturbance energy concentration), the middle straight pipe section is the fatigue accumulation node (the energy dissipation rate is the highest throughout the cycle), and the area near the clamp is the stress concentration low-frequency activation node (significant response under low-frequency vibration during pipeline start-up and shutdown). The abrupt change feature analysis of the three nodes reveals that the frequency step of the initiation node is obvious but the expansion is slow, the damping of the accumulation node continues to rise, and the activation node has recently experienced phase lockout. Based on this, the system generates a damage evolution dynamic type of "stable microcrack with local accelerated propagation", triggers a targeted early warning, and suggests prioritizing the inspection of the end of the path in the next maintenance window.

[0074] In one embodiment, the fatigue damage perturbation spectrum is traced back through a full-cycle temporal drift, and potential damage nodes are predicted based on the dynamic type of damage evolution to identify other potential damage propagation ports, including: Multi-feature correlation mining is performed on the fatigue damage disturbance spectrum to identify associated damage feature groups.

[0075] The associated damage feature group can be a set of damage characterization features with inherent physical coupling relationships identified from the fatigue damage perturbation spectrum map through multi-feature association mining. It can be used to integrate originally isolated heterogeneous features such as frequency shift, phase perturbation, and energy attenuation into a joint damage identifier with structural mechanical significance, improving the semantic consistency and physical interpretability of damage identification. In this embodiment, the associated damage feature group can be obtained by establishing feature co-occurrence and causal dependence relationships between different frequency bands, time windows, or spatial locations in the spectrum map based on graph association rule mining, tensor decomposition, or attention mechanisms. Multi-feature association mining of the fatigue damage perturbation spectrum map to identify the associated damage feature group can be achieved by mining feature combinations with statistical significance or physical coupling in the multi-dimensional feature space of the fatigue damage perturbation spectrum map. Furthermore, this operation can be achieved by using tensor low-rank decomposition to extract cross-band-time coupling patterns and using graph autoencoders to learn association edge weights between spectrum nodes, thereby overcoming the limitations of single-signal dimension analysis, realizing semantic fusion of heterogeneous perturbation features, and enhancing the comprehensive discrimination capability for weak early damage.

[0076] Deeply deconstruct the damage propagation logic of the associated damage feature group to generate damage extension propagation logic.

[0077] The damage propagation transmission logic can be a mechanistic model describing how damage expands spatially within a structure through stress transfer, energy diffusion, or boundary reflection. It can reveal the intrinsic transmission path between the crack initiation point and the distal response, providing a theoretical basis for predicting the propagation direction. In an exemplary embodiment, the damage propagation transmission logic can be obtained by extracting the transmission weights and directions between nodes through graph neural network inference or a rule engine based on fracture mechanics analysis of associated damage feature groups. Deep deconstruction of the damage propagation logic of associated damage feature groups to generate the damage propagation transmission logic can involve analyzing the causal or transmission relationships between the components within the associated damage feature groups and constructing a structured propagation mechanism model. In a specific embodiment, this operation can be achieved by constructing a symbolic inference rule base based on prior knowledge of fracture mechanics for logical deconstruction and using causal discovery algorithms (such as PC algorithms) to infer the directed dependencies between features. This transforms data-driven feature associations into interpretable physical transmission logic, improving the mechanistic basis and reliability of predictions.

[0078] Based on the damage propagation logic, the damage propagation path is traced throughout the entire time cycle to obtain the full-cycle damage propagation chain.

[0079] The damage propagation path can be a set of specific routes that damage may spread in the structural geometry, used to identify key areas that need to be covered by intervention and avoid blind spots in regulation. In this embodiment, the damage propagation path can be generated by combining the structural topology model and the damage propagation transmission logic, using the shortest path algorithm or field theory method to generate potential propagation trajectories. The full-cycle damage propagation chain can be a complete evolutionary sequence from the initial damage to the current state formed by integrating the damage propagation transmission logic and the full-cycle time-series tracing results. It can be used to construct a three-in-one damage cognition framework of history-current situation-trend, supporting high-confidence prediction. Furthermore, the full-cycle damage propagation chain is jointly generated by the damage propagation transmission logic and the full-cycle time-series tracing, serving as the core input for predicting potential damage nodes. Based on the damage propagation transmission logic, the full-cycle damage propagation path is traced back to its full-cycle time-series, resulting in the full-cycle damage propagation chain. This can be achieved by tracing back and reconstructing the complete spatiotemporal evolution process of the damage from its inception to the present along the direction and mechanism defined by the damage propagation transmission logic. Furthermore, this operation can be achieved by reconstructing historical states through inverse integration along the propagation logic on the finite element mesh and by using a hidden Markov model to complete the state path of discrete observation points, thereby forming a damage evolution archive with temporal continuity and spatial coherence, supporting forward-looking prediction.

[0080] Based on the dynamic types of damage evolution and the full-cycle damage propagation chain, potential damage nodes are predicted, and other potential damage propagation ports are identified.

[0081] Predicting potential damage nodes based on damage evolution dynamic types and full-cycle damage propagation chains, and identifying other potential damage propagation ports, can be achieved by fusing known damage evolution dynamic type classification results with the spatial topology of the full-cycle damage propagation chain to predict high-risk nodes that have not yet become explicit and their possible propagation exits. In an exemplary embodiment, this operation can be achieved by embedding the damage propagation chain into a graph neural network, combining evolution type labels for node risk scoring, and using a Bayesian network to fuse propagation probabilities and evolutionary trends for port prediction. This allows for high-confidence localization of latent damage and prediction of propagation boundaries, providing targets for precise intervention.

