AIoT cloud side collaborative Internet of Things management platform and method

Abstract

The invention relates to an AIoT cloud edge collaborative Internet of Things management platform and method, and the platform comprises a multi-mode special sensor network which is disposed at a first abutting surface and a second abutting surface formed between the end part of a pipeline and a connecting piece. The signal acquisition module is used for respectively acquiring a structural damage signal caused by embrittlement caused by active gas and a leakage signal caused by exudation of the active gas; the edge processing layer is used for performing fusion analysis on multi-source data such as acoustic emission, guided waves, vibration and gas concentration; and the cloud platform is used for data interaction, threshold optimization and operation management. The method comprises the steps of structural strength evaluation, trace leakage judgment and grading treatment, and early recognition of trace leakage is achieved through conjoint analysis of structural damage of the abutting surface and gas concentration of a monitoring space. By implementing the scheme, the whole-process sensing of the active gas pipeline from structural damage to leakage can be realized, millisecond-level safety linkage is supported, and the intrinsic safety of underground pipeline operation is improved.

Description

Technical Field

[0001] This application relates to the field of cloud-edge collaborative monitoring, and in particular to an AIoT cloud-edge collaborative Internet of Things management platform and method. Background Technology

[0002] Underground metal pipelines transporting reactive gases are prone to material degradation, interface damage, and micro-leakage under long-term high-pressure conditions. Reactive gases, with their small molecular size, strong diffusivity, and ability to easily penetrate metal lattices, preferentially penetrate and accumulate at structural discontinuities at the pipe ends (such as the contact surface between the pipe end and the connector), thereby inducing local embrittlement and microcrack formation. This type of embrittlement effect caused by reactive gases often exhibits the characteristics of "no macroscopic abnormal parameters" and "progressive damage accumulation."

[0003] When microcracks continue to propagate to the outer wall of the pipe or the interface connecting it to the outside world, they will form a hidden micro-leakage path. The initial leakage of reactive gas is extremely low and cannot cause pressure changes. Therefore, traditional monitoring methods that rely on macroscopic parameters such as pressure and flow rate are difficult to identify early leakage risks in a timely manner.

[0004] Furthermore, reactive gases can easily accumulate in confined spaces, forming flammable or explosive mixtures, requiring monitoring systems capable of ppm-level concentration acquisition and millisecond-level response control. Current technologies still lack dedicated monitoring systems designed to address the risk mechanism of "embrittlement damage + trace leakage." Summary of the Invention

[0005] The purpose of this application is to solve the technical problem that traditional monitoring methods cannot effectively address the embrittlement damage caused by reactive gases and trace leaks.

[0006] According to one aspect of this application, an AIoT cloud-edge collaborative Internet of Things management platform is provided, applied to an underground active gas pipeline, the pipeline comprising: pipeline; A corrosion-resistant inner liner coated on the inner wall of the pipe; And the connectors that connect to the end of the pipe; The main body surface of the connector includes a first surface and a second surface; the end face of the pipe and the end face of the anti-corrosion inner liner abut against the first surface to form a first abutting surface; the outer wall of the pipe abuts against the second surface to form a second abutting surface. Under active gas delivery conditions, active gas molecules penetrate into the first contact surface and accumulate there, inducing microcracks formed by the active gas embrittlement in the pipe body and connector body at the first contact surface; the microcracks extend along the first contact surface and extend to the second contact surface, forming a seepage channel for the active gas. The platform includes: A dedicated sensor network is provided, which has multiple sensor nodes for collecting structural damage signals at the first contact surface and gas leakage signals at the second contact surface. An edge processing layer is used to receive and process the structural damage signal and the gas leak signal; The cloud platform communicates with the edge processing layer for data interaction and storage management.

[0007] Preferably, the sensing nodes deployed on the first contact surface include: Acoustic emission sensor array, used to acquire stress wave signals generated by the propagation of brittle microcracks; An ultrasonic thin guided wave detector is used to acquire the guided wave time domain signal of the pipe segment adjacent to the first contact surface; The sensing nodes deployed on the second contact surface include: Fiber optic acoustic sensors are used to collect pipeline vibration parameters caused by reactive gas leaks. A laser gas sensor is used to collect the concentration of active gases in the monitoring space corresponding to the second contact surface.

[0008] This invention also provides an AIoT cloud-edge collaborative IoT management method, applied to the aforementioned AIoT cloud-edge collaborative IoT management platform, the method comprising: S10: Receive and process stress wave signals, guided wave time domain signals, pipeline vibration parameters, and reactive gas concentration parameters; S20. Based on stress wave signals and guided wave time domain signals, determine whether the pipeline strength is within the preset safety threshold. If not, generate a structural risk warning and trigger a pipeline depressurization operation command. S30. If so, then determine whether a small leak has occurred in the high-risk area based on the pipeline vibration parameters and the active gas concentration parameters. S40. When a leak is detected, the electrically controlled shut-off valve is triggered to close and the ventilation system is started.

[0009] Preferably, the process further includes the following steps before step S10: S50. Construct a three-dimensional pipeline model of the underground active gas pipeline based on the design data and field measurement data of the underground active gas pipeline. S51. In the three-dimensional pipeline model, the physical locations corresponding to the joint ends between two adjacent pipelines and the connection ends between the valve body and the corresponding pipelines are marked as high-risk areas. S52. Generate the high-risk coordinates of each of the high-risk areas in the three-dimensional pipeline model; S53. Designate the enclosed space adjacent to each of the high-risk areas as the monitoring space; S54. Establish the mapping relationship between each of the high-risk areas and the corresponding monitoring space in the three-dimensional pipeline model.

[0010] Preferably, the step S30, "determining whether a trace amount of reactive gas leakage has occurred in the high-risk area based on the pipeline vibration parameters and the reactive gas concentration parameters," specifically includes: S31. Within a preset detection time window, the concentration parameters of the active gas in the monitoring space are sampled multiple times, and the concentration change trend is calculated. S32. When the concentration of reactive gas in a high-risk area continues to rise and exceeds the preset trace leakage threshold within the detection time window, and the corresponding pipeline vibration parameter continues to exceed the preset vibration threshold during the same period, it is determined that a trace reactive gas leak has occurred in the high-risk area.

[0011] Preferably, the step S20 of "determining whether the pipeline strength is within a preset safety threshold" includes: S21. Extract features from the stress wave signal to obtain acoustic emission event counts and accumulated energy, forming a crack activity index. S22. Analyze the time-domain signal of the guided wave to obtain the pipe wall thickness reduction rate data; S23. The crack activity index and the pipe wall thickness reduction rate are combined and calculated to obtain the structural state parameters; S24. When the structural state parameters exceed the safety threshold, the pipeline strength safety assessment is deemed abnormal.

[0012] Preferably, step S23, "integrating the crack activity index with the pipe wall thickness reduction rate," includes: The crack activity index and the pipe wall thickness reduction rate were normalized respectively. The weighting coefficients of the crack activity index and the pipe wall thickness reduction rate in the structural state parameters are determined by the entropy weight method. The normalized indices are weighted and summed according to the weighting coefficients to obtain the dimensionless structural state parameter values.

[0013] Preferably, after S40, the method further includes: S41. After the electrically controlled shut-off valve is triggered, the concentration of active gas in the leakage isolation zone is continuously monitored by the laser gas sensor; S42. If the active gas concentration parameter is lower than the first safety threshold, it is determined to be a controllable leak, a leak alarm is generated, the sampling frequency is increased to the first preset frequency, and the ventilation system is controlled to operate at the first preset power. S43. If the active gas concentration parameter reaches or exceeds the second safety threshold, it is determined to be an uncontrollable leak. The edge processing layer controls the ventilation system to operate at the second preset power within milliseconds, and simultaneously sends the highest level accident alarm to the cloud platform; wherein, the second safety threshold is higher than the first safety threshold, and the second preset power is greater than the first preset power.

[0014] Preferably, the method further includes: S60. The cloud platform receives the crack activity index, pipe wall thickness reduction rate, vibration parameters, active gas concentration parameters, and corresponding structural risk warning, micro-leakage judgment results and control command execution results uploaded by the edge processing layer, forming a historical operation dataset. S61. Based on the historical operation dataset, the critical strength threshold used for strength safety assessment, the micro-leakage threshold used for determining micro-leakage, the first safety threshold, and the second safety threshold are optimized using a machine learning model to generate updated threshold parameters. S62. Send the updated threshold parameters to the edge processing layer.

