Multi-parameter intelligent automatic inspection method and device for dangerous gas pipe network and medium

Abstract

This invention discloses a multi-parameter intelligent automated inspection method, device, and medium for hazardous gas pipeline networks. The method includes: dividing the pipeline network coverage area into several inspection sections, configuring edge computing nodes for real-time data processing in each section, and constructing a hierarchical inspection database; fusing multiple types of parameter data based on a time-series feature extraction network to generate a comprehensive feature sequence; interactively inputting the multi-type parameter data into a dynamically updated inspection model; interactively adjusting the inspection path through a dual-modal command parsing module; the multi-type parameter data including pressure fluctuation values, concentration changes, temperature gradient values, and pipeline deformation; comparing the comprehensive feature sequence with baseline operating data through a multi-level discrimination mechanism to determine the degree of pipeline network operation anomaly; transmitting control signals to the execution components through a control channel and generating an inspection report. This invention improves the real-time performance, accuracy, and intelligence level of inspections and is suitable for the safety monitoring and management of complex hazardous gas pipeline networks.

Description

Technical Field

[0001] This invention relates to a multi-parameter intelligent automated inspection method, device, and medium for hazardous gas pipeline networks, belonging to the field of hazardous gas pipeline network inspection technology. Background Technology

[0002] Traditional pipeline inspection methods rely heavily on manual periodic checks, which have problems such as limited coverage, delayed data collection, and single parameter monitoring, making it difficult to cope with the dynamic changes of complex pipeline systems.

[0003] With the expansion of pipeline networks and the increasing complexity of operating environments, monitoring of single parameters (such as gas concentration) can no longer meet the needs of safety management. The coordinated changes of multiple parameters, such as pressure fluctuations, temperature anomalies, and pipeline deformation, are often key signals for accident early warning. However, traditional inspection methods lack the ability to comprehensively analyze multiple parameters, easily leading to missed detections or misjudgments. Furthermore, pipeline operation has strong temporal characteristics; parameter change patterns differ across different time periods, making static threshold judgment models unable to capture potential long-term risk trends.

[0004] At the data processing level, traditional centralized data management models face transmission delays and computing power bottlenecks in massive monitoring data, especially in remote areas or complex terrains, where data real-time performance and reliability are difficult to guarantee. Furthermore, inspection route planning largely relies on pre-set routes and cannot be dynamically adjusted based on real-time parameter changes, leading to delayed responses in key risk areas. When sensor data is abnormal or manual intervention occurs, the lack of an effective multi-source command fusion mechanism easily leads to command conflicts or execution failures, affecting emergency response efficiency.

[0005] While some existing automated inspection systems have incorporated sensor networks, they still fall short in areas such as multi-parameter fusion, dynamic area division, and edge computing collaboration. This makes it difficult to achieve full-process intelligent control from data acquisition and analysis to regulation and execution. Therefore, an integrated inspection method is needed that can integrate multi-parameter monitoring, dynamic path optimization, hierarchical data processing, and collaborative regulation. Summary of the Invention

[0006] In view of this, the present invention provides a multi-parameter intelligent automated inspection method, system, computer equipment and storage medium for hazardous gas pipeline networks, which improves the real-time performance, accuracy and intelligence of the inspection, and is suitable for the safety monitoring and management of complex hazardous gas pipeline networks.

[0007] The first objective of this invention is to provide a multi-parameter intelligent automated inspection method for hazardous gas pipeline networks.

[0008] The second objective of this invention is to provide a multi-parameter intelligent automated inspection device for hazardous gas pipeline networks.

[0009] A third objective of this invention is to provide a computer device.

[0010] A fourth objective of this invention is to provide a computer-readable storage medium.

[0011] The first objective of this invention can be achieved by adopting the following technical solution:

[0012] A multi-parameter intelligent automated inspection method for hazardous gas pipeline networks, the method comprising:

[0013] The pipeline network coverage area is divided into several inspection sections, and edge computing nodes are configured in each section for real-time data processing to build a hierarchical inspection database.

[0014] Based on the temporal feature extraction network, multiple types of parameter data are fused to generate a comprehensive feature sequence. The multiple types of parameter data are interactively input into the dynamically updated inspection model, and the inspection path is interactively adjusted through the dual-modal command parsing module. The multiple types of parameter data include pressure fluctuation value, concentration change, temperature gradient value and pipeline deformation degree.

[0015] By comparing the comprehensive feature sequence with the baseline operating data through a multi-level discrimination mechanism, the degree of abnormality in pipeline operation is determined, and control signals are transmitted to the actuators through the control channel to generate inspection reports.

[0016] Furthermore, the process of dividing the pipeline network coverage area into several inspection sections, configuring edge computing nodes for real-time data processing in each section, and constructing a hierarchical inspection database specifically includes:

[0017] Set a pipeline network structure complexity index, traverse each pipe segment within the pipeline network coverage area, calculate its complexity index, and record the complexity of each segment. At the same time, identify dynamically changing nodes in the pipeline network and use a region segmentation algorithm to dynamically divide the pipeline network into several inspection segments.

[0018] Each inspection segment is assigned to a different edge computing node. A new global time tag is generated based on the data collection frequency, and the global time tag is sent to each computing node so that each computing node can perform data processing operations based on the assigned inspection segment and the current time tag.

[0019] The processing results of all inspection sections are integrated into the main database to generate a complete set of inspection data, resulting in a hierarchical inspection database.

[0020] Furthermore, the method of dynamically dividing the pipeline network into several inspection sections using a region segmentation algorithm specifically includes:

[0021] The movement trajectory of the inspection terminal is tracked in real time through positioning devices;

[0022] Determine the current key monitoring areas based on the movement trajectory;

[0023] Based on the current key monitoring area, the pipeline network nodes within the safe distance range are expanded outward, and the pipeline network is recursively divided into sub-regions;

[0024] The distance weight between the sub-region and the baseline is used as the data processing priority indicator. Sub-regions with the same distance weight are processed synchronously through time tags, and the sub-regions are dynamically updated in real time according to changes in the movement trajectory.

[0025] Furthermore, the time-series feature extraction network is a composite network structure consisting of memory units for capturing long-term trends and filtering units for processing short-term fluctuations.

[0026] The method for fusing multiple types of parameter data based on the temporal feature extraction network to generate a comprehensive feature sequence specifically includes:

[0027] Align different parameter data in the time dimension using a hardware clock;

[0028] Different weights are assigned based on the sensitivity and correlation of the parameter data;

[0029] Multiple types of parameter data are input into the memory unit and the filtering unit respectively, the features of the multiple types of parameter data are extracted, and the data are weighted and fused according to the weights to generate a comprehensive feature sequence.

[0030] Furthermore, the dual-modal commands include sensor commands and manual commands;

[0031] The process of adjusting the inspection path interactively through the dual-modal command parsing module specifically includes:

[0032] Upon receiving the collected sensor parameter data, the sensor parameter data is integrated into a sensor parameter sequence in chronological order. Based on the generated sensor parameter sequence, a sensor command output path adjustment command is triggered, and the inspection path is adjusted interactively according to the path adjustment command.

[0033] Upon receiving real-time human instruction data, semantic features are extracted based on the received data, triggering a path adjustment instruction for the human instruction output. The inspection path is then adjusted interactively based on the path adjustment instruction.

[0034] When an invalid signal is detected in the interaction between sensor commands and human commands to adjust the path, if the number of invalid signals is higher than the invalid threshold, a command-parameter association table is established, the confidence scores of sensor and human commands are weighted and fused, the matching degree between the dual-modal command and parameter changes is output, and a comprehensive confidence score is set. If the comprehensive confidence score exceeds the total standard, the command is considered valid and the path adjustment is triggered; otherwise, a prompt signal is issued.