[0082] Taking the monitoring of key nodes in a high-pressure gas main pipeline as an example, the gas flow meter abnormality early warning and remote regulation control method in this embodiment can be as follows: Deploy the system of this embodiment in key sections of high-pressure gas pipelines that cross earthquake zones. When the fatigue damage disturbance spectrum shows weak energy leakage in multiple frequency bands, the system identifies a set of associated damage features containing 120Hz frequency drift, sudden increase in phase lag, and 350Hz harmonic enhancement through multi-feature correlation mining. Further analysis reveals that this set of features conforms to the logic of elastic wave propagation along the heat-affected zone of the weld. Based on this, the system performs full-cycle tracing of monitoring data from the past three years and reconstructs a damage propagation chain from the initial micropore to the current subcritical crack. Combining the fact that the damage belongs to the accelerated propagation type of evolution dynamics, the system predicts that the downstream flange sealing surface is a high-risk potential node and identifies its bolt hole edge as the most likely propagation port. This prediction result triggers an advance maintenance plan, avoiding gas outage accidents caused by sudden leaks.

[0083] In one embodiment, multi-time-point damage evolution simulation and damage perturbation diffusion coupling are performed on other potential damage propagation ports to construct a multidimensional fatigue damage propagation trajectory vector field, including: Identify the current structural health status of sensor pipelines based on multi-dimensional resonant timing data streams; The multi-dimensional resonant time-series data stream can be a continuous, real-time transmitted sequence of time-series signals from multiple physical dimensions of the resonant sensing unit. This can be used to support dynamic, high-fidelity identification of structural health status, overcoming the limitations of single-point or static sampling. In an exemplary embodiment, the multi-dimensional resonant time-series data stream can be streamed and transmitted through edge computing nodes to simultaneously acquire data from multiple sensors. The current structural health status of the sensor pipeline can be a representation of the sensor pipeline's instantaneous structural integrity identified based on the latest multi-dimensional resonant time-series data stream. This can be used as the initial condition for real-time stress analysis and damage simulation, ensuring that evolution predictions are consistent with the current physical state. Furthermore, the current structural health status of the sensor pipeline can be identified driven by the multi-dimensional resonant time-series data stream and used as the input basis for real-time dynamic stress distribution analysis. Identifying the current structural health status of the sensor pipeline based on the multi-dimensional resonant time-series data stream can be achieved by using a streaming processing framework to perform real-time feature extraction and state classification on the received multi-dimensional resonant data. Furthermore, this operation can be achieved by using lightweight convolutional neural networks for online inference and sliding window spectral matching for state retrieval, thereby enabling low-latency, high-fidelity perception of structural states and supporting subsequent dynamic analysis.

[0084] Real-time dynamic stress distribution analysis is performed on the current structural health status of the sensor pipeline to extract key stress characteristics; Real-time dynamic stress distribution analysis is a process of calculating the instantaneous stress field at various locations in a pipeline based on the current structural health state, combined with fluid loads and boundary conditions. It can be used to reveal the internal mechanical response of the structure under current operating conditions, providing a physical driving basis for damage evolution. In this embodiment, real-time dynamic stress distribution analysis can be solved quickly using a reduced-order finite element model or a data-physical hybrid surrogate model. Key stress characteristics can be a set of stress parameters extracted from the real-time dynamic stress distribution that have a dominant influence on damage evolution. These can be used to guide the loading conditions in multi-time-point damage evolution simulations, improving simulation realism. Furthermore, key stress characteristics can be obtained by selecting high-contribution stress components through principal component analysis or attention mechanisms.

[0085] Real-time dynamic stress distribution analysis of the current structural health status of the sensor pipeline can be performed by using the current health status as boundary and material parameter inputs to solve for the instantaneous stress field. Furthermore, this operation can be achieved by calling a pre-trained graph neural network surrogate model for rapid stress prediction and running an embedded finite element solver for local fine-grained calculations, thereby establishing a correlation between damage evolution and the real mechanical environment and enhancing the consistency of the simulation with physical reality. Extracting key stress characteristics can be achieved by selecting the components or combinations that have the most driving effect on fatigue damage from the high-dimensional stress field. Further, this operation can be achieved by selecting stresses in high-gradient regions based on sensitivity analysis and compressing the stress field to a low-dimensional feature space using an autoencoder, thereby reducing simulation complexity and focusing on the dominant damage mechanism.

[0086] Based on the key stress characteristics, multi-time-point damage evolution simulations were performed on other potential damage propagation ports to generate damage evolution simulation data; Damage evolution simulation data can be a set of numerical results obtained from multi-time-point simulations of potential damage propagation ports driven by key stress characteristics. It can provide quantitative evidence of damage development over time and support trajectory analysis. In this embodiment, the damage evolution simulation data can be obtained by iteratively calculating crack length and morphology at discrete time steps based on the Paris formula or cohesion model.

[0087] Multi-time-point damage evolution simulations of other potential damage propagation ports based on key stress characteristics can be performed at multiple future time points, using key stress characteristics as load inputs, to execute numerical simulations of crack propagation. Furthermore, this operation can be achieved by employing explicit time integration for transient crack simulation and using incremental step Paris law for long-term trend extrapolation, thereby generating a damage development sequence supported by physical mechanisms. Generating damage evolution simulation data can output crack geometry, stress intensity factors, and local field variables at each time step. Furthermore, this operation can be achieved by storing spatiotemporal field data in HDF5 format and serializing key node evolution records using JSON, thus providing a raw numerical foundation for trajectory analysis.

[0088] Damage trajectory evolution analysis is performed on damage evolution simulation data to generate damage evolution trajectories; Damage trajectory evolution analysis can be a process of extracting spatiotemporal paths and analyzing dynamic parameters from damage evolution simulation data. It can be used to transform discrete simulation results into continuously differentiable damage propagation trajectories. In this embodiment, damage trajectory evolution analysis is achieved by fitting the spatial path using spline interpolation and combining it with time derivatives to calculate the rate and acceleration. The damage evolution trajectory can be a continuous path describing the evolution of damage in space over time and its geometric shape, and can be used as a basic carrier for calculating dynamic parameters such as propagation direction and rate.