[0015] Preferably, the step S61, "optimizing the critical strength threshold for strength safety assessment and the micro-leakage threshold for determining micro-leakage using a machine learning model," specifically includes: Based on the historical operation dataset, the deformation characteristics of the pipeline body under different ambient temperatures and pipeline operating pressure conditions are extracted. A random forest regression model was established with ambient temperature, pipeline pressure, and historical crack propagation rate as inputs and the optimal safety threshold as output. The model hyperparameters are determined by cross-validation, and the root mean square error is used as the model optimization target to dynamically adjust the critical intensity threshold, trace leakage threshold, first safety threshold, and second safety threshold.

[0016] This application offers the following advantages: it enables "early warning" of the root causes of reactive gas leaks, transforming passive response into proactive prevention. By deploying an acoustic emission sensor array and an ultrasonic thin-guided wave detector at the first contact surface, it can directly capture microcrack propagation signals and pipe wall thinning signals caused by embrittlement damage. This allows the system to issue structural risk warnings and trigger early intervention measures such as pressure reduction during the pipeline's structural strength degradation stage, before substantial leaks occur. This fundamentally changes the reactive gas safety management model, enabling the prevention of major leak accidents at their source.

[0017] This invention enhances the accuracy of reactive gas risk perception and resolves the mismatch issue of general models. By constructing dedicated sensor networks for the first contact surface (structural damage zone) and the second contact surface (gas seepage zone), collaborative acquisition of multimodal signals is achieved. A "defense-in-depth" system for reactive gas leaks is established, realizing a closed-loop process from early warning to response. This application does not improve sensors in isolation but constructs a defense-in-depth system encompassing "structural health assessment → trace leak detection → tiered safety response." The fusion processing logic of the edge processing layer ensures that the system can execute a scientific decision-making process of "intensity assessment first, then leak detection." Once a leak is confirmed, the system can leverage the advantages of edge computing to directly control the electrically controlled shut-off valve and ventilation system within milliseconds, forming a rapid and automatic closed loop from early risk identification to final safety control. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram illustrating the structure of the AIoT cloud-edge collaborative Internet of Things management platform according to an embodiment of this application; Figure 2 This is a partial cross-sectional view of the pipe and connector at the end connection in one embodiment of this application; Figure 3 This is a logical block diagram of another embodiment of the AIoT cloud-edge collaborative IoT management method described in this application.

[0020] The following are the definitions of the reference numerals: 100, AIoT cloud-edge collaborative IoT management platform; 10, dedicated sensor network; 20, edge processing layer; 30, cloud platform; 40, connector; 41, first contact surface; 42, second contact surface; 43, first surface; 44, second surface; 50, pipe; 51, anti-corrosion inner liner; 52, outer wall; 53, pipe end face; 54, inner wall. Detailed Implementation

[0021] To facilitate understanding of this application, a more complete description will be provided below with reference to the accompanying drawings. Preferred embodiments of this application are shown in the drawings. However, this application can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of this application.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein in the specification of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0023] Please refer to Figure 1-2 One embodiment of this application provides an AIoT cloud-edge collaborative Internet of Things management platform 100, applied to an underground active gas pipeline, the pipeline including: Pipeline 50; Corrosion-resistant inner liner 51 coated on the inner wall 54 of pipe 50; And the connector 40 that connects to the end of the pipe 50; The main surface of the connector 40 includes a first surface 43 and a second surface 44; the end face of the pipe 53 and the end face of the anti-corrosion inner liner 51 abut against the first surface 43 to form a first abutting surface 41; the outer wall of the pipe 52 abuts against the second surface 44 to form a second abutting surface 42. Under active gas transport conditions, active gas molecules penetrate into the first contact surface 41 and accumulate there, inducing microcracks formed by the active gas embrittlement of the pipe 50 body and the connector 40 body at the first contact surface 41; the microcracks extend along the first contact surface 41 and extend to the second contact surface 42, forming an active gas seepage channel. The platform includes: A dedicated sensor network 10 is provided with multiple sensor nodes for collecting structural damage signals at the first contact surface 41 and gas leakage signals at the second contact surface 42. Edge processing layer 20 is used to receive and process structural damage signals and gas leak signals; The cloud platform 30 is connected to the edge processing layer 20 for data interaction and storage management.

[0024] In this embodiment, it should be noted that the AIoT cloud-edge collaborative IoT management platform 100 is used for targeted online monitoring of key structural components of underground active gas pipelines. The monitoring objects include not only the main body material of the pipeline 50, but also the anti-corrosion inner liner 51 coated on the inner wall 54 and the composite structural interface formed by the connector 40 connected to the end of the pipeline 50. The main surface of the connector 40 has a first surface 43 and a second surface 44. The first surface 43 abuts against the pipeline end face 53 and the end of the inner liner, forming a first abutment surface 41, which is the axial discontinuous interface of the pipeline. The second surface 44 contacts the outer wall 52 of the pipeline, forming a second abutment surface 42, which is the external interface where the radial continuity of the pipeline is interrupted.

[0025] Under long-term high-pressure transportation conditions of reactive gases (such as hydrogen), reactive gas molecules will preferentially penetrate along the material interface of the first contact surface 41. Since the end face simultaneously contains the metal body of the pipe 50, the metal of the connector 40, and the termination end of the anti-corrosion inner liner 51, a multi-material intersection area is formed, which is prone to microscopically loose bonding. This causes the reactive gas to accumulate at this point and induces microcracks in the metal matrix caused by embrittlement. As the microcracks initiate and expand, they tend to propagate outward along the first contact surface 41, eventually extending to the second contact surface 42 and penetrating at the second contact surface 42, forming a leakage channel for the reactive gas to seep out from the inside of the pipe 50 to the outside.

[0026] To identify the entire process from "brittle damage - crack propagation - micro-leakage," this embodiment deploys structural damage detection nodes (such as acoustic emission arrays and thin-guided wave detectors) on the first contact surface 41 using a dedicated sensor network 10. This collects stress wave signals or wall thickness attenuation signals generated by the dynamic propagation of brittle microcracks in the metal matrix. Leakage detection nodes (such as fiber optic acoustic sensors and laser gas sensors) are deployed on the second contact surface 42 to collect vibration and concentration parameters caused by the seepage of reactive gases. The edge processing layer 20 synchronously receives and fuses the aforementioned multimodal data, achieving an integrated assessment of local structural health and micro-leakage status. The cloud platform 30 is used for data storage, model analysis, and threshold optimization, thereby constructing a cloud-edge-device collaborative end-to-end monitoring system.

[0027] The technical solution implemented in this embodiment enables simultaneous monitoring of both the "structural damage source" and the "minor leakage phenomenon" in underground reactive gas pipelines. Compared to traditional methods that rely on macroscopic parameters such as pressure and flow rate to determine leakage, this embodiment can achieve early identification of structural damage at the micro-crack stage, thus significantly advancing the warning window. Simultaneously, by setting high-risk detection points at the first contact surface 41 and the second contact surface 42, the platform can completely capture the actual migration path of reactive gas along the interface, achieving full-process perception from the formation of embrittlement damage to leakage. Furthermore, leveraging the real-time computing capabilities of the edge processing layer 20, rapid response can be executed immediately upon the occurrence of anomalies, while the computing resources of the cloud platform 30 can achieve threshold self-learning optimization, continuously improving monitoring accuracy over time. Therefore, this embodiment not only improves the inherent safety of underground reactive gas pipelines but also effectively enhances the system's intelligence, adaptability, and real-time response capabilities.

[0028] Furthermore, the sensing nodes deployed on the first contact surface 41 include: Acoustic emission sensor array, used to acquire stress wave signals generated by the propagation of brittle microcracks; An ultrasonic thin-walled guided wave detector is used to acquire the guided wave time-domain signal of the pipe segment adjacent to the first contact surface 41. The sensing nodes deployed on the second contact surface 42 include: Fiber optic acoustic sensors are used to collect vibration parameters of pipeline 50 caused by reactive gas leakage. A laser gas sensor is used to collect the concentration of active gas in the monitoring space corresponding to the second contact surface 42.