[0035] Furthermore, the multi-level discrimination mechanism compares the comprehensive feature sequence with the baseline operating data to determine the degree of pipeline network operation anomaly, transmits control signals to the execution components through the control channel, and generates an inspection report, specifically including:

[0036] A baseline operating data is constructed as the inspection target, and the baseline operating data is aligned with the real-time operating data of the pipeline network in the time dimension. The baseline operating data includes baseline pressure value, baseline concentration value, baseline temperature value, and baseline deformation.

[0037] For each time point, the degree of deviation between the comprehensive feature sequence and the baseline running data is calculated. Based on the importance of the parameters, weights are assigned to each type of parameter to generate a weighted deviation degree.

[0038] The weighted deviation is mapped to the intensity range of the control signal using a standardized function, and the corresponding control signals are generated through the control channel and transmitted to the execution unit.

[0039] Based on the control signals, an inspection report is generated using a multi-level discrimination mechanism. This multi-level discrimination mechanism includes a primary discrimination mechanism that calculates deviations in real time and provides control, an intermediate discrimination mechanism that periodically evaluates the progress and effectiveness of inspections, and a high-level discrimination mechanism that comprehensively analyzes long-term inspection data and provides optimization suggestions.

[0040] Furthermore, the step of generating corresponding control signals through the control channel and transmitting the control signals to the execution unit specifically includes:

[0041] If the control channel is threshold triggered, the degree of deviation is reflected by the switch signal, and the switch control signal is transmitted to the execution unit.

[0042] If the control channel is a trend warning, the degree of deviation is displayed through the warning level, and the warning level is displayed in the monitoring interface and transmitted to the execution unit.

[0043] If the control channel is pattern matched, the degree of deviation will be indicated by the action command, and the action prompt information will be transmitted to the execution unit.

[0044] Furthermore, the actuating components include a valve regulating module, an alarm module, and a data recording module;

[0045] The valve adjustment module is used to adjust the valve opening according to the switch control signal;

[0046] The alarm module is used to trigger an audible and visual alarm based on the warning level.

[0047] The data recording module is used to record abnormal events based on action prompt information.

[0048] The second objective of this invention can be achieved by adopting the following technical solution:

[0049] A multi-parameter intelligent automated inspection device for hazardous gas pipeline networks, the device comprising:

[0050] The building unit is used to divide the pipeline network coverage area into several inspection sections, configure edge computing nodes for real-time data processing, and build a hierarchical inspection database.

[0051] The inspection path adjustment unit is used to fuse multiple types of parameter data based on the time-series feature extraction network to generate a comprehensive feature sequence, interactively input the multiple types of parameter data into the dynamically updated inspection model, and interactively adjust the inspection path through the dual-modal command parsing module. The multiple types of parameter data include pressure fluctuation value, concentration change, temperature gradient value and pipeline deformation degree.

[0052] The judgment unit is used to compare the comprehensive feature sequence with the baseline operating data through a multi-level discrimination mechanism to determine the degree of abnormality in pipeline operation, transmit control signals to the execution components through the control channel, and generate an inspection report.

[0053] The third objective of this invention can be achieved by adopting the following technical solution:

[0054] A computer device includes a processor and a memory for storing processor-executable programs. When the processor executes the program stored in the memory, it implements the above-described intelligent automated inspection method for multi-parameter hazardous gas pipeline networks.

[0055] The fourth objective of this invention can be achieved by adopting the following technical solution:

[0056] A computer-readable storage medium storing a program, which, when executed by a processor, implements the above-described intelligent automated inspection method for multi-parameter hazardous gas pipeline networks.

[0057] The present invention has the following advantages over the prior art:

[0058] 1. This invention significantly improves the real-time performance and reliability of data processing by constructing a hierarchical inspection database and an edge computing node collaborative architecture. It divides the pipeline network into inspection sections and assigns edge nodes, reducing the computational burden on centralized processing. Simultaneously, the global time stamp synchronization mechanism ensures time consistency of data across multiple regions, resolving data latency and conflict issues in traditional centralized management. The region segmentation algorithm dynamically divides key monitoring areas and sub-regions based on the inspection terminal trajectory, and uses distance weights to set data processing priorities, concentrating resources on high-risk areas and improving inspection efficiency and targeting.

[0059] 2. This invention employs a multi-parameter fusion technology based on a time-series feature extraction network to achieve deep integration of multiple types of data, including pressure, concentration, temperature, and deformation. Through time alignment, weight allocation, and a memory-filter composite unit design, it can capture long-term operating trends while sensitively responding to short-term fluctuations, overcoming the limitations of traditional single-parameter monitoring. The dual-modal command parsing module fuses sensor signals and manual commands, filters invalid inputs through a confidence scoring mechanism, and generates a comprehensive command through weighted fusion when the two conflict, ensuring the accuracy and flexibility of path adjustment and solving the problem of multi-source command coordination.

[0060] 3. This invention combines a multi-level discrimination mechanism with control channels to achieve accurate judgment of anomaly severity and graded response. Multi-channel control signals, including threshold triggering, trend warning, and pattern matching, can drive coordinated actions of valve adjustment, alarm modules, and data recording based on the anomaly level, forming a closed-loop control process and improving the timeliness and effectiveness of emergency response. The construction of a hierarchical database not only supports historical data tracing but also provides a data foundation for the dynamic updating of the inspection model. Continuous optimization of the discrimination criteria through long-term data accumulation enhances the method's adaptability and evolutionary capabilities.

[0061] 4. This invention realizes the transformation of hazardous gas pipeline networks from passive inspection to proactive early warning, from single parameter to multi-dimensional analysis, and from static path to dynamic optimization, which significantly reduces the accident rate and improves the safety and intelligent management level of pipeline network operation. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0063] Figure 1 This is a flowchart of the multi-parameter intelligent automated inspection method for hazardous gas pipelines according to Embodiment 1 of the present invention.

[0064] Figure 2 This is a flowchart illustrating the construction of the hierarchical inspection database in Embodiment 1 of the present invention.

[0065] Figure 3 This is a flowchart of the region segmentation algorithm in Embodiment 1 of the present invention.

[0066] Figure 4 This is a structural block diagram of the temporal feature extraction network in Embodiment 1 of the present invention.

[0067] Figure 5This is a flowchart of the temporal feature extraction network fusing multiple types of parameter data in Embodiment 1 of the present invention.

[0068] Figure 6 This is a flowchart for determining the degree of abnormality in pipeline operation according to Embodiment 1 of the present invention.

[0069] Figure 7 This is a structural block diagram of the multi-parameter intelligent automated inspection device for hazardous gas pipelines according to Embodiment 2 of the present invention.

[0070] Figure 8 This is a structural block diagram of the computer device according to Embodiment 3 of the present invention. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0072] Example 1:

[0073] like Figure 1 As shown in the figure, this embodiment provides a multi-parameter intelligent automated inspection method for hazardous gas pipeline networks, which includes the following steps:

[0074] S101. Divide the pipeline network coverage area into several inspection sections, configure edge computing nodes for real-time data processing, and build a hierarchical inspection database.

[0075] Furthermore, step S101 is as follows: Figure 2 As shown, it specifically includes:

[0076] S1011. Set the pipeline network structure complexity index, traverse each pipe segment within the pipeline network coverage area, calculate its complexity index, and record the complexity of each segment. At the same time, identify dynamically changing nodes in the pipeline network and use a region segmentation algorithm to dynamically divide the pipeline network into several inspection sections.