[0089] Damage trajectory evolution analysis of damage evolution simulation data can be performed by fitting a sequence of crack tip positions to construct a continuous spatiotemporal trajectory and analyzing its differential characteristics. Furthermore, this operation can be achieved by fitting the spatial trajectory with B-spline curves and smoothing noisy trajectory points using Kalman filtering, thereby transforming discrete simulation results into a continuous path that can be used to construct a vector field. Generating the damage evolution trajectory can be achieved by outputting a continuous path representation containing spatial coordinates and timestamps. Further, this operation can be achieved by expressing it as a parametric curve and storing it as a sequence of mesh vertices, thus forming the geometric framework of damage propagation and supporting the calculation of dynamic parameters.

[0090] Calculate the direction of expansion, rate of expansion, strain potential energy, and probability of fracture failure of the damage evolution trajectory; The propagation direction can be the tangential vector of the damage evolution trajectory at a local location, indicating the spatial direction of the next crack propagation. It can be used to determine the spatial orientation of the reverse perturbation required for excitation intervention. The propagation rate can be the distance the damage front advances per unit time, reflecting the speed of damage development. It is used to assess the urgency of intervention and participates in the assignment of vector field intensity. Strain potential energy can be the elastic strain energy density stored near the damage tip due to stress concentration. It can be used as an energy source to drive spontaneous crack propagation and affects the calculation of fracture failure probability. Fracture failure probability can be the statistical probability of complete fracture of the structure under the current stress and damage state. It can be used to quantify the risk level and participate in the weight allocation in damage perturbation diffusion coupling. In this embodiment, the fracture failure probability can be evaluated based on Weibull distribution or Monte Carlo simulation combined with the uncertainty of material fracture toughness.

[0091] Calculating the extension direction, extension rate, strain potential energy, and fracture failure probability of the damage evolution trajectory can be achieved by performing differential operations on the trajectory and combining local field variables to calculate four types of dynamic parameters. Furthermore, this operation can be implemented by calculating the direction and rate using finite difference methods and by combining J-integral and material toughness to calculate the fracture probability, thereby quantifying the multidimensional properties of damage extension and providing a basis for assigning values ​​to the vector field. Damage perturbation diffusion coupling based on the extension direction, extension rate, strain potential energy, and fracture failure probability can be achieved by using the four types of parameters as weights and vector components to construct a spatial perturbation propagation model. Furthermore, this operation can be achieved by constructing a weighted diffusion partial differential equation to solve the perturbation field and using a graph attention network to aggregate the influence of multiple parameters, thereby realizing multi-factor fusion damage influence field modeling and improving the physical realism of the vector field.

[0092] Based on the aforementioned expansion direction, expansion rate, strain potential energy, and fracture failure probability, damage perturbation diffusion coupling is performed to construct a multidimensional fatigue damage expansion trajectory vector field.

[0093] Constructing a multidimensional fatigue damage propagation trajectory vector field can be achieved by organizing the coupled perturbation field into a spatial vector distribution, with each location containing direction, intensity, and risk information. Furthermore, this operation can be implemented by storing the vector field in a regular grid and representing high-gradient regions in the form of unstructured point clouds, thereby forming a physical guidance field that can be directly used for excitation intervention simulation.

[0094] Taking a remote monitoring station for a high-pressure gas transmission and distribution trunk line as an example, the gas flow meter abnormality early warning and remote regulation control method in this embodiment can be deployed downstream of the key throttling valve of the high-pressure gas transmission trunk line. The system continuously receives multi-dimensional resonant time-series data streams and identifies a slight damping abnormality at the bottom of the sensor U-tube in real time. Then, dynamic stress analysis is initiated, and it is found that the area is subjected to significant alternating bending stress under start-up and shutdown conditions. The maximum principal stress amplitude is extracted as a key feature. For the potential expansion port at the upstream elbow connection, the system performs multi-time point simulation to generate the trajectory of crack expansion along the weld heat-affected zone in the next 72 hours. The trajectory analysis shows that the expansion direction is axially deviated by 30 degrees, the rate is 8 micrometers per day, the local strain potential energy is close to the critical value, and the probability of fracture failure reaches 12%. Based on this, the system constructs a vector field and selects to apply a 180Hz phase modulation excitation perpendicular to the expansion direction in the subsequent excitation simulation, successfully inducing micro-closure of the crack surface and stabilizing the resonant frequency within the allowable error band.

[0095] In one embodiment, a multi-strategy excitation intervention is simulated on the multidimensional fatigue damage propagation trajectory vector field, followed by excitation control decision optimization to construct a remote adjustment optimization strategy for the excitation frequency, including: Based on the flow meter parameter adjustment space, identify the excitation parameter resources that can be remotely called by the gas flow meter; based on the remotely called excitation parameter resources, perform multi-strategy anti-damage intervention simulation on the multidimensional fatigue damage propagation trajectory vector field, and generate damage control simulation data of multiple excitation adjustment schemes; The recovery speed and residual damage propagation rate of the sensor pipeline resonance state after excitation intervention are calculated based on the damage control simulation data. The secondary fracture risk probability is predicted based on the recovery speed and residual damage propagation rate to generate a secondary risk probability. An adaptive anomaly warning is generated based on the secondary risk probability to generate an adaptive damage anomaly warning signal. The secondary residual damage density is calculated from the damage control simulation data to generate the secondary residual damage density; the damage diffusion trend is analyzed from the secondary residual damage density to obtain the secondary damage diffusion trend map. Based on the secondary damage diffusion trend diagram, the excitation control decision is optimized, and a remote adjustment optimization strategy for the excitation frequency is constructed. Based on the remote adjustment optimization strategy for the excitation frequency and the adaptive damage anomaly early warning signal, the fatigue damage anomaly early warning and remote adjustment control operation of the gas flow meter structure is carried out.