[0029] In this embodiment, it should be noted that the sensing nodes deployed at the first contact surface 41 are mainly for monitoring structural damage caused by embrittlement due to reactive gases. Specifically, several sensing units of the acoustic emission sensing array are attached to the outer surface of the pipe wall 52 and the connector 40 near the first contact surface 41. When embrittled microcracks in the metal matrix near the first contact surface 41 initiate, propagate, or become transiently unstable, they release high-frequency stress wave signals. The acoustic emission sensing array picks up these stress waves in real time and outputs continuous acoustic emission event counts and energy characteristics, thereby reflecting the timing and intensity of crack activity in the region of the first contact surface 41. Meanwhile, the transmitting and receiving probes of the ultrasonic thin guided wave detector are arranged in a counter-or collinear manner on the pipe walls of the pipe 50 on both sides of the first contact surface 41. By exciting guided waves into the pipe wall and receiving their time-domain response, the changes in the guided wave propagation characteristics caused by brittle damage, wall thickness reduction or interface debonding are detected, thereby obtaining the thickness reduction rate and overall cross-sectional stiffness change of the pipe section adjacent to the first contact surface 41, and realizing a quantitative assessment of the structural integrity of the area.

[0030] The sensing nodes deployed on the second contact surface 42 monitor the actual leakage behavior of reactive gas. The sensing cable of the fiber optic acoustic sensor is laid axially along the pipe 50 and crosses the second contact surface 42. When reactive gas is ejected through the leak channel formed by microcracks, it generates flow noise and structural vibration response in a specific frequency band near the second contact surface 42. The fiber optic acoustic sensor can acquire changes in vibration parameters along the pipeline in a distributed manner and locate the source of leakage disturbance in time and space. The laser gas sensor is deployed in the monitoring space facing the second contact surface 42 to collect the concentration of reactive gas in this locally sealed or semi-sealed space with high sensitivity. When the leakage persists, the concentration in this area will show a cumulative upward trend over time, thus providing a reliable quantitative basis for the determination and graded treatment of trace leaks. These multiple sensors are connected to the edge processing layer 20 through a dedicated sensor network 10, forming a collaborative sensing system targeting the "embrittlement damage source" of the first contact surface 41 and the "leakage manifestation end" of the second contact surface 42.

[0031] The technical solution implemented in this embodiment enables integrated and refined monitoring of the "embrittlement microcrack evolution process" and "reactive gas leakage behavior" at the same connection structure. On the one hand, acoustic emission and thin guided wave signals reflect the structural failure state near the first contact surface 41 from the perspective of material internal structure and cross-sectional stiffness, allowing the system to identify high-risk areas before a significant leak occurs. On the other hand, fiber optic acoustics and laser gas detection characterize the actual leakage intensity and accumulation trend of the second contact surface 42 from the dimensions of vibration and concentration, ensuring that minute leaks are not missed. By integrating four types of sensor information on the same platform, the reliability of anomaly identification is improved, the probability of false alarms and missed alarms caused by a single sensing mechanism is reduced, and a rich data foundation is provided for subsequent leak location, graded treatment, and threshold adaptive optimization, thereby significantly improving the intrinsic safety level of underground reactive gas pipelines at critical interfaces.

[0032] like Figure 3 As shown, the present invention also provides an AIoT cloud-edge collaborative IoT management method, applied to the aforementioned AIoT cloud-edge collaborative IoT management platform, the method comprising: S10. Receive and process stress wave signals, guided wave time-domain signals, pipeline vibration parameters, and reactive gas concentration parameters. In this step, it should be noted that: the platform acquires the stress wave signal output from the acoustic emission sensor array deployed at the first contact surface, and the guided wave time-domain signal output from the ultrasonic thin guided wave detector, through a dedicated sensor network, to reflect the structural damage state in the area adjacent to the first contact surface; simultaneously, it acquires the distributed vibration parameters along the pipeline generated by the fiber optic acoustic sensor deployed at the second contact surface, and the reactive gas concentration parameters in the monitoring space collected by the laser gas sensor.

[0033] Various signals are synchronously accessed to the edge processing layer using timestamps. The preprocessing module performs filtering, normalization, feature extraction, and time window slicing, enabling subsequent analysis steps to make judgments based on high-quality data.

[0034] In particular, in a preferred embodiment, the active gas is hydrogen, and the laser gas sensor can detect hydrogen concentration at the ppm level, which is suitable for hydrogen with characteristics such as easy diffusion and extremely low lower explosive limit.

[0035] S20. Based on the stress wave signal and the guided wave time-domain signal, determine whether the pipeline strength is within the preset safety threshold. If not, generate a structural risk warning and trigger a pipeline depressurization operation command. In this step, it should be noted that the edge processing layer extracts the acoustic emission event count, accumulated energy, and spectral characteristics from the stress wave signal to construct the "crack activity index"; and performs modal analysis, wave velocity calculation, and signal attenuation comparison on the guided wave time-domain signal to obtain the pipe wall thickness reduction rate or interface damage index in the region adjacent to the first contact surface.

[0036] Subsequently, the edge processing layer fuses and calculates the above indicators to obtain "structural state parameters", and compares them with a preset safety threshold to determine whether there is a trend of strength degradation or embrittlement in the first contact surface area.

[0037] This step takes precedence over leak detection, ensuring that the safety of pipeline materials is assessed first, and avoiding missing the opportunity to intervene by relying solely on leak phenomena when cracks propagate too rapidly.

[0038] S30. If so, determine whether a minor leak has occurred in the high-risk area based on the pipeline vibration parameters and the active gas concentration parameters. It should be noted that in this step, the system only enters the leak detection process when S20 determines that the structural state parameters do not exceed the safety threshold, thus realizing a chain logic of "structure priority - leak post-progression".

[0039] In this step, the concentration parameters of the reactive gas are continuously sampled within a preset detection time window, and their rate of increase, fluctuation characteristics, and cumulative changes are analyzed using a trend extraction model. Simultaneously, pipeline vibration parameters are used to identify flow disturbances and characteristic vibrations (e.g., broadband high-frequency disturbances generated by hydrogen leakage) when the reactive gas is ejected through microcracks.

[0040] The edge processing layer integrates two types of signals: vibration and concentration. If the concentration continues to rise within the time window and exceeds the micro-leakage threshold, and the corresponding vibration parameter continues to exceed the preset vibration threshold, then it is determined that a micro-leakage has occurred in the second contact surface area.

[0041] This "dual-parameter cross-validation" design effectively avoids the risk of false alarms from a single sensor.

[0042] S40. When a leak is detected, the electrically controlled shut-off valve is triggered to close and the ventilation system is started.

[0043] In this step, it should be noted that when S30 determines that there is a trace leak, the edge processing layer immediately sends a shut-off command to the electronically controlled shut-off valve to quickly block the continued delivery of active gas; at the same time, it sends a start command to the ventilation system to dilute and disperse the active gas accumulated in the monitoring space through a preset air volume.

[0044] If the reactive gas is hydrogen, shutting off the valve and rapid ventilation are particularly critical: hydrogen diffuses quickly and has a wide explosion limit range, so the source must be cut off immediately at the initial stage of the leak and the concentration of hydrogen in the environment must be reduced as soon as possible to prevent the risk of explosion.

[0045] The technical solution implemented in this embodiment enables dual-path monitoring of embrittlement damage caused by reactive gases and minute leaks. The first contact surface monitors the "structural damage source," while the second contact surface monitors the "leakage manifestation end," forming a complete monitoring link from microcracks to leaks.

[0046] Significantly improves the sensitivity and reliability of leak detection. Multimodal sensing (stress wave, guided wave, vibration, concentration) is fused and determined through edge computing, avoiding false detections and missed detections caused by traditional single sensing methods.

[0047] It enables cloud-edge linkage and ultra-real-time response for safe handling. The edge processing layer can complete risk assessment and drive shutdown and ventilation actions in milliseconds, while the cloud platform realizes historical data accumulation, dynamic threshold optimization, and equipment management.

[0048] This method is applicable to high-risk gases with strong diffusivity and high reactivity, such as hydrogen. Compared to traditional pressure / flow monitoring solutions, this embodiment models the molecular-scale characteristics, embrittlement mechanisms, and trace leakage modes of reactive gases such as hydrogen, resulting in a highly specialized monitoring strategy.