[0077] In constructing the hierarchical inspection database, this embodiment requires setting a pipeline network structure complexity index. The setting of this complexity index needs to comprehensively consider factors such as the pipeline network topology, the connection method of pipe segments, pipe diameter, material, and geographical environment. For example, the more branches and connections a pipe segment has in the pipeline network, the higher its complexity index will be. Pipe segments with smaller diameters may be more prone to blockages, resulting in a higher complexity index. Pipe segments located in complex terrain or densely populated areas also need to have a higher complexity index.

[0078] After setting the pipeline network structure complexity index, each pipe segment within the pipeline network coverage area is traversed. Following the established complexity index calculation method, the complexity index of each pipe segment is calculated one by one, and the complexity of each segment is recorded in detail. The pipeline network structure complexity index (C) is quantitatively calculated through a weighted summation of multiple influencing factors, as shown in the following formula:

[0079]

[0080] in, ( These are the weighting coefficients for topology, pipe diameter, material, geographical environment, and operating time, respectively, which need to be preset in conjunction with the pipeline network risk level; T is the topology factor, D is the pipe diameter factor, M is the material factor, E is the geographical environment factor, and C is the topology factor. t As a runtime factor, each factor has been normalized and its value range is [0,1] to ensure that the final quantification result of the complexity index C falls within the [0,1] interval, so as to achieve horizontal comparability of the complexity of different pipe sections.

[0081] During the recording process, it is necessary to accurately label the geographical location, pipe section number, material, pipe diameter, and other information of each section to facilitate subsequent data query and analysis.

[0082] At the same time, it is necessary to identify dynamically changing nodes in the pipeline network. These dynamically changing nodes may include nodes whose parameters such as pressure and concentration change due to changes in gas flow, or nodes that become abnormal due to external environmental influences. For example, when the gas consumption in a certain area suddenly increases, the corresponding pipeline network node in that area may become a dynamically changing node; or when the pipeline is subjected to external impact or corrosion, the corresponding node will also change dynamically.

[0083] After identifying dynamically changing nodes, a region segmentation algorithm is used to dynamically divide the pipeline network into several inspection sections. The specific operation of the region segmentation algorithm is as follows: Figure 3 As shown below:

[0084] S10111. Track the movement trajectory of the inspection terminal in real time through positioning equipment.

[0085] In this embodiment, the movement trajectory of the inspection terminal is tracked in real time using a GPS positioning device or other high-precision positioning equipment. The movement trajectory includes the location coordinates and direction of travel of the inspection terminal. The positioning device needs to record the position and direction of the inspection terminal at regular intervals to ensure the accuracy and real-time performance of the trajectory.

[0086] S10112. Determine the current key monitoring area based on the movement trajectory of the inspection terminal.

[0087] Generally, areas where inspection terminals spend a longer time or move at a slower speed may have potential safety hazards and should be identified as key monitoring areas. For example, when an inspection terminal stops near a certain pipe section and conducts a detailed inspection, the area where that pipe section is located can be considered a current key monitoring area.

[0088] S10113. Based on the current key monitoring area, expand the pipeline network nodes outward within the safe distance range and recursively divide the pipeline network into sub-regions.

[0089] In this embodiment, the safety distance needs to be determined based on factors such as the nature of the hazardous gas and the pressure rating of the pipeline. For example, for pipelines with high pressure and high concentration of hazardous gases, the safety distance should be set larger; while for pipelines with low pressure and low concentration, the safety distance can be appropriately reduced.

[0090] In this embodiment, during the recursive partitioning process, it is necessary to ensure that the pipeline nodes in each sub-region have similar complexity and dynamic change characteristics in order to carry out targeted inspections and data processing. When partitioning, the topology of the pipeline network can be used to start from the key monitoring area and gradually expand outward to divide adjacent pipeline nodes into the same sub-region.

[0091] S10114. Based on the distance weight between the sub-region and the baseline as the data processing priority indicator, sub-regions with the same distance weight are processed synchronously through time tags, and the sub-regions are dynamically updated in real time according to the changes in the movement trajectory.

[0092] In this embodiment, the distance between a sub-region and the baseline (i.e., the key monitoring area) is used as the data processing priority indicator. The closer a sub-region is to the baseline, the higher its data processing priority, because these areas are more likely to have safety hazards and require timely data processing and analysis. For example, sub-regions within 10 meters of the key monitoring area have the highest data processing priority, those within 10-20 meters have the second highest priority, and so on.

[0093] Sub-regions with the same distance weight are synchronized using time tags. The time tags are generated based on the data acquisition frequency, which can be set according to the importance and hazard level of the pipeline network. For example, for important high-pressure pipelines, the data acquisition frequency can be set to once per minute; while for general low-pressure pipelines, the data acquisition frequency can be set to once every 10 minutes. Each sub-region performs data processing operations simultaneously upon receiving the time tag to ensure consistency and synchronization in data processing.

[0094] The sub-regions are dynamically updated in real time based on changes in the movement trajectory. When the movement trajectory of the inspection terminal changes, the key monitoring area may change accordingly. Therefore, it is necessary to redetermine the key monitoring area and re-divide and update the sub-regions according to the above steps to ensure that the division of inspection sections always matches the actual inspection needs.

[0095] S1012. Assign each inspection section to different edge computing nodes, generate a new global time tag based on the data acquisition frequency, and send the global time tag to each computing node so that each computing node can perform data processing operations based on the assigned inspection section and the current time tag.

[0096] In this embodiment, after dividing the inspection sections, each inspection section is assigned to a different edge computing node. The allocation of edge computing nodes needs to consider factors such as the node's computing power, storage capacity, and physical distance from the inspection section to ensure the efficiency and reliability of data processing. For example, inspection sections with higher complexity are assigned to edge computing nodes with stronger computing power to ensure that data can be processed in a timely manner.

[0097] A new global time stamp is generated based on the data acquisition frequency and sent to each computing node. This global time stamp unifies the time base across all computing nodes, ensuring that data processing on each node remains consistent in time. Each node performs data processing operations based on its assigned inspection section and the current time stamp. These operations include preprocessing the collected pressure fluctuations, concentration changes, temperature gradients, and pipeline deformation data, such as filtering, denoising, and normalization, to improve data quality and usability.

[0098] S1013. Integrate the processing results of all inspection sections into the main database to generate a complete inspection data set, resulting in a hierarchical inspection database.

[0099] The main database in this embodiment needs to have powerful data storage and management capabilities, capable of storing a large amount of inspection data and supporting fast data query and retrieval. During the integration process, the data of each section needs to be formatted and standardized to ensure data consistency and compatibility, thereby generating a complete inspection data set, which yields a hierarchical inspection database. This inspection data set contains detailed parameter data and processing results of each section of the pipeline network, providing comprehensive data support for subsequent inspection analysis and anomaly determination.

[0100] S102. Based on the temporal feature extraction network, multiple types of parameter data are fused to generate a comprehensive feature sequence. The multiple types of parameter data are interactively input into the dynamically updated inspection model, and the inspection path is interactively adjusted through the dual-modal command parsing module.

[0101] like Figure 4 As shown, the time-series feature extraction network in this embodiment is a composite network structure consisting of a memory unit for capturing long-term trends and a filtering unit for processing short-term fluctuations. The memory unit can adopt a time-series neural network structure such as Long Short-Term Memory (LSTM) or Gated Recurrent Unit (GRU). Its core function is to store and remember long-term trend information in historical data. For example, for time-series data of pressure fluctuation values, the memory unit can learn the overall trend of pressure changes over a period of time, such as whether there is a continuous upward or downward trend. The filtering unit can be a low-pass filter or other algorithm module that can smooth short-term fluctuations. Its function is to remove high-frequency noise and short-term abnormal fluctuations from the data and retain the basic trend of the data. For example, when the concentration sensor experiences a brief numerical change due to external interference, the filtering unit can smooth the change to avoid misjudging it as a real abnormal change in concentration.