[0096] The flowmeter parameter adjustment space can be a multi-dimensional feasible domain composed of all remotely controllable excitation-related parameters in the gas flowmeter. This domain can be used to define the boundaries of practically executable intervention operations, ensuring the physical feasibility of the deduced strategy. In this embodiment, the flowmeter parameter adjustment space can extract constraints such as frequency range, amplitude upper limit, and phase adjustment step size through device communication protocol parsing and hardware drive capability mapping. The remotely invoked excitation parameter resources can be a set of specific excitation control capabilities currently activated via remote commands within the flowmeter parameter adjustment space. These can be used as the input basis for multi-strategy anti-damage intervention deduction, connecting the digital model and the physical actuator. Furthermore, the remotely invoked excitation parameter resources are identified by the flowmeter parameter adjustment space and used to constrain the range of scheme generation for multi-strategy anti-damage intervention deduction. Multi-strategy anti-damage intervention deduction can be a process of simulating various suppressive excitation schemes for a multi-dimensional fatigue damage propagation trajectory vector field under the constraints of remotely invoked excitation parameter resources. This can be used to generate candidate intervention strategies with engineering feasibility, avoiding theoretically optimal but unexecutable schemes. In one specific embodiment, multi-strategy anti-damage intervention simulation can inject different combinations of excitation parameters into a digital twin model to observe the suppression response to the damage vector field.

[0097] Damage control simulation data from multiple excitation adjustment schemes can be a set of simulation results on damage state changes output after each excitation adjustment scheme runs in a digital model. This data provides a unified benchmark for subsequent quantitative assessment, supporting the calculation of stabilization rate and residual damage. Furthermore, the damage control simulation data from multiple excitation adjustment schemes can include, but is not limited to, time-domain resonance recovery curves, crack length evolution sequences, and stress intensity factor variation trajectories. The stabilization rate of the sensor pipeline resonance state after excitation intervention can be an indicator of the rate at which the resonance frequency or Q factor returns to a healthy baseline value after the application of excitation intervention. This can be used to measure the effectiveness of the intervention in restoring resonance stability and serves as a key evaluation dimension for control decisions. In this embodiment, the stabilization rate of the sensor pipeline resonance state after excitation intervention is calculated by fitting the time series of the resonance parameters after intervention, determining the time required for convergence to the tolerance band or the exponential decay rate. The residual damage propagation rate can be a quantitative value of the crack propagation rate or damage growth trend that still exists after excitation intervention. This can be used to reflect the residual damage activity that the intervention failed to completely suppress and to assess long-term risk. Furthermore, the residual damage propagation rate can be extracted from simulation data based on the Paris formula or cohesive model, representing the crack increment per unit time after intervention.

[0098] Secondary fracture risk probability prediction can be based on the residual amount of stabilization rate and damage propagation rate, combined with failure criteria, to probabilistically model the likelihood of recurrence of structural fracture. This can be used to transform deterministic indicators into risk probabilities, supporting dynamic adjustment of adaptive warning thresholds. In this embodiment, secondary fracture risk probability prediction can employ Monte Carlo simulation or Bayesian networks to fuse uncertainty parameters and output the probability of fracture events. The secondary risk probability can be a specific numerical result output by the secondary fracture risk probability prediction, which can be used as a direct trigger for adaptive anomaly warnings, replacing fixed thresholds.

[0099] Adaptive damage anomaly early warning signals can be early warning instructions dynamically generated based on secondary risk probabilities. Their trigger thresholds are adjusted in real-time according to the intervention effect, which can improve the early warning system's adaptability to post-intervention states and avoid false alarms or missed alarms caused by fixed thresholds. Secondary residual damage density can be the spatial distribution density of micro-damage still existing in local areas of the sensor pipeline after vibration intervention. It can be used to quantify the accumulation of latent damage that the intervention failed to remove, revealing potential weak areas. The secondary damage diffusion trend map can be a spatial visualization model of the possible subsequent expansion direction and intensity of damage obtained based on secondary residual damage density analysis. It can be used to reveal whether new risk paths have formed after intervention, providing closed-loop feedback for vibration control decision optimization. The remote adjustment optimization strategy for excitation frequency can be the final set of excitation frequency control instructions re-optimized after considering the secondary damage diffusion trend and secondary risk probability. It can be used to ensure that the adjustment strategy not only effectively suppresses current damage but also avoids secondary fracture risks and latent diffusion paths.

[0100] Identifying remotely accessible excitation parameter resources for gas flow meters based on their parameter adjustment space involves analyzing the device's communication interface and drive capabilities to extract the actual remotely controllable excitation parameter range and type. Furthermore, this operation can be achieved by automatically resolving the parameter space through the device description file or by probing the available adjustment range through online detection commands, ensuring the physical feasibility of subsequent deduction strategies and preventing theoretical solutions from deviating from hardware limitations. Deducing multi-strategy damage intervention for the multidimensional fatigue damage propagation trajectory vector field based on remotely accessible excitation parameter resources involves injecting various excitation combinations into the digital model under parameter resource constraints to simulate their suppression effect on the damage vector field. Further, this operation can be implemented using Latin hypercube sampling to cover the parameter space or by exploring efficient intervention strategies based on reinforcement learning agents, thereby generating engineering-feasible candidate intervention schemes and improving strategy practicality. Calculating the stabilization speed and residual damage propagation rate of the sensor pipeline resonance state after excitation intervention in damage control simulation data can be achieved by fitting the resonance parameter convergence curve and crack length growth curve to the simulation data and extracting the corresponding indicators. Furthermore, this operation can be achieved by fitting the stabilization process using an exponential decay model or by calculating the instantaneous expansion rate using the difference method, thereby establishing a comparable evaluation system for intervention effects.

[0101] Predicting the probability of secondary fracture risk based on the residual damage propagation rate and recovery rate can be achieved by using these indicators as inputs and combining them with the material's fracture toughness and load spectrum to calculate the probability of re-fracture. Furthermore, this operation can be implemented by constructing a surrogate model to accelerate Monte Carlo simulations or by using Bayesian updates to fuse prior failure data, thereby transforming deterministic results into risk language and supporting dynamic early warning. Generating the secondary risk probability can be achieved by outputting the numerical results of the risk prediction model. Further, this operation can be implemented by outputting the results as a percentage or as a logarithmic risk score, thus providing a quantitative basis for adaptive early warning. Adaptive anomaly early warning based on the secondary risk probability can be achieved by dynamically determining whether to trigger an early warning and the warning level according to a preset risk level mapping rule. Further, this operation can be implemented by using a fuzzy rule engine to determine the warning level or by using a dynamic threshold sliding window mechanism, thereby enabling the warning threshold to adaptively adjust according to the intervention effect and improving system robustness.