[0049] The overall safety level of underground reactive gas pipelines will be improved. Through a complete closed-loop system of "structural strength assessment + trace leak detection + cloud-edge collaborative response", proactive intervention can be achieved in the early stages of an accident, effectively preventing the leakage from escalating and secondary disasters from occurring.

[0050] Furthermore, prior to S10, it also included: S50. Construct a three-dimensional pipeline model of the underground active gas pipeline based on the design data and field measurement data. In this step, it should be noted that the three-dimensional pipeline model is based on design data such as as-built drawings, BIM model, pipeline material, burial depth, slope, and interface type. The actual laying location, spatial direction, and node coordinates of the pipeline are obtained through methods such as on-site distance measurement, ground-penetrating radar detection, and pipeline trackers.

[0051] The model is constructed using a unified three-dimensional coordinate system, which presents each pipe section, each connector, and its contact surface as a geometric entity, ensuring that the mapping of subsequent monitoring points has an accurate spatial reference.

[0052] For preferred hydrogen pipeline scenarios, the model can also overlay layers of hydrogen safety-related environmental parameters, such as hydrogen detection zones, ventilation zones, and enclosed sections of pipe corridors, to provide a basis for subsequent monitoring space delineation.

[0053] S51. In the 3D pipeline model, the physical locations corresponding to the joints between two adjacent pipelines and the connection between the valve body and the corresponding pipeline are marked as high-risk areas. It should be noted in this step that the marking of high-risk areas is based on the physicochemical properties of reactive gases (e.g., hydrogen has a small molecular size, strong diffusivity, and is prone to embrittlement). Due to factors such as abrupt changes in material structure, stress concentration, and exposure of the inner liner termination section at the contact surface of the connector, these areas are highly susceptible to becoming the starting point for reactive gas embrittlement and leakage.

[0054] Therefore, the system automatically identifies all structural nodes that meet the following conditions and marks them as high-risk areas: the first contact surface of the pipe-connector; the second contact surface of the pipe-connector; the connection end between the pipe and the valve body; and structural discontinuities such as welded ends and plug-in ends between pipe sections.

[0055] This step is particularly critical in hydrogen pipeline embodiments because hydrogen embrittlement and hydrogen diffusion often occur first at these interface locations.

[0056] S52. Generate the high-risk coordinates of each high-risk area in the 3D pipeline model. In this step, it should be noted that the system automatically assigns a unique spatial coordinate label to each high-risk area and establishes a two-way mapping between it and the actual geographic coordinates (latitude and longitude).

[0057] These high-risk coordinates are used to: accurately guide sensor installation; achieve spatial binding between sensor nodes and 3D models; and facilitate subsequent leak location, visualization, and inspection path generation.

[0058] For hydrogen-based implementations, generating high-risk coordinates is particularly important because hydrogen leaks are characterized by rapid diffusion and uneven accumulation, requiring precise spatial calibration to achieve regional-level identification.

[0059] S53. Identify enclosed spaces adjacent to each high-risk area as monitoring spaces. In this step, it should be noted that the so-called "monitoring space" includes various limited spaces that may generate reactive gas accumulation, such as: underground valve wells, enclosed sections of pipe corridors, relatively enclosed spaces in tunnels, and cavities formed when pipelines pass through structures. The system automatically identifies enclosed or semi-enclosed spaces within a preset range (such as 1–3m) from high-risk areas based on the three-dimensional model and marks them as the corresponding monitoring spaces.

[0060] This step is especially important if the active gas is hydrogen, because hydrogen has a low density and can easily accumulate at the top of a confined space, so the location of the hydrogen monitoring points needs to be determined in advance.

[0061] S54. Establish the mapping relationship between each high-risk area and the corresponding monitoring space in the three-dimensional pipeline model. In this step, it should be noted that this step is used to clarify the one-to-one correspondence between structural damage monitoring points (such as near the first contact surface) and concentration monitoring points (within the monitoring space).

[0062] The system calculates the potential leak location in high-risk areas based on factors such as spatial distance, pipeline orientation, and gas diffusion path, and binds it to the monitoring space to form the following mapping link: First contact surface (crack initiation zone) → Second contact surface (exudation zone) → Adjacent monitoring space (accumulation zone) This mapping relationship ensures that structural damage monitoring and gas leak monitoring maintain a logical connection, concentration changes can indicate the corresponding structural damage location, and subsequent early warning and linkage control have a clear target.

[0063] In the hydrogen pipeline embodiment, due to the extremely fast diffusion rate of hydrogen, this mapping can significantly improve the accuracy of leak location and avoid misjudgment or missed detection.

[0064] The technical solution implemented in this embodiment enables a fully digital spatial representation of underground reactive gas pipelines from their structural health state to their gas leakage state. By constructing a three-dimensional pipeline model and clearly defining high-risk areas, monitoring spaces, and the mapping relationship between the two, the platform establishes a clear spatial correlation between structural damage signals and concentration change signals, ensuring that subsequent intelligent monitoring and risk identification have a physical basis and spatial constraints.

[0065] The introduction of 3D models effectively avoids the shortcomings of traditional pipeline monitoring, such as "point-based monitoring and difficulty in eliminating blind spots." It enables accurate mapping of sensor deployment, high-risk area location, and leak tracing in 3D space, significantly improving monitoring accuracy. Furthermore, this spatial mapping system supports concentration field simulation, leak diffusion prediction, and automatic inspection path planning, providing structured input for subsequent cloud-based big data computation and parameter optimization.

[0066] This technical solution is particularly advantageous for typical reactive gases such as hydrogen. Hydrogen is characterized by its small molecular size, easy diffusion, embrittlement, and wide explosive limits, making its leakage risks both insidious and sudden. This embodiment precisely defines the hydrogen embrittlement zone, the hydrogen seepage zone, and the hydrogen accumulation space, enabling the monitoring system to proactively protect against the unique mechanisms of hydrogen. The overall solution forms a tight link between "structural damage - seepage leakage - spatial accumulation," allowing the platform to make predictive judgments before actual hazards occur, thereby significantly improving the intrinsic safety level of underground hydrogen pipelines.

[0067] Furthermore, S30's "determining whether a trace amount of reactive gas leakage has occurred in a high-risk area based on pipeline vibration parameters and reactive gas concentration parameters" specifically includes: S31. Within a preset detection time window, the concentration parameters of the reactive gas in the monitoring space are sampled multiple times, and the concentration change trend is calculated. It should be noted that the core of this step is to collect time-series data and identify the changing trends of the dynamic characteristics of the reactive gas concentration in the monitoring space.

[0068] The preset detection time window can be set according to the diffusion rate of reactive gas, the volume of the monitoring space, and the ventilation capacity of the pipe gallery; for example, it can be set to a range of 10 to 120 seconds.

[0069] In the actual sampling process: Concentration data is collected in real time by laser gas sensors deployed in the monitoring space. These sensors have a resolution of ppm or higher. The system performs trend analysis on the sampling sequence, such as linear growth fitting, second-order fitting, exponential growth identification, and rate of change calculation. Concentration change indicators (such as slope, growth rate, average increment, etc.) are generated for subsequent leak detection.

[0070] If the object being monitored is hydrogen, due to the low molecular density and fast diffusion rate of hydrogen, the sampling frequency in this step is usually higher (e.g., 1–10 Hz) in order to quickly capture early signs of gas accumulation at the top or in a localized area of ​​the monitoring space.

[0071] S32. When the concentration of reactive gas in a high-risk area continuously rises within the detection time window and exceeds a preset trace leakage threshold, and the corresponding pipeline vibration parameters continuously exceed a preset vibration threshold during the same period, a trace reactive gas leak is determined to have occurred in the high-risk area. It should be noted that this step employs a "dual-factor joint judgment mechanism," using the coordinated verification of concentration increase signals and structural acoustic signals to avoid false alarms. Specifically, if the reactive gas concentration in the monitored space continuously rises and exceeds the trace leakage threshold (e.g., 3 ppm, 10 ppm, etc., which can be set according to the gas type), it is considered an initial indication of gas migration into the monitored space. For hydrogen, the leakage concentration threshold can be lower, as the lower explosive limit of hydrogen is 4%, and its early accumulation requires higher sensitivity monitoring. Pipeline vibration data is obtained by a fiber optic acoustic sensor (DAS), which can capture characteristic frequency band vibrations caused by leaked gas ejection, seepage, or crossing interfaces; the system determines whether the preset vibration threshold is exceeded based on indicators such as vibration amplitude, frequency components, and energy changes.