[0102] When fusing multiple types of parameter data based on time-series feature extraction networks, it is necessary to collect four types of parameter data: pressure fluctuation, concentration change, temperature gradient, and pipeline deformation. Pressure fluctuation data collection requires pressure sensors installed at key nodes in the pipeline network, such as bends and near valves, where pressure fluctuations are likely to occur. The sensors acquire pressure data in real time at a set sampling frequency, recording pressure changes over time. Concentration change data collection relies on gas concentration sensors. Based on the characteristics of hazardous gases, appropriate sensor types are selected and installed in areas where gas leaks or significant concentration changes may occur, such as pipe joints and flange connections, to continuously monitor dynamic changes in gas concentration. Temperature gradient data collection requires deploying a temperature sensor array, rationally distributed along the axial and radial directions of the pipeline to capture temperature differences at different locations and calculate the temperature gradient value. Pipeline deformation data collection can utilize laser rangefinders or strain gauges, installed at pipeline support points, bends, and other easily deformable locations to measure pipeline deformation in real time.

[0103] In actual inspection operations, when using a multi-level discrimination mechanism combined with control channels to achieve graded response to anomalies, the system first aligns the real-time generated comprehensive feature sequence with the baseline operating data (including baseline pressure, baseline concentration, baseline temperature, and baseline deformation) of each inspection section stored in the hierarchical inspection database in the time dimension. The anomaly level is then determined by calculating the weighted deviation. If a minor anomaly is determined, the threshold-triggered control channel is triggered, generating a switch signal that is transmitted to the valve adjustment module to fine-tune the valve opening to balance the pipeline pressure. At the same time, the data recording module automatically records the adjustment time, valve number, and parameter changes before and after adjustment. If the anomaly is determined to be moderate, the trend warning control channel is activated, displaying a yellow warning level on the monitoring interface. The alarm module emits an intermittent buzzer, simultaneously triggering the inspection terminal to adjust its path and prioritizing secondary data collection on the abnormal section. If the anomaly is determined to be severe, the mode matching control channel is activated, generating action commands based on the preset anomaly handling mode. The alarm module emits a continuous audible and visual alarm, the valve regulating module quickly closes the upstream and downstream valves of the abnormal section, and the data recording module fully records the time, location, peak parameter deviation, and handling details of the anomaly, forming a closed-loop control process of "anomaly determination - signal transmission - execution response - data recording". Regarding the hierarchical database supporting the dynamic updating of the inspection model, the database regularly integrates historical inspection data from each inspection section, including parameter fluctuation patterns at different times, records of anomaly handling effects, and data on the impact of environmental factors. This data is used as the input source for the inspection model to optimize the parameter weight allocation and anomaly detection threshold in the model. For example, when a certain area of ​​the pipeline network is in a high-temperature environment for a long time, causing the temperature gradient value to frequently approach the warning threshold without triggering an actual fault, the inspection model can adjust the weight of the temperature gradient parameter in the weighted deviation calculation based on the historical data accumulated in the database. This makes the anomaly detection criteria more consistent with the actual operating conditions of the area, gradually improving the adaptability of the method to different operating scenarios.

[0104] Furthermore, the process of fusing multiple types of parameter data based on a temporal feature extraction network in this embodiment is as follows: Figure 5 As shown, it specifically includes:

[0105] S1021. Align different parameter data in the time dimension using a hardware clock.

[0106] In this embodiment, after collecting various parameter data, different parameter data are aligned in the time dimension through a hardware clock. The hardware clock provides a unified time reference for each sensor and data acquisition device, ensuring that parameter data of different types and locations have consistency in timestamps. For example, data collected by pressure sensors and concentration sensors at the same time will be tagged with the same timestamp for subsequent time series analysis.

[0107] S1022. Assign different weights based on the sensitivity and correlation of the parameter data.

[0108] In this embodiment, the sensitivity of parameter data refers to the degree to which changes in the parameter affect the safe operation of the pipeline network. For example, changes in concentration may directly reflect gas leaks and are highly sensitive; while changes in temperature gradient values ​​may, to some extent, indicate pipeline anomalies, but their sensitivity is relatively low. Correlation refers to the degree of mutual influence between different parameters. For instance, pressure fluctuations and concentration changes may change simultaneously during a gas leak, showing a strong correlation. Based on these characteristics, a corresponding weighting coefficient is assigned to each parameter. The weighting coefficient ranges from 0 to 1, with a larger value indicating greater importance for the parameter.

[0109] S1023. Input the multi-class parameter data into the memory unit and the filtering unit respectively, extract the features of the multi-class parameter data, and perform weighted fusion according to the weights to generate a comprehensive feature sequence.

[0110] In this embodiment, for each type of parameter data, such as pressure fluctuation value, its time series data is first input into the memory unit. The memory unit extracts the long-term trend characteristics of the parameter by learning historical data. At the same time, the time series data of the same pressure fluctuation value is input into the filtering unit. The filtering unit processes the data and extracts the short-term fluctuation characteristics.

[0111] After extracting features from multiple parameter data, a weighted fusion is performed based on previously assigned weights. Specifically, for each time point, the long-term trend features extracted by the memory unit and the short-term fluctuation features extracted by the filtering unit are weighted and summed according to their corresponding parameter weights to obtain the comprehensive feature of that parameter at that time point. Then, the comprehensive features of the four types of parameters (pressure fluctuation value, concentration change, temperature gradient value, and pipe deformation degree) are fused to generate a comprehensive feature sequence. For example, at a certain time point, the comprehensive feature weight of the pressure fluctuation value is 0.3, the comprehensive feature weight of the concentration change value is 0.4, the comprehensive feature weight of the temperature gradient value is 0.2, and the comprehensive feature weight of the pipe deformation degree is 0.1. These four features are added together according to their respective weights to obtain the comprehensive feature at that time point. This process is repeated to form the comprehensive feature sequence of the entire time series.

[0112] When adjusting the inspection path interactively through the dual-modal command parsing module, the dual-modal commands include sensor commands and manual commands. For sensor commands, a combination of feature matching and threshold judgment is used for command parsing. First, parameter change features are extracted. From the collected parameter data such as pressure fluctuation values ​​and concentration changes, the change patterns, trends, and amplitudes of the parameters are analyzed. For example, when the concentration change rises sharply in a short period of time and exceeds a certain preset threshold, this sharp rise feature is extracted. Then, the time-series dependency is determined, that is, whether the change of the current parameter has a time dependency relationship with the change of historical parameters. For example, whether the previous abnormal pressure fluctuation value led to the current abnormal concentration change. Through this time-series dependency analysis, the cause and nature of the parameter change can be determined more accurately.

[0113] Based on the results of feature matching and threshold judgment, a path adjustment command is output. Simultaneously, a data confidence score is set to determine the validity of sensor input. The calculation of the data confidence score considers multiple factors, such as sensor accuracy, data consistency, and sampling frequency. For example, data collected by high-precision sensors has a relatively high confidence score; and when data collected by multiple sensors is consistent, the confidence score also increases. If the data confidence score is lower than the preset confidence standard, an invalid signal is output, indicating that the sensor command is unreliable and cannot be used as a basis for adjusting the inspection path. If the confidence score is not lower than the confidence standard, an adjustment signal is output, indicating that the inspection path needs to be adjusted.

[0114] For human commands, semantic analysis is performed based on a natural language understanding model. Human commands can be input in the form of text or speech. First, the input text or speech is converted into a machine-processable format. For speech commands, speech recognition is required to convert speech into text. For text commands, the next step is performed directly. Then, through semantic extraction and intent recognition, key information is extracted from the text to understand the operator's intent. For example, if the input human command is "go to pipeline A area to check the concentration", semantic extraction will identify key information such as "pipeline A area" and "concentration", and intent recognition will determine that the operator wants the inspection terminal to go to area A to check the concentration.