[0102] Calculating secondary residual damage density from damage control simulation data can be performed by calculating the damage density in local areas based on the damage field distribution after intervention. Further, this operation can be achieved by using kernel density estimation to calculate spatial damage density or by using statistical equivalent damage indices based on grid cells, thereby identifying hidden high-risk areas not covered by the intervention. Generating secondary residual damage density can be achieved by outputting the density calculation results as a scalar field or a heat map. Further, this operation can be achieved by generating a GeoTIFF format density map or outputting structured grid data, thus providing input for diffusion trend analysis. Damage diffusion trend analysis of secondary residual damage density can be performed by combining material properties and stress fields to predict possible future expansion paths of high-density areas. Further, this operation can be achieved by using streamline tracing algorithms to simulate diffusion direction or applying graph neural networks to propagate density risk, thereby revealing potential new risk channels after intervention. Obtaining a secondary damage diffusion trend map can be achieved by visualizing or structurally storing the trend analysis results. Further, this operation can be achieved by generating vector arrow diagrams to represent diffusion direction or constructing spatiotemporal cubes to express trend evolution, thereby providing spatial guidance for control decision optimization.

[0103] Optimizing excitation control decisions based on secondary damage diffusion trend maps can add constraints to avoid secondary diffusion paths to the existing multi-objective optimization. Furthermore, this operation can be achieved by adding a diffusion suppression term to the NSGA-II objective function or using a constraint satisfaction problem solver to handle new constraints, thus ensuring that the final strategy balances current suppression and long-term safety. Constructing a remote excitation frequency adjustment optimization strategy can encapsulate the optimization results into an executable frequency adjustment instruction set. Further, this operation can be achieved by generating a frequency sequence with time scheduling or outputting PID parameter tuning suggestions, thus forming the final closed-loop control output. Performing fatigue damage anomaly early warning and remote adjustment control operations for gas flow meter structures based on the remote excitation frequency adjustment optimization strategy and adaptive damage anomaly early warning signals can simultaneously execute early warning information release and remote excitation parameter distribution, achieving a perception-decision-execution closed loop. Furthermore, this operation can be achieved by parallel processing of early warning and adjustment instructions using an edge controller or by utilizing a 5G slicing network to ensure low-latency instruction distribution, thus completing the full transformation from passive monitoring to active control.

[0104] Taking remote operation and maintenance of a high-pressure gas gate station as an example, the gas flow meter abnormality early warning and remote adjustment control method in this embodiment can be as follows: When a Coriolis flow meter of a high-pressure gas gate station detects signs of micro-cracks at the root of the U-shaped pipe during operation, the system first identifies that the device supports 100-200Hz frequency adjustment and ±0.5V amplitude control, constituting excitation parameter resources that can be remotely invoked; then, three frequency sweep strategies are deduced in the digital twin. The simulation shows that a small excitation near 150Hz can make the resonant frequency stabilize within 6 hours, but the residual damage density in the weld heat-affected zone is still high; based on this, the probability of secondary fracture is calculated to be 18%, triggering a medium-level adaptive early warning; further analysis of the secondary damage diffusion trend map reveals that the damage may extend along the circumferential direction to the flange connection; the system optimizes the strategy accordingly, superimposing phase modulation on the 150Hz to enhance local energy dissipation, and finally issues a new strategy and simultaneously pushes the early warning to the operation and maintenance platform, realizing proactive intervention under controllable risk.

[0105] In one embodiment, a remote adjustment optimization strategy for the excitation frequency is constructed, including: A multi-objective weighted algorithm is used to solve the objective function in the risk-adverse decision optimization strategy. The objective function includes the target of resonant amplitude stability after excitation frequency adjustment, the target of minimizing measurement error, and the target of suppressing pipeline stress fatigue. The multi-objective weighted algorithm can be a numerical method for solving optimization problems involving multiple conflicting objectives. It can be used to find the Pareto optimal solution among resonance stability, metrological accuracy, and structural safety, avoiding intervention imbalances caused by a single dominant objective. In this embodiment, the multi-objective weighted algorithm can normalize each objective and linearly combine them according to preset or adaptive weights, transforming it into a single-objective optimization problem for solution. The objective function in the risk-resistant decision optimization strategy can be a mathematical expression describing the multi-dimensional performance requirements that the excitation parameter optimization must simultaneously satisfy. It can provide an optimization basis for the multi-objective weighted algorithm, ensuring that the intervention strategy balances accuracy, stability, and lifespan. Furthermore, the objective function in the risk-resistant decision optimization strategy can include, but is not limited to, the resonance amplitude stability objective, the metrological error minimization objective, and the pipeline stress fatigue suppression objective. The resonance amplitude stability objective after excitation frequency adjustment can be an optimization sub-objective requiring the sensor resonance response amplitude fluctuation to be controlled within an allowable range after excitation intervention. It can be used to ensure signal-to-noise ratio and measurement repeatability, preventing additional noise introduced by amplitude oscillations. Minimizing measurement error can be an optimization sub-objective aimed at reducing the deviation between the flowmeter output value and the actual mass flow rate. This can be used to directly improve the accuracy of energy measurement and meet the needs of trade settlement or pipeline balancing. Suppressing pipeline stress fatigue can be an optimization sub-objective aimed at reducing the amplitude of alternating stress in key parts of the sensor pipeline through vibration parameter design. This can be used to slow down the microcrack propagation rate and extend the service life of the equipment.