[0072] Joint judgment logic: Continuous increase in concentration indicates that gas is gradually accumulating in the monitoring space; Continuous abnormal vibration parameters indicate flow disturbance caused by micro-cracks in the pipe wall or contact surface. If both conditions are met, it is determined that a minor leak has occurred in the high-risk area.

[0073] This joint strategy avoids the following sources of misjudgment: The concentration has increased but the source is unknown (e.g., external gas is introduced during surrounding operations). There is only abnormal vibration but no gas leakage (e.g., mechanical vibration interference, environmental noise).

[0074] Especially in the hydrogen embodiment, this step can effectively identify small-scale leaks that form in a short time after hydrogen propagates from the crack to the second contact surface, significantly improving the accuracy of early leak identification.

[0075] The technical solution implemented in this embodiment enables agile identification and highly reliable determination of minute leaks in high-risk areas. By sampling the concentration of reactive gas multiple times within the detection time window and calculating its changing trend, the system can capture early signals of initial gas accumulation in the monitoring space, unaffected by instantaneous disturbances or random fluctuations. Simultaneously, combined with synchronous analysis of pipeline vibration parameters, the system possesses the ability to verify seepage behavior from the perspective of structural disturbances, thereby cross-validating two different physical quantities: "concentration detection" and "acoustic monitoring."

[0076] This two-factor detection mechanism makes the identification of minute leaks more accurate and reliable. For reactive gases such as hydrogen, which have high diffusivity and strong embrittlement, when a microcrack formed at the first contact surface extends to the second contact surface, it often leaks out in a very small flux. This embodiment can detect the presence of hydrogen before it reaches a dangerous concentration, thus achieving early interception of potential accidents.

[0077] Through the joint judgment method in this embodiment, the entire system can avoid frequent false alarms while maintaining high sensitivity, ensuring monitoring quality, reliability and safety in engineering applications, and providing accurate triggering conditions for automatic emergency response of pipelines (such as shut-off and ventilation), thus building a more targeted early leakage protection system for active gas pipelines.

[0078] Furthermore, S20, "determining whether the pipeline strength is within the preset safety threshold" includes: S21. Feature extraction is performed on the stress wave signal to obtain the acoustic emission event count and cumulative energy, forming a crack activity index. In this step, it should be noted that the feature extraction of the stress wave signal mainly targets the stress wave signal collected by the acoustic emission sensor array deployed near the first contact surface. The edge processing layer first performs noise reduction, filtering, and thresholding on the original acoustic emission waveform to identify acoustic emission events related to the initiation and propagation of brittle microcracks; then, the acoustic emission event count per unit time is counted, and the waveform energy of each event is integrated to obtain the cumulative energy. The event count reflects the frequency of crack activity, and the cumulative energy reflects the energy release level during crack propagation; together, they constitute the "crack activity index" characterizing the degree of crack evolution in the region near the first contact surface. Under the preferred hydrogen gas condition, since hydrogen-induced embrittlement crack propagation typically exhibits high-frequency, small-scale, and multiple-burst characteristics, the acoustic emission event count and cumulative energy can sensitively reflect early changes in hydrogen-induced embrittlement damage.

[0079] S22. Analyze the guided wave time-domain signal to obtain the pipe wall thickness reduction rate data. In this step, it should be noted that the guided wave time-domain signal analysis utilizes the guided wave excitation and response signals acquired by ultrasonic thin-walled guided wave detectors deployed on the outer walls of the pipes on both sides of the first contact surface to detect changes in the cross-sectional stiffness and thickness of the pipe wall along the axial direction. Specifically, the system performs time-history and spectrum analysis on the guided wave time-domain signal. By comparing changes in parameters such as guided wave propagation time, amplitude attenuation, and dispersion characteristics at different times and detection periods, the wall thickness reduction of the pipe section adjacent to the first contact surface is inferred. Combined with the reference guided wave response obtained during initial calibration, the wall thickness reduction rate relative to the reference state at the current moment can be calculated. For pipelines transporting hydrogen, since hydrogen-induced corrosion and hydrogen embrittlement can lead to local wall thickness loss or changes in elastic modulus, even minor changes in guided wave propagation characteristics can be quantitatively reflected through this step, thus providing direct evidence for structural strength assessment.

[0080] S23. The crack activity index and the pipe wall thickness reduction rate are fused together to obtain structural state parameters. In this step, it should be noted that the crack activity index and the pipe wall thickness reduction rate are fused together to obtain "structural state parameters" that comprehensively reflect the structural health status of the first contact surface region. In practical implementation, the edge processing layer can first normalize the crack activity index and the pipe wall thickness reduction rate to make them comparable on the same numerical scale; then, according to preset weights or weights determined based on statistical / machine learning methods, the two types of indices are weighted and summed or nonlinearly combined to form dimensionless structural state parameters. This parameter considers both the crack activity caused by brittle microcracks and the wall thickness reduction caused by factors such as corrosion and stress coupling, thus characterizing the overall load-bearing capacity and structural safety margin of the pipeline in the region adjacent to the first contact surface with a single index. In the preferred hydrogen scenario, this structural state parameter can comprehensively characterize the wall thickness degradation caused by hydrogen-induced embrittlement cracking and hydrogen-induced corrosion simultaneously.

[0081] S24. When the structural state parameters exceed the safety threshold, the pipeline strength safety assessment is deemed abnormal. It should be noted that this step, where the structural state parameters exceed the safety threshold, constitutes an assessment and classification of the aforementioned fusion results. The safety threshold can be set in the initial stage of the system based on design strength, material properties, allowable wall thickness reduction ratio, and historical operating experience, or it can be adaptively optimized through a cloud platform based on long-term operating data. When the real-time calculated structural state parameters are less than or within the safety threshold range, the system determines that the pipeline strength in the area adjacent to the first contact surface is at an acceptable level, allowing entry into the minor leakage assessment process. Once the structural state parameters exceed the safety threshold, the system considers that there is a significant risk of strength degradation or failure in that area, triggering a structural risk warning and enabling the implementation of safety control strategies such as pressure reduction operation and shutdown maintenance. In the hydrogen embodiment, this assessment is particularly critical, as it can provide early warning of potential hydrogen-induced embrittlement failure before any obvious leakage or concentration anomaly occurs.

[0082] The technical solution implemented in this embodiment enables the establishment of a quantitative assessment mechanism for the structural safety status of critical connection areas in underground reactive gas pipelines at the micro-damage level. Through joint analysis of stress wave and guided wave signals, the system can not only identify the degree of crack activity caused by embrittlement effects but also capture changes in wall thickness caused by corrosion or stress, thus incorporating the two key factors of "crack activity" and "residual wall thickness" into a unified structural state parameter. Comparison of this parameter with a preset safety threshold transforms structural strength assessment from a qualitative judgment into a measurable and traceable quantitative process. For reactive gas pipelines, such as those carrying hydrogen, the strength assessment steps provided in this embodiment can identify high-risk areas before microcracks develop into penetrating leakage channels. By connecting with subsequent micro-leakage assessment and emergency response steps, a layered protection system of "structure first, leakage later" is formed, effectively improving the intrinsic safety level and operational reliability of the entire pipeline under high-pressure hydrogen and other reactive gas transportation conditions.

[0083] Furthermore, S23, "integrating the crack activity index with the pipe wall thickness reduction rate in the calculation," includes: The crack activity index and the pipe wall thickness reduction rate were normalized respectively.

[0084] The weighting coefficients of crack activity index and pipe wall thickness reduction rate in structural state parameters were determined by entropy weight method.

[0085] The normalized indices are weighted and summed according to the weighting coefficients to obtain the dimensionless structural state parameter values.

[0086] In this step, it should be noted that the "crack activity index" and the "wall thickness reduction rate" are derived from the acoustic emission sensor array and the ultrasonic thin-guided wave detector, respectively. The two reflect different physical mechanisms, different dimensions, and different rates of change. In order to achieve comparability and integrated analysis between the two, they need to be normalized first.

[0087] Specifically, in this embodiment, Min–Max normalization can be used to distribute each index within the [0,1] interval, which preserves its trend and facilitates subsequent weight calculation.