[0115] Based on the semantic analysis results, a path adjustment instruction is output, and an instruction confidence score is set to judge the validity of the manual instruction. The determination of the instruction confidence score needs to consider factors such as the clarity, completeness, and grammatical correctness of the instruction. For example, clear, complete, and grammatically correct instructions have a higher confidence score; while vague, incomplete, or grammatically incorrect instructions have a lower confidence score. If the instruction confidence score is lower than the confidence standard, an invalid signal is output, indicating that the manual instruction cannot be correctly understood or executed; if the confidence score is not lower than the confidence standard, an adjustment signal is output, instructing the inspection path to be adjusted according to the manual instruction.

[0116] When adjusting the inspection path interactively through the dual-modal command parsing module, there are three different working modes, each used to handle different types of command inputs and interactive adjustment logic, specifically including:

[0117] Attitude 1: Path adjustment triggered by sensor parameter sequence.

[0118] Upon receiving the collected parameter data (pressure fluctuation values, concentration changes, temperature gradient values, and pipe deformation), the sensor parameter data is integrated into a sensor parameter sequence in chronological order. For example, within a certain time period, the pressure sensor collects pressure values ​​of 1.2MPa, 1.25MPa, and 1.3MPa at a certain node in the pipeline, the corresponding gas concentration values ​​collected by the concentration sensor are 20ppm, 22ppm, and 25ppm, the temperature sensor collects temperature values ​​of 25℃, 26℃, and 27℃, and the pipe deformation sensor collects deformation values ​​of 0.1mm, 0.12mm, and 0.15mm. These data are arranged in chronological order to form a sensor parameter sequence.

[0119] The sensor parameter sequence is analyzed to extract features such as the pattern, magnitude, and rate of parameter changes. For example, when the concentration changes from 20 ppm to 25 ppm within 10 minutes, and the pressure fluctuation value also shows a continuous upward trend, the system will identify this combination of features. Then, the sensor will trigger a path adjustment command output. The generation of this command combines the results of feature matching and threshold judgment. For example, if the concentration change exceeds the preset warning threshold of 25 ppm and the pressure fluctuation value exceeds the threshold of 1.3 MPa, it will be determined that the inspection path needs to be adjusted, increasing the inspection frequency of the area or changing the inspection route to prioritize the inspection of that area.

[0120] Subsequently, the inspection terminal performs interactive adjustment of the inspection path based on the output adjustment signal. Specifically, the inspection terminal receives path adjustment instructions, such as turning from the current inspection route to an area with abnormal concentration and pressure, or inserting key inspection steps for the abnormal area into the original route. During the execution process, the position and actions of the inspection terminal are monitored in real time to ensure the accuracy and timeliness of the path adjustment.

[0121] Posture 2: Path adjustment based on human command data triggering.

[0122] When receiving real-time human instructions, whether in text or voice format, semantic processing is performed. For voice instructions, the system converts them into text content through the speech recognition module; for text instructions, the system directly enters the semantic extraction stage. For example, if an operator inputs the voice instruction "go to the bend of pipe No. 3 to check the temperature", the speech recognition module converts it into text, and the system extracts key semantic features such as "bend of pipe No. 3" and "check the temperature".

[0123] Through semantic analysis, a path adjustment instruction is triggered to generate a manual command. This instruction is generated based on an accurate understanding of the operator's intent. For example, the intent of the instruction mentioned above might be to direct the inspection terminal to a designated location for a specific temperature check. The system will calculate the optimal path to the bend in pipe No. 3 based on the pipeline topology and the current location of the inspection terminal, and generate the corresponding path adjustment instruction.

[0124] Subsequently, the system executes an interactive adjustment of the inspection path based on the path adjustment command. After receiving the path adjustment command, the inspection terminal moves to the No. 3 pipe bend along the planned optimal path and activates the temperature sensor for specialized detection. During the movement, the system updates the location information of the inspection terminal in real time and adjusts the path according to the actual situation to avoid affecting the inspection progress due to abnormal conditions or physical obstacles in other areas of the pipeline network.

[0125] Attitude 3: Invalid signal situations that occur when processing sensor commands and manual commands interact to adjust the path.

[0126] When invalid signals are detected in the interaction between sensor commands and human commands to adjust the path, the number of invalid signals will be counted first. For example, if the sensor commands output invalid signals 3 times because the data confidence score is lower than the standard, and the human commands output invalid signals 2 times because the command confidence score is lower than the standard, the number of invalid signals will be 5.

[0127] If the number of invalid signals exceeds the invalid threshold (preset, such as 5 times), an instruction-parameter association table is established. This table records the correspondence between sensor instructions, manual instructions, and parameter changes. For example, the pressure fluctuation value, concentration change, and other parameter data corresponding to a certain sensor instruction, as well as the operation intention and parameter focus corresponding to a certain manual instruction.

[0128] The confidence scores of sensor and human commands are weighted and fused. This weighting process considers the importance and reliability of different command types, assigning different weight coefficients to the confidence scores of sensor and human commands. For example, the confidence score of a sensor command has a weight of 0.6, and the confidence score of a human command has a weight of 0.4. The overall confidence score is then calculated through weighted calculation.

[0129] The system outputs the matching degree between bimodal commands and parameter changes, and sets a comprehensive confidence level. The matching degree is calculated based on the information recorded in the command-parameter association table, analyzing the degree of fit between the bimodal commands and actual parameter changes. For example, a bimodal command instructs the inspection terminal to check the concentration in pipeline area A. If the concentration sensor in pipeline area A does indeed collect abnormal concentration change data, it indicates that the matching degree between the command and the parameter changes is high.

[0130] If the overall confidence score exceeds the total standard, such as 60 points, the instruction is considered valid, triggering the path adjustment operation. The inspection terminal will adjust the inspection path according to the valid instruction and execute the corresponding inspection task. Otherwise, a prompt signal will be issued to remind the operator to pay attention to the validity of the instruction or abnormalities in the parameter data, so that the operator can take appropriate measures in a timely manner.

[0131] During the interactive adjustment of the three postures, the operation data of each step will be recorded in real time, including the type and content of the instruction, the confidence score, the adjusted path information, and the execution status of the inspection terminal. This data will be used for subsequent inspection analysis and system optimization to ensure that the adjustment of the inspection path is always based on accurate instructions and reliable data, thereby improving the efficiency and safety of hazardous gas pipeline inspection.

[0132] S103. By comparing the comprehensive feature sequence with the baseline operating data through a multi-level discrimination mechanism, the degree of abnormality in pipeline operation is determined, and the control signal is transmitted to the execution component through the control channel to generate an inspection report.

[0133] Furthermore, this step S103 is as follows: Figure 6 As shown, it specifically includes:

[0134] S1031. Construct baseline operating data as the inspection target, and align the baseline operating data with the real-time operating data of the pipeline network in the time dimension.

[0135] In this embodiment, when determining the degree of abnormality in pipeline network operation, it is necessary to construct benchmark operating data as the inspection target. Benchmark operating data includes benchmark pressure, benchmark concentration, benchmark temperature, and benchmark deformation. These data are determined based on historical monitoring data under normal pipeline network operation conditions. For example, under normal operating conditions, the pressure of a certain section of natural gas pipeline is usually stable at around 2.0 MPa, so the benchmark pressure value is set to 2.0 MPa; the concentration of natural gas in the pipeline is normally 98%, so the benchmark concentration value is set to 98%; the temperature of the pipeline environment is generally maintained between 20℃ and 25℃, so the benchmark temperature value is set to 22℃ after comprehensive consideration; the deformation of the pipeline during normal operation is extremely small, and long-term monitoring shows that its deformation usually does not exceed 0.05 mm, so the benchmark deformation value is set to 0.05 mm.