[0106] The optimal excitation drive frequency parameters and the optimal drive voltage gain parameters are generated based on the solution results. The optimal excitation drive frequency parameter can be the best excitation frequency value determined after multi-objective optimization. It can be used as one of the core commands for remote adjustment to guide the system back to the optimal operating point. Furthermore, the optimal excitation drive frequency parameter can be generated by a multi-objective weighted algorithm and sent to the excitation control module via a remote communication protocol. For example, the optimal excitation drive frequency parameter may include, but is not limited to, the fundamental frequency offset compensation value, the harmonic auxiliary excitation frequency, and the sweep frequency center frequency. The optimal drive voltage gain parameter can be the optimized excitation drive signal amplitude amplification coefficient, which can be used to control the excitation energy intensity, ensuring resonance stability while avoiding overdrive that could cause new damage. Generating the optimal excitation drive frequency parameter and the optimal drive voltage gain parameter based on the solution results can be achieved by extracting the frequency and gain—two key control variables—from the optimization solution. Furthermore, this operation can be implemented by selecting a compromise solution from the Pareto front to extract parameters, projecting them onto the feasible region according to engineering constraints, and then rounding them, thereby forming a directly executable physical control command.

[0107] The optimal excitation drive frequency parameters and optimal drive voltage gain parameters are sent to the excitation control module of the gas flow meter via a remote communication protocol. The remote communication protocol can be a standardized communication rule used to transmit optimized parameters from the cloud or edge computing nodes to the field flow meter, enabling secure, reliable, and low-latency issuance of control commands. The excitation control module of the gas flow meter is a hardware control unit within the flow meter responsible for executing excitation commands and driving the sensor pipeline vibration. It receives remote parameters and implements physical excitation, serving as the execution end of closed-loop control. In one specific embodiment, the excitation control module of the gas flow meter can integrate a DDS signal generator, a power amplifier, and a feedback sampling circuit, supporting external parameter configuration. Sending the optimal excitation drive frequency parameters and optimal drive voltage gain parameters to the excitation control module of the gas flow meter via the remote communication protocol can be achieved by encapsulating the parameters into a standard message format and transmitting them to the field device over a network. Furthermore, this operation can be implemented by publishing JSON format parameter packages via MQTT topics and updating device resources using the CoAPPUT method, thereby enabling remote, contactless control suitable for unattended scenarios.

[0108] The excitation control module receives parameters and performs closed-loop feedback control based on the phase-locked loop circuit, adjusting the excitation frequency of the sensor circuit to the preset optimal resonant operating point in real time. The phase-locked loop (PLL) circuit can be a feedback control circuit used to track and lock the current resonant frequency of the sensor circuit in real time. It can ensure that the excitation frequency is always synchronized with the actual resonant point, improving the closed-loop control accuracy. In one embodiment, the PLL circuit can compare the phase difference between the excitation and response using a phase detector, and adjust the voltage-controlled oscillator output via a loop filter. The preset optimal resonant operating point can be an ideal combination of excitation frequency and amplitude determined through optimized calculations. This allows the system to achieve an optimal balance between accuracy, stability, and lifespan, and can be used as the target state for closed-loop control, guiding the dynamic adjustment of the PLL. The excitation control module receives parameters and performs closed-loop feedback control based on the PLL circuit. This can be achieved by the PLL continuously monitoring the response phase and dynamically adjusting the output frequency after the module loads new parameters. Furthermore, this operation can be implemented by using a digital PLL in an FPGA to achieve phase tracking, or by using an analog PLL in conjunction with an ADC / DAC to form a hybrid closed loop. This ensures that the excitation always tracks the actual resonant point, improving control robustness. Real-time adjustment of the sensor circuit's excitation frequency to the preset optimal resonant operating point can be achieved by the PLL driving the excitation source to gradually approach and stabilize at the optimized operating frequency. Furthermore, this operation can be achieved by using a gradual frequency ramp to avoid shocks and implementing a step jump followed by rapid convergence, thereby restoring resonant stability and offsetting frequency drift caused by damage.

[0109] Record the resonant frequency drift and measurement deviation data before and after adjustment, and perform online iterative updates of model parameters to generate a dynamically corrected risk-resistant decision optimization strategy.

[0110] The resonant frequency drift before and after adjustment can be the difference in the actual resonant frequency of the sensor before and after the excitation parameter adjustment. This can be used to quantify the correction effect of the intervention on the frequency drift and for model calibration. Measurement deviation data can be the error record between the flowmeter output value and the reference standard before and after adjustment. This can be used to reflect the actual improvement in measurement accuracy brought about by the intervention. Online iterative updates of model parameters can be the process of dynamically correcting the model parameters in the risk-resistant decision optimization strategy using measured adjustment effect data. This can be used to enable the optimization strategy to have self-learning capabilities and adapt to equipment aging and changes in operating conditions. Furthermore, online iterative updates of model parameters can use recursive least squares, Kalman filtering, or online gradient descent to update the objective function weights or constraint coefficients. The dynamically corrected risk-resistant decision optimization strategy can be a version of the optimization strategy with stronger environmental adaptability after online iterative updates. This can be used to improve the accuracy and reliability of subsequent interventions and form a continuous evolution capability. Furthermore, the dynamically corrected risk-resistant decision optimization strategy can be generated by the online iterative updates of model parameters and used for the next round of multi-objective optimization solutions.

[0111] Taking the intelligent operation and maintenance of urban gas high-pressure regulating stations as an example, the gas flow meter abnormality early warning and remote adjustment control method in this embodiment can be as follows: Due to long-term vibration, the resonant frequency of an ultrasonic gas flow meter in a regulating station slowly drifts. The system constructs a three-objective function including amplitude stability, metering error, and stress suppression. The optimal excitation frequency is obtained by using a multi-objective weighted algorithm, which is 182.3Hz and the voltage gain is 1.05 times. This parameter is sent to the field excitation control module via the MQTT protocol. The module's built-in digital phase-locked loop immediately starts closed-loop tracking and locks the excitation frequency to the new operating point within 10 seconds. After adjustment, the recorded frequency drift decreases from +0.8Hz to +0.05Hz, and the metering deviation decreases from 1.2% to 0.15%. The system uses this data to update the weight of the metering error term in the objective function online and generates a dynamically corrected risk-resistant decision optimization strategy for the next intervention, thereby continuously improving the control accuracy.