[0088] After normalization, the weight coefficients of the two indicators in the structural state parameters are automatically determined using the entropy weight method. The entropy weight method can adaptively assign higher weights to indicators with stronger discriminative capabilities based on the dispersion and information distribution characteristics of each indicator. For example, in the hydrogen scenario of the preferred embodiment, early hydrogen embrittlement damage often manifests as high-frequency jumps in the crack activity index, while the pipe wall thickness reduction rate is relatively gradual in the early stages; the entropy weight method thus automatically increases the weight of the crack activity index, enhancing the sensitivity to early characteristics of hydrogen embrittlement.

[0089] Subsequently, based on the determined weighting coefficients, the normalized indices are weighted and summed to obtain dimensionlessly uniform structural state parameters. These parameters are mathematically dimensionless, but in an engineering sense, they comprehensively reflect the structural health status of the region adjacent to the first contact surface. Larger parameter values ​​indicate more frequent crack activity, more significant wall thickness reduction, and lower structural safety margin.

[0090] In the preferred scenario of hydrogen transportation, since hydrogen embrittlement damage and hydrogen corrosion damage often evolve in a coupled manner and have strong local abrupt changes, the above-mentioned fusion processing can unify the two key damage information of dynamic microcrack propagation and pipe wall degradation into a quantifiable health index, thereby achieving accurate characterization of the structural failure risk caused by reactive gas.

[0091] The technical solution implemented in this embodiment can significantly improve the quantitative assessment of structural damage. By normalization and entropy weighting, this embodiment can integrate data from two different sources and with different physical meanings—acoustic emission and guided wave detection—into a single structural state parameter. This transforms the damage assessment process from fragmented judgment to a unified index measurement, providing a more stable and consistent basis for subsequent strength and safety determination.

[0092] Furthermore, this fusion method can enhance the system's sensitivity to the "embrittlement-leakage" failure mechanism unique to reactive gases.

[0093] Especially in highly reactive gas scenarios such as hydrogen, hydrogen embrittlement exhibits sudden cracking activity, while pipe wall thinning has a cumulative characteristic, resulting in significant temporal differences between the two types of damage. The entropy weight method can automatically adjust the weights, enabling the system to maintain high sensitivity to key risk characteristics at different damage stages, thereby improving early identification capabilities.

[0094] Furthermore, by generating dimensionless structural state parameters that can be calculated in real time, this embodiment achieves visualization, quantification, and traceability of structural safety assessment.

[0095] This parameter can serve as a unified triggering basis for determining the execution intensity of the edge processing layer, adjusting the early warning level, and triggering safety controls, enabling the entire cloud-edge collaborative system to have adaptive and dynamic adjustment capabilities, and significantly enhancing the inherent safety level of pipelines under complex working conditions and during long-term operation.

[0096] Furthermore, after S40, it also includes: S41. After the electrically controlled shut-off valve is triggered, the concentration of reactive gas in the leak isolation zone is continuously monitored using a laser gas sensor. It should be noted that S41 occurs after the electrically controlled shut-off valve has been triggered and closed. At this point, the leak source is physically isolated from the main pipeline network, but residual reactive gas may still accumulate in the leak isolation zone. Therefore, continuous monitoring of this local space is necessary.

[0097] To ensure monitoring accuracy, this embodiment employs a laser gas sensor with ppm-level resolution, capable of maintaining stable detection even at extremely low leakage levels of reactive gas (preferably hydrogen). The sensor is installed in the monitoring space corresponding to the second contact surface, enabling it to capture minute changes in gas concentration in that area in real time and transmit the data at a high sampling rate to the edge processing layer for subsequent leakage level determination.

[0098] If the target gas is hydrogen, the laser gas sensor can use TDLS (tunable diode laser absorption spectroscopy) or TDLAS technology to avoid errors caused by the rapid diffusion and weak adsorption of hydrogen, and to ensure a highly sensitive response to changes in hydrogen concentration.

[0099] S42. If the reactive gas concentration parameter is lower than the first safety threshold, it is determined to be a controllable leak, a leak alarm is generated, the sampling frequency is increased to the first preset frequency, and the ventilation system is controlled to operate at the first preset power. In this step, it should be noted that when the edge processing layer determines that the reactive gas concentration in the leak isolation zone is lower than the first safety threshold, the system considers the current leak scale to be small and the leak rate to be low, and this state is defined as "controllable leak".

[0100] The first safety threshold is set based on the physicochemical properties of the target reactive gas. For example, in the preferred embodiment of hydrogen, this threshold is typically 4% (volume fraction) below the lower explosive limit of hydrogen, ensuring that the environment remains within an absolutely safe range.

[0101] In this state, to maintain high-frequency monitoring of the leak area, the system will automatically increase the sampling frequency of the laser gas sensor, switching it from the normal monitoring mode to a first preset frequency mode, for example, from 1Hz to 10Hz, to quickly capture changes in the leak. Simultaneously, the ventilation system only needs to operate stably at the first preset power, generally corresponding to normal exhaust or low-speed extraction, to ensure that leaked gas does not accumulate and escalate the risk.

[0102] S43. If the active gas concentration parameter reaches or exceeds the second safety threshold, it is determined to be an uncontrollable leak. Within milliseconds, the edge processing layer controls the ventilation system to operate at the second preset power, and simultaneously sends the highest-level accident alarm to the cloud platform. The second safety threshold is higher than the first safety threshold, and the second preset power is greater than the first preset power. It should be noted in this step that... If the concentration of reactive gas in the leak isolation zone reaches or exceeds the second safety threshold, it indicates that the current gas accumulation rate is too fast or the leakage is large, which is a dangerous state. This step defines this state as "uncontrollable leakage". The second safety threshold is higher than the first safety threshold. In the preferred embodiment of hydrogen, it can be matched with or slightly lower than the lower explosive limit of hydrogen (4% volume fraction) to ensure that the system has taken emergency intervention before entering the potential explosion zone.

[0103] In the event of an uncontrollable leak, the edge processing layer leverages its localized computing capabilities to directly control the ventilation system to increase its operating power to the second preset power, i.e., the maximum extraction capacity, within milliseconds, thereby expelling the reactive gas from the isolation zone as quickly as possible. Simultaneously, the edge processing layer immediately sends the highest-level emergency alarm to the cloud platform via a high-speed communication link, enabling the remote management system to quickly intervene, such as automatically initiating emergency response procedures and coordinating with ground safety devices.

[0104] This millisecond-level linkage capability is something that conventional cloud platforms cannot achieve, and it is a significant advantage of the "cloud-edge collaborative architecture" adopted in this invention.

[0105] The technical solution implemented in this embodiment can form a complete chain of "rapid identification - graded response - proactive intervention" in leakage emergency response.

[0106] By entering continuous monitoring mode the instant the shut-off valve closes, the system ensures real-time sensing of gas dynamics within the isolation zone, fundamentally preventing missed detections and blind spots. The high sensitivity of the laser gas sensor allows for immediate response to minute concentration fluctuations, making risk identification more accurate.

[0107] This implementation method can also automatically identify the leakage level based on the leakage scale and execute corresponding treatment measures to achieve intelligent and differentiated safety control.

[0108] For controllable leaks, the area is ventilated in a controlled manner under safe conditions by increasing the sampling frequency and using low-power ventilation. For uncontrollable leaks, full-power ventilation and accident reporting are performed at millisecond speeds, significantly shortening the time between the accumulation of hazardous substances and explosive concentrations, thereby maximizing on-site safety.

[0109] Furthermore, through cloud-edge collaboration, the system possesses the dual capabilities of rapid response and global decision-making.

[0110] The edge processing layer is responsible for real-time judgment and emergency response, while the cloud platform can perform data accumulation, analysis, trend prediction and strategy optimization after an accident, enabling the overall safety system to have continuous learning capabilities, thereby continuously improving its response level to reactive gas leaks in long-term operation.

[0111] Furthermore, the methods also include: S60. The cloud platform receives crack activity indicators, pipe wall thickness reduction rate, vibration parameters, active gas concentration parameters, and corresponding structural risk warnings, trace leakage judgment results, and control command execution results uploaded by the edge processing layer, forming a historical operation dataset. It should be noted that the goal of this step is to build a data foundation covering long-term operational status for subsequent threshold optimization and system adaptive capability improvement.