[0136] After establishing the baseline operating data, it is necessary to align it with the real-time operating data of the pipeline network in terms of time. For example, if the pressure value at a certain moment in the real-time operating data is 2.1 MPa, the concentration value is 95%, the temperature value is 28℃, and the deformation is 0.08 mm, then these real-time data need to be matched with the baseline operating data at the corresponding moment to ensure that the comparison time points are consistent in order to accurately calculate the degree of deviation.

[0137] S1032. For each time point, calculate the degree of deviation between the comprehensive feature sequence and the baseline running data, assign weights to each type of parameter according to the importance of the parameters, and generate a weighted deviation degree.

[0138] In this embodiment, for each time point, the deviation between the comprehensive characteristic sequence and the baseline operating data is calculated. Taking a certain time point as an example, the pressure fluctuation characteristic value in the comprehensive characteristic sequence is 2.1 MPa, the baseline pressure value is 2.0 MPa, and the deviation between the two is 0.1 MPa; the concentration change characteristic value is 95%, the baseline concentration value is 98%, and the deviation is -3%; the temperature gradient characteristic value is 28℃, the baseline temperature value is 22℃, and the deviation is 6℃; the pipeline deformation characteristic value is 0.08 mm, the baseline deformation is 0.05 mm, and the deviation is 0.03 mm.

[0139] Each parameter is assigned a weight based on its importance. Different parameters have varying degrees of impact on the safe operation of the pipeline network. For example, concentration changes directly relate to the risk of gas leakage and are therefore of high importance, so they are assigned a weight of 0.4. Pressure fluctuations affect the pressure-bearing safety of the pipeline and are of secondary importance, so they are assigned a weight of 0.3. Temperature gradients may affect the physical properties of the gas and the material properties of the pipeline, so they are assigned a weight of 0.2. Pipeline deformation reflects the structural safety of the pipeline and is assigned a weight of 0.1.

[0140] A weighted deviation is generated based on the deviations and weights of each parameter. The calculation method is: pressure deviation × pressure weight + concentration deviation × concentration weight + temperature deviation × temperature weight + deformation deviation × deformation weight. Continuing with the example above, the weighted deviation is 0.1 × 0.3 + (-3%) × 0.4 + 6 × 0.2 + 0.03 × 0.1. It is important to note that in actual calculations, the deviations of each parameter need to be standardized or converted to a unified unit to ensure the rationality of the weighted calculation.

[0141] S1033. Use a standardized function to map the weighted deviation level to the intensity range of the control signal, and generate corresponding control signals through the control channel, and transmit the control signals to the execution unit.

[0142] The control channel in this embodiment includes threshold triggering, trend warning, and pattern matching. When transmitting the control signal, a normalization function is used to map the weighted deviation degree to the intensity range of the control signal. For example, the intensity range of the control signal is set to 0 to 100. The normalization function converts the calculated weighted deviation degree into a value within this range. The larger the value, the greater the deviation degree and the higher the required control signal strength.

[0143] Furthermore, corresponding control signals are generated through the control channel and transmitted to the execution unit, specifically including:

[0144] If the control channel is threshold-triggered, the degree of deviation is reflected through a switching signal, and a switching control signal is transmitted to the actuator. For example, when the weighted deviation degree is converted into a control signal strength value of 60, the threshold switch is triggered, and a switching control signal is output to the actuator. Assuming the actuator is a valve regulating module, this signal may instruct the closure of part of the valve to regulate the pipeline pressure and reduce the degree of deviation.

[0145] If the control channel is a trend warning system, the degree of deviation is displayed through the warning level, and the warning level is shown on the monitoring interface. The warning level is then transmitted to the actuator. Warning levels are typically divided into different levels, such as low, medium, high, and emergency. When the control signal strength value is between 0 and 30, a low warning level is displayed, and the warning indicator on the monitoring interface is green; when the strength value is between 31 and 60, a medium warning level is displayed, and the indicator is yellow; when the strength value is between 61 and 80, a high warning level is displayed, and the indicator is orange; when the strength value is between 81 and 100, an emergency warning level is displayed, and the indicator is red.

[0146] If the control channel is pattern matching, the degree of deviation is indicated by an action command, and the action command information is transmitted to the execution unit. For example, when the pattern matching algorithm identifies that the weighted deviation degree corresponds to the pattern of pipeline leakage, the action command "The pipeline may be leaking, please go and check immediately" is generated, and the prompt information is played through the voice broadcast device in the execution unit. At the same time, the action command is sent to the data recording module for recording.

[0147] S1034. Based on the control signal, generate an inspection report using a multi-level discrimination mechanism.

[0148] The multi-level discrimination mechanism in this embodiment includes a primary discrimination that calculates deviations in real time and provides adjustment, an intermediate discrimination that periodically evaluates inspection progress and effectiveness, and a high-level discrimination that comprehensively analyzes long-term inspection data and provides optimization suggestions. The details are as follows:

[0149] Primary discrimination: Real-time calculation of deviation and provision of control. For example, during the inspection process, real-time data of each parameter is received, the weighted deviation is calculated, and control signals are generated in real time according to the deviation to control the execution unit to perform corresponding operations, such as adjusting the valve opening or triggering an alarm.

[0150] Intermediate Assessment: Periodically evaluate the progress and effectiveness of inspections. Using a daily cycle, evaluate the inspection data for each day to see if the planned inspection tasks were completed, whether parameter deviations in each area have improved, and the effectiveness of control measures. For example, check whether all designated inspection sections were inspected that day, whether the parameter deviations in each section were within acceptable limits, and whether valve adjustments, alarms, and other operations were timely and effective.

[0151] Advanced Judgment: Comprehensive analysis of long-term inspection data to provide optimization suggestions. Analysis of inspection data over a month or longer identifies patterns and potential problems in pipeline operation. For example, analysis may reveal that a certain section of pipeline frequently experiences significant pressure deviations at specific times each month, possibly due to regular variations in gas consumption during those periods. Based on this, suggestions for optimizing scheduling can be proposed to reduce pressure deviations.

[0152] Throughout the process, data from each stage needs to be accurately recorded and stored for subsequent retrieval and analysis. For example, the process of constructing baseline operating data, the acquisition time and values ​​of real-time operating data, the calculation process of deviation, the generation and transmission records of control signals, and the content of inspection reports all need to be fully recorded in the database to provide detailed data support for the safe operation and maintenance of the pipeline network.

[0153] The execution components in this embodiment include a valve adjustment module, an alarm module, and a data recording module. These modules function through the control signals transmitted via the control channel. Taking a section of a hazardous gas pipeline that transports carbon monoxide as an example, when the pressure fluctuation value in the pipeline exceeds the reference pressure value of 0.5 MPa and the concentration change reaches 120% of the reference concentration value, a control signal is generated and transmitted to the execution components.

[0154] The valve regulating module adjusts the valve opening based on on / off control signals. For example, when the control signal indicates that pressure needs to be reduced due to excessive pressure, the valve regulating module receives the corresponding on / off control signal and slowly adjusts the opening of the regulating valve on the pipeline from 60% to 40% to reduce the gas flow rate and thus alleviate the pressure in the pipeline. During the adjustment process, the drive motor inside the valve regulating module operates according to the signal command, driving the valve core to move through the gear transmission mechanism. At the same time, the position sensor provides real-time feedback on the valve opening status to ensure accurate execution of the adjustment action.