[0112] In addition, refer to Figure 2 To achieve the above objectives, the present invention also provides a gas flow meter abnormality early warning and remote adjustment control system, the system comprising: The map construction module 10 is used to acquire multi-dimensional time-series data of the gas flow meter resonant sensing unit, perform resonant characteristic information analysis and structural state logic hierarchy reconstruction, and construct a structural health status perception map of the flow meter sensor pipeline. The spectrum mining module 20 is used to calculate the local offset amplitude of multi-resonance features in the structural health status perception map and to mine structural damage disturbance features in order to construct a fatigue damage disturbance spectrum map. The type generation module 30 is used to perform deep feature analysis and node damage mutation feature analysis on the fatigue damage disturbance spectrum map to generate the damage evolution dynamic type of the mutation node. The port identification module 40 is used to perform full-cycle time-series drift tracing of the fatigue damage disturbance spectrum and predict potential damage nodes based on the damage evolution dynamic type, and identify other potential damage propagation ports. The vector modeling module 50 is used to simulate damage evolution at multiple time points and damage disturbance diffusion coupling for other potential damage propagation ports, and to construct a multidimensional fatigue damage propagation trajectory vector field. The strategy optimization module 60 is used to perform multi-strategy excitation intervention simulation on the multidimensional fatigue damage propagation trajectory vector field, and then perform excitation control decision optimization to construct an optimized strategy for remote adjustment of excitation frequency.

[0113] Other embodiments or specific implementations of the gas flow meter abnormality early warning and remote adjustment control system of the present invention can be referred to the above-mentioned method embodiments, and will not be repeated here.

[0114] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for abnormal early warning and remote adjustment control of a gas flow meter, characterized in that, The method includes: Acquire multi-dimensional time-series data of the resonant sensing unit of the gas flow meter, perform resonant characteristic information analysis and structural state logic hierarchy reconstruction, and construct a structural health status perception map of the flow meter sensor pipeline. The local offset amplitude of multi-resonance features is calculated from the structural health status perception map, and structural damage disturbance features are mined to construct a fatigue damage disturbance spectrum map. Deep feature analysis and node damage mutation feature analysis are performed on the fatigue damage disturbance spectrum to generate the damage evolution dynamic type of mutation nodes. The fatigue damage disturbance spectrum is traced back to its full-cycle temporal drift, and potential damage nodes are predicted based on the damage evolution dynamic type to identify other potential damage propagation ports. Multi-time point damage evolution simulation and damage perturbation diffusion coupling are performed on other potential damage propagation ports to construct a multidimensional fatigue damage propagation trajectory vector field; A multi-strategy excitation intervention was simulated for the multidimensional fatigue damage propagation trajectory vector field, and then the excitation control decision was optimized to construct a remote adjustment optimization strategy for the excitation frequency.

2. The gas flow meter abnormality early warning and remote adjustment control method as described in claim 1, characterized in that, The process of acquiring multi-dimensional time-series data of the gas flow meter resonant sensing unit, performing resonant characteristic information analysis and structural state logic hierarchy reconstruction, and constructing a structural health status perception map of the flow meter sensor pipeline includes: Multi-dimensional resonant timing data stream is obtained based on the excitation drive unit and sensing detection unit of the gas flow meter; Heterogeneous feature information is parsed from the multi-dimensional resonant timing data stream to generate heterogeneous resonant feature information, which includes the resonant frequency of the pipeline, the driving phase difference, the feedback voltage of the excitation coil, and the resonant quality factor Q value. The heterogeneous resonance characteristic information is time-stamp aligned and numerically standardized to construct multidimensional structural state standardized features. Multi-feature correlation analysis was performed on the standardized features of multidimensional structural states to extract the causal relationships between different features. Structural dimension mining is performed on the standardized features of the multidimensional structural state, and logical hierarchy is reconstructed based on the causal relationship to construct a structural health status perception map of the flow meter sensor pipeline.

3. The gas flow meter abnormality early warning and remote adjustment control method as described in claim 1, characterized in that, The calculation of local offset amplitude of multi-resonance features in the structural health status perception spectrum and the mining of structural damage disturbance features to construct a fatigue damage disturbance spectrum include: The sensor pipeline structure health status perception map is decomposed into time-series states to generate multiple continuous resonance detection windows; The local offset amplitude of the multi-resonance characteristics is calculated for multiple continuous resonance detection windows to obtain the resonance offset amplitude value of each window; Based on the resonant offset amplitude value, transient large offset distortion is identified, and structural state distortion points are marked. Based on a preset fatigue damage sensitivity threshold, damage risk sensitivity assessment is performed on structural state distortion points to extract potential damage perception factors. Dynamic damage perturbation features are mined from potential damage sensing factors to construct a fatigue damage perturbation spectrum.

4. The gas flow meter abnormality early warning and remote adjustment control method as described in claim 3, characterized in that, The step of dynamically mining the damage perturbation features of potential damage sensing factors to construct a fatigue damage perturbation spectrum includes: Periodic damage perturbation characteristics are analyzed for potential damage sensing factors, and periodic resonance offset characteristics are extracted. Based on the periodic resonance offset characteristics, multi-time-point offset fitting is performed to construct the damage disturbance waveform curve; Calculate the drift frequency intensity and trend slope of the damage disturbance waveform curve; Based on the drift frequency intensity and trend slope, drift frequency continuity is mined to construct a fatigue damage disturbance spectrum.

5. The gas flow meter abnormality early warning and remote adjustment control method as described in claim 1, characterized in that, The process of performing deep feature analysis and node damage mutation feature analysis on the fatigue damage disturbance spectrum to generate the damage evolution dynamic type of mutation nodes includes: Deep feature analysis is performed on the fatigue damage disturbance spectrum, and damage feature annotation and encoding are performed to obtain the damage disturbance feature code; Based on the damage disturbance feature encoding, the sensor pipeline structure health status perception map is correlated and mapped and the topology is located to obtain the map location information of the damage factor; Based on the location information of the map, damage propagation path mining is performed to extract the damage propagation path; A path abrupt change key node analysis was performed on the damage propagation path to extract microcrack initiation nodes, fatigue accumulation nodes, and low-frequency activation nodes of stress concentration in the path. The node state abrupt change characteristics of the microcrack initiation node, fatigue accumulation node, and low-frequency activation node of stress concentration are analyzed to generate the damage evolution dynamic type of the abrupt node.