[0112] The cloud platform, acting as a centralized data processing center, continuously receives multi-source monitoring data uploaded from the edge processing layer, including: Crack activity index: Composed of characteristics such as acoustic emission event count and cumulative energy, it is used to describe the activity of generation and propagation of microcracks caused by hydrogen embrittlement (or other reactive gas embrittlement). Pipe wall thickness reduction rate: obtained by guided wave detection signal analysis, reflecting the trend of material thickness reduction caused by active gas infiltration, corrosion or hydrogen-induced damage; Vibration parameters: acquired by fiber optic acoustic sensors, reflecting the vibration characteristics of the outer wall that may be excited by the vibration of leaking airflow; Active gas concentration parameter (preferably hydrogen concentration): acquired by a laser gas sensor, providing a leak indication at the ppm level; And corresponding structural risk warning results, trace leakage judgment results, control command execution feedback and other status variables.

[0113] The cloud platform organizes, denoises, and labels the above data in chronological order, and constructs historical operating datasets across different operating conditions (temperature, pressure, season, operating load) to provide long-term, stable, and high-quality data samples for subsequent learning models.

[0114] S61. Based on historical operational datasets, optimize the critical strength threshold used for strength safety assessment, the micro-leakage threshold used to determine micro-leakage, the first safety threshold, and the second safety threshold using a machine learning model, generating updated threshold parameters. It should be noted that the core of this step is to use a machine learning model to perform closed-loop optimization of the system's key thresholds, allowing them to be dynamically updated as operating conditions change.

[0115] The optimized thresholds include: Critical strength threshold: used to determine whether a structure is healthy and safe; Micro-leakage threshold: used to determine if there are early signs of leakage; First safety threshold and second safety threshold: used to distinguish between controllable and uncontrollable leaks.

[0116] Machine learning models extract multidimensional features based on historical operating datasets, such as: material expansion behavior caused by changes in ambient temperature, changes in stress characteristics caused by fluctuations in pipeline operating pressure, the sensitivity of crack propagation rate to changes in gas type (preferably hydrogen), and the correlation between concentration, vibration and crack characteristics under different operating conditions.

[0117] Subsequently, a regression model (preferably a random forest regression model) is constructed with operating parameters as input and the optimal threshold as output, and the optimal parameter combination of the model is determined by cross-validation.

[0118] Using indicators such as root mean square error (RMSE) as optimization targets, the generated threshold parameters can adapt to environmental changes while avoiding false alarms or missed alarms.

[0119] The final output is the updated set of threshold parameters.

[0120] S62. Send the updated threshold parameters to the edge processing layer. It should be noted that the purpose of this step is to effectively transmit the cloud-optimized threshold parameters to the edge processing layer, enabling the monitoring system to use the latest strategies for judgment and response in real time.

[0121] After receiving the updated threshold, the edge processing layer writes it into the local parameter library and calls it in subsequent logic such as structural strength assessment, micro-leakage determination, and controllable / uncontrollable leakage determination.

[0122] This step ensures that the system can continuously adapt to changes in monitoring sensitivity caused by different seasons, different operating pressures, and different degrees of material aging, thereby achieving self-evolution of the monitoring model and long-term reliable operation.

[0123] The technical solution implemented in this embodiment enables the monitoring system to possess long-term learning and adaptive evolution capabilities. By continuously receiving multimodal data from the edge processing layer in the cloud, a historical operational dataset covering the entire lifespan is formed. This allows the system to not only record every crack propagation and every early sign of leakage, but also capture the impact of different operating conditions on monitoring sensitivity, thus providing a solid foundation for intelligent optimization.

[0124] This embodiment can also achieve dynamic optimization of threshold parameters, enabling the monitoring strategy to automatically adjust with changes in the environment.

[0125] In applications where temperature, pressure, and gas type (such as hydrogen) vary significantly, traditional fixed thresholds are prone to false alarms or missed alarms. By using machine learning models to extract operating condition-related features and outputting optimal thresholds, the system can maintain high accuracy and a low false alarm rate even in variable environments.

[0126] In addition, by sending optimized threshold parameters from the cloud to the edge processing layer, the system forms a cloud-edge closed loop, improving response accuracy.

[0127] The edge side uses new thresholds for real-time judgment, making monitoring and decision-making more aligned with the current operating status, further improving the accuracy of micro-crack early warning, micro-leakage identification and leakage classification and handling, and enhancing the overall safe operation capability of the active gas pipeline network (preferably hydrogen pipeline network).

[0128] Furthermore, S61's "optimization of the critical strength threshold used for strength safety assessment and the micro-leakage threshold used to determine micro-leakage through machine learning models" specifically includes: Based on historical operational datasets, the deformation characteristics of the pipeline body under different ambient temperatures and pipeline operating pressure conditions are extracted. It should be noted that the first part of S61 aims to extract environmental and load factors that are highly relevant to structural safety from long-term operational data.

[0129] In active gas (preferably hydrogen) pipelines, both ambient temperature and operating pressure cause periodic thermal expansion and contraction and pressure deformation of the pipeline body, thereby affecting: the baseline level of crack activity indicators, the rate of change of pipe wall thickness reduction, material yield strength, and sensitivity to hydrogen-induced damage. For example, when the ambient temperature rises, the diffusion coefficient of the metallic material increases, the active gas permeation rate may increase, and the crack propagation law will also change accordingly; when the operating pressure rises, the additional stress on the inner wall of the pipeline increases, which may accelerate the transformation of hydrogen-induced (or other active gas-induced) microcracks into macrocracks.

[0130] Therefore, based on historical operating datasets, the cloud platform extracts multidimensional features including but not limited to the following: the coupling pattern between periodic temperature fluctuations and crack activity changes, the sensitivity of waveguide thickness attenuation under different pressure levels, and the comprehensive impact of temperature-pressure combination conditions on leak precursor signals, and uses these features as training inputs for subsequent machine learning models.

[0131] A random forest regression model is established, taking ambient temperature, pipeline pressure, and historical crack propagation rate as inputs and the optimal safety threshold as output. It should be noted that the second part of S61 uses a regression model to fit the nonlinear mapping relationship between "complex operating conditions → optimal threshold".

[0132] The random forest regression model is preferred because it can handle nonlinear and high-dimensional features, is insensitive to noise, can automatically assess the importance of features such as temperature, pressure, and crack propagation rate, and can achieve robust predictions even with limited data.

[0133] The model inputs include: ambient temperature sequence, pipeline operating pressure sequence, crack propagation rate (derived from changes in acoustic emission event density), guided wave attenuation characteristics (reflecting the trend of thickness weakening), and gas concentration change gradient (particularly applicable to easily diffusing gases such as hydrogen).

[0134] The model outputs the optimal critical intensity threshold, the optimal trace leakage threshold, the best matching value between the first / second safe concentration threshold and the current operating conditions. This type of model can automatically adjust the thresholds based on factors such as the difference in diffusion rate of active gas (preferably hydrogen) under different operating conditions and changes in hydrogen embrittlement sensitivity.

[0135] The model hyperparameters are determined using cross-validation, and the root mean square error (RMSE) is used as the optimization objective to dynamically adjust the critical strength threshold, micro-leakage threshold, first safety threshold, and second safety threshold. It should be noted that the third part of S61 uses k-fold cross-validation to determine the key hyperparameters of the random forest, such as the number of decision trees, maximum tree depth, and feature sampling ratio for each split. Simultaneously, the root mean square error (RMSE) is used as the optimization objective to dynamically adjust the following thresholds: critical strength threshold, micro-leakage threshold, first safety threshold, and second safety threshold. The application of RMSE here ensures that the thresholds have higher generalization ability across multiple operating environments, avoiding adaptation to only a specific working condition.

[0136] When the active gas is hydrogen, due to hydrogen's strong diffusivity and high embrittlement sensitivity, the model can automatically capture the changing patterns of hydrogen leakage and embrittlement behavior under different temperatures, pressures, and high-risk interface conditions, thereby outputting more accurate dynamic thresholds.

[0137] The technical solution implemented in this embodiment enables the system to adapt to changes in operating conditions. By extracting the coupled influence characteristics of temperature and pressure on crack propagation and thickness reduction, the system can "understand" the impact of reactive gases (especially hydrogen) on materials, leakage trends, and monitoring signals under complex environments, thereby achieving real-time perception of changes in operating conditions.