[0155] The alarm module triggers audible and visual alarms based on the warning level. For example, if the current anomaly is determined to be a high-level warning based on the weighted deviation, the alarm module will immediately activate the audible and visual alarms. Specifically, the alarm light installed on-site will begin flashing red, while a buzzer will emit a rapid alarm sound to alert personnel to potential danger. The audible and visual signal strength of the alarm module must be clearly visible and audible within a certain range. For example, at a distance of 10 meters from the alarm module, the light should flash twice per second, and the sound intensity should reach 80 decibels, so that staff can detect the danger promptly.

[0156] The data logging module records abnormal events based on action prompts. When the valve regulating module performs a regulating action or the alarm module triggers an alarm, the data logging module receives the corresponding action prompts. For example, when the valve regulating module adjusts the opening from 60% to 40%, the data logging module records the time of the regulating action as 10:15 AM on June 30, 2025, the valve number as V-001, the opening before adjustment as 60%, and the opening after adjustment as 40%. When the alarm module triggers a high-level warning, the data logging module records the alarm time as 10:12 AM on June 30, 2025, the alarm location as Pipeline 1 in Area A of the pipeline network, and the warning level as high-level.

[0157] In another scenario, when the pipeline's deformation exceeds a baseline value due to an abnormal temperature rise, a control signal is generated. The valve adjustment module may receive this signal to partially close the valve to reduce gas flow. The alarm module triggers a yellow warning light and an intermittent buzzer alarm. The data recording module simultaneously records the time, location, deformation, and specific valve adjustment actions of the temperature anomaly. After the valve adjustment is completed, a confirmation signal is sent to the alarm module. Once the alarm module confirms that the temperature has begun to decrease, it stops the alarm or lowers the alarm level.

[0158] Throughout the entire operation of the execution unit, the modules transmit data and exchange commands via standardized communication protocols, ensuring the stability and timeliness of signal transmission. For example, the valve regulating module and the data recording module use the Modbus communication protocol, while the alarm module and the data recording module use the CAN bus communication protocol. These communication methods can adapt to the complex environment of industrial sites, reduce signal interference, and ensure smooth collaborative operation. Simultaneously, the data recording module stores all recorded information in a local database and periodically uploads it to the main database, allowing managers to query and analyze the execution unit's operational status at any time, providing a basis for pipeline network maintenance and optimization.

[0159] It should be noted that although the above-described method operations are depicted in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all of the illustrated operations must be performed to achieve the desired result. On the contrary, the order of execution of the depicted steps can be changed. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0160] Example 2:

[0161] like Figure 7 As shown in the figure, this embodiment provides a multi-parameter intelligent automated inspection device for hazardous gas pipeline networks. The device includes a construction unit 701, an inspection path adjustment unit 702, and a judgment unit 703. The specific functions of each module are as follows:

[0162] The construction unit 701 is used to divide the pipeline network coverage area into several inspection sections, configure edge computing nodes for real-time data processing, and build a hierarchical inspection database.

[0163] The inspection path adjustment unit 702 is used to fuse multiple types of parameter data based on the time-series feature extraction network to generate a comprehensive feature sequence, interactively input the multiple types of parameter data into the dynamically updated inspection model, and interactively adjust the inspection path through the dual-modal command parsing module. The multiple types of parameter data include pressure fluctuation value, concentration change, temperature gradient value and pipeline deformation degree.

[0164] The judgment unit 703 is used to compare the comprehensive feature sequence with the baseline operating data through a multi-level discrimination mechanism to determine the degree of abnormality in pipeline operation, transmit control signals to the execution components through the control channel, and generate an inspection report.

[0165] The specific implementation of each module in this embodiment can be found in Embodiment 1 above, and will not be repeated here. It should be noted that the device provided in this embodiment is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional units as needed, that is, the internal structure can be divided into different functional units to complete all or part of the functions described above.

[0166] Example 3:

[0167] This embodiment provides a computer device, such as... Figure 8 As shown, it includes a processor 802, a memory, an input device 803, a display device 804, and a network interface 805 connected via a device bus 801. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium 806 and an internal memory 807. The non-volatile storage medium 806 stores operating devices, computer programs, and a database. The internal memory 807 provides an environment for the operation of the operating devices and computer programs in the non-volatile storage medium. When the processor 802 executes the computer program stored in the memory, it implements the multi-parameter intelligent automated inspection method for hazardous gas pipelines described in Embodiment 1 above, as follows:

[0168] The pipeline network coverage area is divided into several inspection sections, and edge computing nodes are configured in each section for real-time data processing to construct a hierarchical inspection database. Based on a time-series feature extraction network, multiple types of parameter data are fused to generate a comprehensive feature sequence. The multi-type parameter data is interactively input into the dynamically updated inspection model, and the inspection path is interactively adjusted through a dual-modal command parsing module. The multi-type parameter data includes pressure fluctuation values, concentration changes, temperature gradient values, and pipeline deformation. A multi-level discrimination mechanism is used to compare the comprehensive feature sequence with the baseline operating data to determine the degree of pipeline network operation abnormality. Control signals are transmitted to the execution components through the control channel to generate an inspection report.

[0169] Example 4:

[0170] This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the multi-parameter intelligent automated inspection method for hazardous gas pipelines described in Embodiment 1 above, as follows:

[0171] The pipeline network coverage area is divided into several inspection sections, and edge computing nodes are configured in each section for real-time data processing to construct a hierarchical inspection database. Based on a time-series feature extraction network, multiple types of parameter data are fused to generate a comprehensive feature sequence. The multi-type parameter data is interactively input into the dynamically updated inspection model, and the inspection path is interactively adjusted through a dual-modal command parsing module. The multi-type parameter data includes pressure fluctuation values, concentration changes, temperature gradient values, and pipeline deformation. A multi-level discrimination mechanism is used to compare the comprehensive feature sequence with the baseline operating data to determine the degree of pipeline network operation abnormality. Control signals are transmitted to the execution components through the control channel to generate an inspection report.

[0172] It should be noted that the computer-readable storage medium in this embodiment can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0173] In this embodiment, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this embodiment, the computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable program. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable storage medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable storage medium can be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination thereof.

[0174] The computer-readable storage medium described above can be used to write computer programs for executing this embodiment in one or more programming languages ​​or combinations thereof. These programming languages ​​include object-oriented programming languages—such as Java, Python, and C++—and conventional procedural programming languages—such as C or similar programming languages. The program can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0175] In summary, this invention realizes the transformation of hazardous gas pipeline networks from passive inspection to proactive early warning, from single parameter to multi-dimensional analysis, and from static path to dynamic optimization, significantly reducing the accident rate and improving the safety and intelligent management level of pipeline network operation.

[0176] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope disclosed in the present invention, based on the technical solution and inventive concept of the present invention, shall fall within the scope of protection of the present invention.