6. The gas flow meter abnormality early warning and remote adjustment control method as described in claim 1, characterized in that, The process of tracing the full-cycle temporal drift of the fatigue damage disturbance spectrum and predicting potential damage nodes based on the damage evolution dynamics type, as well as identifying other potential damage propagation ports, includes: Multi-feature correlation mining is performed on the fatigue damage disturbance spectrum to identify associated damage feature groups; Deeply deconstruct the damage propagation logic of the associated damage feature group to generate damage extension propagation logic; Based on the damage propagation logic, the damage propagation path is traced throughout the entire time cycle to obtain the full-cycle damage propagation chain. Based on the aforementioned dynamic damage evolution type and full-cycle damage propagation chain, potential damage nodes are predicted, and other potential damage propagation ports are identified.

7. The gas flow meter abnormality early warning and remote adjustment control method as described in claim 1, characterized in that, The process of performing multi-time-point damage evolution simulation and damage perturbation diffusion coupling on other potential damage propagation ports to construct a multidimensional fatigue damage propagation trajectory vector field includes: Identify the current structural health status of sensor pipelines based on multi-dimensional resonant timing data streams; Real-time dynamic stress distribution analysis is performed on the current structural health status of the sensor pipeline to extract key stress characteristics; Based on the key stress characteristics, multi-time-point damage evolution simulations were performed on other potential damage propagation ports to generate damage evolution simulation data; Damage trajectory evolution analysis is performed on damage evolution simulation data to generate damage evolution trajectories; Calculate the direction of expansion, rate of expansion, strain potential energy, and probability of fracture failure of the damage evolution trajectory; Based on the aforementioned expansion direction, expansion rate, strain potential energy, and fracture failure probability, damage perturbation diffusion coupling is performed to construct a multidimensional fatigue damage expansion trajectory vector field.

8. The gas flow meter abnormality early warning and remote adjustment control method as described in claim 1, characterized in that, The process involves multi-strategy excitation intervention deduction of the multidimensional fatigue damage propagation trajectory vector field, followed by excitation control decision optimization, and the construction of a remote adjustment optimization strategy for the excitation frequency, including: Based on the flow meter parameter adjustment space, identify the excitation parameter resources that can be remotely called by the gas flow meter; according to the excitation parameter resources that can be remotely called, perform multi-strategy anti-damage intervention simulation on the multidimensional fatigue damage propagation trajectory vector field, and generate damage control simulation data of multiple excitation adjustment schemes; Calculate the recovery speed and residual damage propagation rate of the sensor pipeline resonance state after the excitation intervention of the damage control simulation data; predict the secondary fracture risk probability based on the recovery speed and residual damage propagation rate to generate a secondary risk probability; and generate an adaptive anomaly warning signal based on the secondary risk probability. The secondary residual damage density is calculated from the damage control simulation data to generate a secondary residual damage density; the damage diffusion trend is analyzed from the secondary residual damage density to obtain a secondary damage diffusion trend map. Based on the secondary damage diffusion trend diagram, the excitation control decision is optimized, and a remote adjustment optimization strategy for the excitation frequency is constructed. Based on the remote adjustment optimization strategy for the excitation frequency and the adaptive damage anomaly early warning signal, the fatigue damage anomaly early warning and remote adjustment control operation of the gas flow meter structure is carried out.

9. The gas flow meter abnormality early warning and remote adjustment control method as described in claim 8, characterized in that, The constructed excitation frequency remote adjustment optimization strategy includes: A multi-objective weighted algorithm is used to solve the objective function in the risk-adverse decision optimization strategy. The objective function includes the target of resonance amplitude stability after excitation frequency adjustment, the target of minimizing measurement error, and the target of suppressing pipeline stress fatigue. The optimal excitation drive frequency parameters and the optimal drive voltage gain parameters are generated based on the solution results. The optimal excitation drive frequency parameters and optimal drive voltage gain parameters are sent to the excitation control module of the gas flow meter via a remote communication protocol. The excitation control module receives parameters and performs closed-loop feedback control based on the phase-locked loop circuit, adjusting the excitation frequency of the sensor circuit to the preset optimal resonant operating point in real time. Record the resonant frequency drift and measurement deviation data before and after adjustment, and perform online iterative updates of model parameters to generate a dynamically corrected risk-resistant decision optimization strategy.

10. A gas flow meter abnormality early warning and remote adjustment control system, characterized in that, The system includes: The graph construction module is used to acquire multi-dimensional time-series data of the resonant sensing unit of the gas flow meter, perform resonant characteristic information analysis and structural state logic hierarchy reconstruction, and construct a structural health status perception graph of the flow meter sensor pipeline. The spectrum mining module is used to calculate the local offset amplitude of multi-resonance features in the structural health status perception map and to mine structural damage disturbance features in order to construct a fatigue damage disturbance spectrum map. The type generation module is used to perform deep feature analysis and node damage mutation feature analysis on the fatigue damage disturbance spectrum map, and generate the damage evolution dynamic type of the mutation node. The port identification module is used to trace the full-cycle temporal drift of the fatigue damage disturbance spectrum and predict potential damage nodes based on the damage evolution dynamic type, and identify other potential damage propagation ports. The vector modeling module is used to simulate damage evolution at multiple time points and couple damage disturbance diffusion for other potential damage propagation ports, and to construct a multidimensional fatigue damage propagation trajectory vector field. The strategy optimization module is used to perform multi-strategy excitation intervention simulation on the multidimensional fatigue damage propagation trajectory vector field, and then optimize the excitation control decision to construct an optimized strategy for remote adjustment of excitation frequency.

Images