[0138] This embodiment can also establish a nonlinear mapping between high-dimensional features and the optimal threshold through a machine learning model, thereby improving the scientific nature and sensitivity of threshold setting.

[0139] Especially in hydrogen pipeline networks, the rate of hydrogen embrittlement and micro-leakage behavior are greatly affected by temperature and pressure. Traditional fixed threshold strategies are difficult to adapt to changing environments, while this technology can automatically optimize based on real data, making the judgment criteria more in line with reality.

[0140] Furthermore, by employing cross-validation and error minimization strategies, the threshold output by the model is made to have higher stability and generalization ability.

[0141] Regardless of seasonal changes, adjustments to operating methods, or aging of pipeline materials, the system can ensure operation under optimal thresholds, significantly reducing false alarms and missed alarms, and improving the reliability of microcrack early warning and micro-leakage detection.

[0142] In summary, this embodiment achieves intelligent, dynamic, and adaptive updating of thresholds, making the AIoT cloud-edge collaborative IoT management platform an active safety system with long-term evolution capabilities, thereby significantly enhancing the intrinsic safety level of underground active gas (preferably hydrogen) pipeline networks.

[0143] The embodiments described above are merely illustrative of several implementations of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the scope of protection of this application. Therefore, the scope of protection of this patent application should be determined by the appended claims.

Claims

1. An AIoT cloud-edge collaborative IoT management platform, characterized in that, Applied to underground active gas pipelines, the pipeline includes: pipeline; A corrosion-resistant inner liner coated on the inner wall of the pipe; And the connectors that connect to the end of the pipe; The main body surface of the connector includes a first surface and a second surface; the end face of the pipe and the end face of the anti-corrosion inner liner abut against the first surface to form a first abutting surface; the outer wall of the pipe abuts against the second surface to form a second abutting surface. Under active gas delivery conditions, active gas molecules penetrate into the first contact surface and accumulate there, inducing microcracks formed by the active gas embrittlement in the pipe body and connector body at the first contact surface; the microcracks extend along the first contact surface and extend to the second contact surface, forming a seepage channel for the active gas. The platform includes: A dedicated sensor network is provided, which has multiple sensor nodes for collecting structural damage signals at the first contact surface and gas leakage signals at the second contact surface. An edge processing layer is used to receive and process the structural damage signal and the gas leak signal; The cloud platform communicates with the edge processing layer for data interaction and storage management.

2. The AIoT cloud-edge collaborative IoT management platform according to claim 1, characterized in that, The sensing nodes deployed on the first contact surface include: Acoustic emission sensor array, used to acquire stress wave signals generated by the propagation of brittle microcracks; An ultrasonic thin guided wave detector is used to acquire the guided wave time domain signal of the pipe segment adjacent to the first contact surface; The sensing nodes deployed on the second contact surface include: Fiber optic acoustic sensors are used to collect pipeline vibration parameters caused by reactive gas leaks. A laser gas sensor is used to collect the concentration of active gases in the monitoring space corresponding to the second contact surface.

3. An AIoT cloud-edge collaborative IoT management method, applied to the AIoT cloud-edge collaborative IoT management platform as described in claim 1 or 2, characterized in that, The method includes: S10: Receive and process stress wave signals, guided wave time domain signals, pipeline vibration parameters, and reactive gas concentration parameters; S20. Based on stress wave signals and guided wave time domain signals, determine whether the pipeline strength is within the preset safety threshold. If not, generate a structural risk warning and trigger a pipeline depressurization operation command. S30. If so, then determine whether a small leak has occurred in the high-risk area based on the pipeline vibration parameters and the active gas concentration parameters. S40. When a leak is detected, the electrically controlled shut-off valve is triggered to close and the ventilation system is started.

4. The AIoT cloud-edge collaborative IoT management method according to claim 3, characterized in that, Before S10, the following is also included: S50. Construct a three-dimensional pipeline model of the underground active gas pipeline based on the design data and field measurement data of the underground active gas pipeline. S51. In the three-dimensional pipeline model, the physical locations corresponding to the joint ends between two adjacent pipelines and the connection ends between the valve body and the corresponding pipelines are marked as high-risk areas. S52. Generate the high-risk coordinates of each of the high-risk areas in the three-dimensional pipeline model; S53. Designate the enclosed space adjacent to each of the high-risk areas as the monitoring space; S54. Establish the mapping relationship between each of the high-risk areas and the corresponding monitoring space in the three-dimensional pipeline model.

5. The AIoT cloud-edge collaborative IoT management method according to claim 4, characterized in that, The step S30, "determining whether a trace amount of reactive gas leakage has occurred in the high-risk area based on the pipeline vibration parameters and reactive gas concentration parameters," specifically includes: S31. Within a preset detection time window, the concentration parameters of the active gas in the monitoring space are sampled multiple times, and the concentration change trend is calculated. S32. When the concentration of reactive gas in a high-risk area continues to rise and exceeds the preset trace leakage threshold within the detection time window, and the corresponding pipeline vibration parameter continues to exceed the preset vibration threshold during the same period, it is determined that a trace reactive gas leak has occurred in the high-risk area.

6. The AIoT cloud-edge collaborative IoT management method according to claim 3, characterized in that, The step S20, "determining whether the pipeline strength is within the preset safety threshold," includes: S21. Extract features from the stress wave signal to obtain acoustic emission event counts and accumulated energy, forming a crack activity index. S22. Analyze the time-domain signal of the guided wave to obtain the pipe wall thickness reduction rate data; S23. The crack activity index and the pipe wall thickness reduction rate are combined and calculated to obtain the structural state parameters; S24. When the structural state parameters exceed the safety threshold, the pipeline strength safety assessment is deemed abnormal.

7. The AIoT cloud-edge collaborative IoT management method according to claim 6, characterized in that, The step S23, "integrating the crack activity index with the pipe wall thickness reduction rate in the calculation," includes: The crack activity index and the pipe wall thickness reduction rate were normalized respectively. The weighting coefficients of the crack activity index and the pipe wall thickness reduction rate in the structural state parameters are determined by the entropy weight method. The normalized indices are weighted and summed according to the weighting coefficients to obtain the dimensionless structural state parameter values.

8. The AIoT cloud-edge collaborative IoT management method according to claim 3, characterized in that, Following S40, the following is also included: S41. After the electrically controlled shut-off valve is triggered, the concentration of active gas in the leakage isolation zone is continuously monitored by the laser gas sensor; S42. If the active gas concentration parameter is lower than the first safety threshold, it is determined to be a controllable leak, a leak alarm is generated, the sampling frequency is increased to the first preset frequency, and the ventilation system is controlled to operate at the first preset power. S43. If the active gas concentration parameter reaches or exceeds the second safety threshold, it is determined to be an uncontrollable leak. The edge processing layer controls the ventilation system to operate at the second preset power within milliseconds, and simultaneously sends the highest level accident alarm to the cloud platform; wherein, the second safety threshold is higher than the first safety threshold, and the second preset power is greater than the first preset power.

9. The AIoT cloud-edge collaborative IoT management method according to claim 8, characterized in that, The method further includes: S60. The cloud platform receives the crack activity index, pipe wall thickness reduction rate, vibration parameters, active gas concentration parameters, and corresponding structural risk warning, micro-leakage judgment results and control command execution results uploaded by the edge processing layer, forming a historical operation dataset. S61. Based on the historical operation dataset, the critical strength threshold used for strength safety assessment, the micro-leakage threshold used for determining micro-leakage, the first safety threshold, and the second safety threshold are optimized using a machine learning model to generate updated threshold parameters. S62. Send the updated threshold parameters to the edge processing layer.

10. The AIoT cloud-edge collaborative IoT management method according to claim 9, characterized in that, The "optimization of the critical strength threshold for strength safety assessment and the micro-leakage threshold for determining micro-leakage using a machine learning model" in S61 specifically includes: Based on the historical operation dataset, the deformation characteristics of the pipeline body under different ambient temperatures and pipeline operating pressure conditions are extracted. A random forest regression model was established with ambient temperature, pipeline pressure, and historical crack propagation rate as inputs and the optimal safety threshold as output. The model hyperparameters are determined by cross-validation, and the root mean square error is used as the model optimization target to dynamically adjust the critical intensity threshold, trace leakage threshold, first safety threshold, and second safety threshold.

Images