Claims

1. A multi-parameter intelligent automated inspection method for hazardous gas pipeline networks, characterized in that, The method includes: The pipeline network coverage area is divided into several inspection sections, and edge computing nodes are configured in each section for real-time data processing to build a hierarchical inspection database. Based on the fusion of multiple types of parameter data by the time-series feature extraction network, a comprehensive feature sequence is generated. The multiple types of parameter data are interactively input into the dynamically updated inspection model, and the inspection path is interactively adjusted through the dual-modal command parsing module. The multiple types of parameter data include pressure fluctuation value, concentration change, temperature gradient value and pipeline deformation degree. The time-series feature extraction network is a composite network structure composed of memory units for capturing long-term trends and filtering units for processing short-term fluctuations. By comparing the comprehensive feature sequence with the baseline operating data through a multi-level discrimination mechanism, the degree of pipeline network operation abnormality is determined, and control signals are transmitted to the execution components through the control channel to generate an inspection report; The method for fusing multiple types of parameter data based on the temporal feature extraction network to generate a comprehensive feature sequence specifically includes: Align different parameter data in the time dimension using a hardware clock; Different weights are assigned based on the sensitivity and correlation of the parameter data; Multiple types of parameter data are input into the memory unit and the filtering unit respectively, the features of the multiple types of parameter data are extracted, and the data are weighted and fused according to the weights to generate a comprehensive feature sequence. The process involves comparing the comprehensive feature sequence with baseline operating data using a multi-level discrimination mechanism to determine the degree of pipeline network operation anomaly, transmitting control signals to the actuators via control channels, and generating inspection reports. Specifically, this includes: A baseline operating data is constructed as the inspection target, and the baseline operating data is aligned with the real-time operating data of the pipeline network in the time dimension. The baseline operating data includes baseline pressure value, baseline concentration value, baseline temperature value, and baseline deformation. For each time point, the degree of deviation between the comprehensive feature sequence and the baseline running data is calculated. Based on the importance of the parameters, weights are assigned to each type of parameter to generate a weighted deviation degree. The weighted deviation is mapped to the intensity range of the control signal using a standardized function, and the corresponding control signals are generated through the control channel and transmitted to the execution unit. Based on the control signals, an inspection report is generated using a multi-level discrimination mechanism. This multi-level discrimination mechanism includes a primary discrimination mechanism that calculates deviations in real time and provides control, an intermediate discrimination mechanism that periodically evaluates the progress and effectiveness of inspections, and a high-level discrimination mechanism that comprehensively analyzes long-term inspection data and provides optimization suggestions.

2. The multi-parameter intelligent automated inspection method for hazardous gas pipelines according to claim 1, characterized in that, The process of dividing the pipeline network coverage area into several inspection sections, configuring edge computing nodes for real-time data processing in each section, and constructing a hierarchical inspection database specifically includes: Set a pipeline network structure complexity index, traverse each pipe segment within the pipeline network coverage area, calculate its complexity index, and record the complexity of each segment. At the same time, identify dynamically changing nodes in the pipeline network and use a region segmentation algorithm to dynamically divide the pipeline network into several inspection segments. Each inspection segment is assigned to a different edge computing node. A new global time tag is generated based on the data collection frequency, and the global time tag is sent to each computing node so that each computing node can perform data processing operations based on the assigned inspection segment and the current time tag. The processing results of all inspection sections are integrated into the main database to generate a complete set of inspection data, resulting in a hierarchical inspection database.

3. The multi-parameter intelligent automated inspection method for hazardous gas pipelines according to claim 2, characterized in that, The method of dynamically dividing the pipeline network into several inspection sections using a region segmentation algorithm specifically includes: The movement trajectory of the inspection terminal is tracked in real time through positioning devices; Determine the current key monitoring areas based on the movement trajectory; Based on the current key monitoring area, the pipeline network nodes within the safe distance range are expanded outward, and the pipeline network is recursively divided into sub-regions; The distance weight between the sub-region and the baseline is used as the data processing priority indicator. Sub-regions with the same distance weight are processed synchronously through time tags, and the sub-regions are dynamically updated in real time according to changes in the movement trajectory.

4. The multi-parameter intelligent automated inspection method for hazardous gas pipelines according to claim 1, characterized in that, The dual-modal commands include sensor commands and manual commands; The dual-modal command parsing module interactively adjusts the inspection path, specifically including: Upon receiving the collected sensor parameter data, the sensor parameter data is integrated into a sensor parameter sequence in chronological order. Based on the generated sensor parameter sequence, a sensor command output path adjustment command is triggered, and the inspection path is adjusted interactively according to the path adjustment command. Upon receiving real-time human instruction data, semantic features are extracted based on the received data, triggering a path adjustment instruction for the human instruction output. The inspection path is then adjusted interactively based on the path adjustment instruction. When an invalid signal is detected in the interaction between sensor commands and human commands to adjust the path, if the number of invalid signals is higher than the invalid threshold, a command-parameter association table is established, the confidence scores of sensor and human commands are weighted and fused, the matching degree between the dual-modal command and parameter changes is output, and a comprehensive confidence score is set. If the comprehensive confidence score exceeds the total standard, the command is considered valid and the path adjustment is triggered; otherwise, a prompt signal is issued.

5. The multi-parameter intelligent automated inspection method for hazardous gas pipelines according to claim 4, characterized in that, The process of generating corresponding control signals through the control channel and transmitting these control signals to the execution unit specifically includes: If the control channel is threshold triggered, the degree of deviation is reflected by the switch signal, and the switch control signal is transmitted to the execution unit. If the control channel is a trend warning, the degree of deviation is displayed through the warning level, and the warning level is displayed in the monitoring interface and transmitted to the execution unit. If the control channel is pattern matched, the degree of deviation will be indicated by the action command, and the action prompt information will be transmitted to the execution unit.

6. A multi-parameter intelligent automated inspection device for hazardous gas pipeline networks, characterized in that, The device includes: The database construction unit is used to divide the pipeline network coverage area into several inspection sections, configure edge computing nodes for real-time data processing, and build a hierarchical inspection database. The inspection path adjustment unit is used to fuse multiple types of parameter data based on the time-series feature extraction network to generate a comprehensive feature sequence, interactively input the multiple types of parameter data into the dynamically updated inspection model, and interactively adjust the inspection path through the dual-modal command parsing module. The multiple types of parameter data include pressure fluctuation value, concentration change, temperature gradient value and pipeline deformation degree. The time-series feature extraction network is a composite network structure composed of a memory unit for capturing long-term trends and a filtering unit for processing short-term fluctuations. The anomaly determination unit is used to compare the comprehensive feature sequence with the baseline operating data through a multi-level discrimination mechanism to determine the degree of anomaly in the pipeline network operation, transmit control signals to the execution components through the control channel, and generate an inspection report. The method for fusing multiple types of parameter data based on the temporal feature extraction network to generate a comprehensive feature sequence specifically includes: Align different parameter data in the time dimension using a hardware clock; Different weights are assigned based on the sensitivity and correlation of the parameter data; Multiple types of parameter data are input into the memory unit and the filtering unit respectively, the features of the multiple types of parameter data are extracted, and the data are weighted and fused according to the weights to generate a comprehensive feature sequence. The process involves comparing the comprehensive feature sequence with baseline operating data using a multi-level discrimination mechanism to determine the degree of pipeline network operation anomaly, transmitting control signals to the actuators via control channels, and generating inspection reports. Specifically, this includes: A baseline operating data is constructed as the inspection target, and the baseline operating data is aligned with the real-time operating data of the pipeline network in the time dimension. The baseline operating data includes baseline pressure value, baseline concentration value, baseline temperature value, and baseline deformation. For each time point, the degree of deviation between the comprehensive feature sequence and the baseline running data is calculated. Based on the importance of the parameters, weights are assigned to each type of parameter to generate a weighted deviation degree. The weighted deviation is mapped to the intensity range of the control signal using a standardized function, and the corresponding control signals are generated through the control channel and transmitted to the execution unit. Based on the control signals, an inspection report is generated using a multi-level discrimination mechanism. This multi-level discrimination mechanism includes a primary discrimination mechanism that calculates deviations in real time and provides control, an intermediate discrimination mechanism that periodically evaluates the progress and effectiveness of inspections, and a high-level discrimination mechanism that comprehensively analyzes long-term inspection data and provides optimization suggestions.

7. A computer device comprising a processor and a memory for storing a processor-executable program, characterized in that, When the processor executes the program stored in the memory, it implements the multi-parameter intelligent automated inspection method for hazardous gas pipelines as described in any one of claims 1-5.

8. A computer-readable storage medium storing a program, characterized in that, When the program is executed by the processor, it implements the multi-parameter intelligent automated inspection method for hazardous gas pipelines as described in any one of claims 1-5.

Images