Multi-element fusion pipeline early warning method and device

Abstract

The invention relates to a multi-element fusion pipeline early warning method and device, and the method comprises the steps: carrying out the compensation of an optical fiber vibration signal through a polarization state compensation model and the optical fiber basic data of a to-be-detected crude oil pipeline, and obtaining the effective signal intensity; detecting and classifying the construction instrument type and the construction action in the current video image to obtain a confidence coefficient belonging to a danger mining behavior; performing false alarm detection and threat detection according to the historical alarm point location data to obtain a false alarm correction factor and a threat triggering probability; performing risk quantification on the effective signal intensity, the confidence coefficient and the triggering threat probability according to the misinformation correction factor to obtain a pipeline early warning result; through fusion of optical fiber vibration signal compensation, video behavior recognition and historical data analysis, a multi-dimensional risk assessment model is constructed, the technical problems of signal distortion, high false alarm rate, response lag and the like in frozen soil region pipeline monitoring are effectively solved, and the method has the advantages of improving the monitoring precision, reducing the false alarm rate and realizing intelligent graded response.

Description

Technical Field

[0001] This invention relates to the field of pipeline safety monitoring technology, and in particular to a pipeline early warning method and device that integrates multiple elements. Background Technology

[0002] Pipeline safety monitoring methods can monitor pipelines in petrochemical, oil, and natural gas pipelines. Current technologies often employ fiber optic vibration monitoring, but in certain special environments, such as permafrost, farmland (where signal strength is weak in silt), deserts (sandy environments), and highways / railways, fiber optic vibration monitoring is not ideal due to environmental characteristics. The environmental characteristics of crude oil pipelines in permafrost regions are extremely unique and complex, posing one of the most severe challenges in pipeline engineering. The environmental characteristics of crude oil pipelines in permafrost regions can be summarized as: a dynamically unstable, extremely temperature-sensitive, highly corrosive, and poorly accessible complex system. While using fiber optic vibration sensing for pipeline safety monitoring in permafrost regions is technically feasible, the environmental characteristics prevent fiber optic vibration monitoring technology from adapting to extreme environments. This leads to problems such as false alarms, missed alarms, and response delays due to environmental noise, special geological conditions, and a lack of predictability for periodic human activities (such as busy farming seasons and peak traffic).

[0003] Therefore, there is an urgent need to propose a multi-factor integrated pipeline early warning method and device to solve the technical problems of false alarms, missed alarms and response delays in the existing fiber optic vibration monitoring method for early warning of crude oil pipelines in permafrost areas due to factors such as environmental noise, special geological conditions and lack of predictive ability for periodic human activities. Summary of the Invention

[0004] In view of this, it is necessary to provide a multi-factor integrated pipeline early warning method and device to solve the technical problems of false alarms, missed alarms and response delays in the existing fiber optic vibration monitoring method for early warning of crude oil pipelines in permafrost areas due to factors such as environmental noise, special geological conditions and lack of predictive ability for periodic human activities.

[0005] To address the aforementioned problems, in a first aspect, the present invention provides a multi-factor fusion pipeline early warning method, comprising: Acquire the fiber optic vibration signal, current video image, historical alarm point data, and basic fiber optic data of the crude oil pipeline to be inspected; The vibration signal of the optical fiber is compensated using a polarization state compensation model and the basic data of the optical fiber to obtain an effective signal strength. The types of construction equipment and construction actions in the current video image are detected and classified to obtain the confidence level of belonging to dangerous excavation behavior; Based on the historical alarm point data, false alarm detection and threat detection are performed to obtain the false alarm correction factor and the probability of triggering the threat. The pipeline early warning result is obtained by quantifying the effective signal strength, the confidence level, and the trigger threat probability based on the false alarm correction factor.

[0006] In one possible implementation, the fiber optic basic data includes the fiber's inherent attenuation value; the compensation of the fiber vibration signal using a polarization state compensation model and the fiber optic basic data to obtain an effective signal strength includes: Based on the environmental data in the fiber optic basic data, determine the additional attenuation value caused by environmental factors; The total attenuation coefficient is obtained by applying the additional attenuation value and the inherent attenuation value of the optical fiber to the polarization state compensation model. The effective signal strength is obtained by calculating the total attenuation coefficient and the optical fiber vibration signal.

[0007] In one possible implementation, the basic optical fiber data includes optical fiber distance, initial burial depth, soil type coefficient, and soil density; The step of determining the additional attenuation value caused by environmental factors based on the environmental data in the fiber optic basic data includes: The soil type is obtained by multiplying the soil density and the soil type coefficient. The target burial depth is obtained based on the initial burial depth, the preset initial attenuation coefficient, and the preset attenuation rate constant. Based on the target burial depth, the soil type, and the fiber optic distance, the additional attenuation value caused by environmental factors is obtained.

[0008] In one possible implementation, the detection and classification of construction equipment types and construction actions in the current video image to obtain a confidence level indicating that the activity belongs to a dangerous excavation includes: Retrieve historical video images; The YOLOv7-tiny model is trained based on the historical video images to obtain the trained YOLOv7-tiny model; the input of the trained YOLOv7-tiny model is the video images, and the output includes the type of construction equipment and the construction actions. The YOLOv7-tiny model is used to detect and classify the types of construction equipment and construction actions in the current video image to obtain the confidence level of dangerous excavation behavior.

[0009] In one possible implementation, the step of performing false alarm detection and threat detection based on the historical alarm location data to obtain a false alarm correction factor and a threat trigger probability includes: False alarm detection is performed on the historical alarm point data based on a long short-term memory network to obtain a false alarm correction factor; The time and region are divided according to the patterns of human activity, resulting in the division; Based on the historical alarm location data, the actual number of alarms and the total number of alarms for each region in each time period in the division results are calculated to obtain the trigger threat probability for each region in each time period.

[0010] In one possible implementation, the step of quantifying the risk of the effective signal strength, the confidence level, and the trigger threat probability based on the false alarm correction factor to obtain the pipeline early warning result includes: Based on the preset allocation weights and the false alarm correction factor, the effective signal strength, the confidence level, and the trigger threat probability are risk-quantified to obtain the final risk value; The final risk value is determined based on a preset graded response mechanism to obtain a pipeline early warning result.

[0011] In one possible implementation, the step of judging the final risk value according to a preset graded response mechanism to obtain a pipeline early warning result includes: When the final risk value is greater than the first preset value, the pipeline early warning result is determined to be a high-level early warning, and the valve is automatically closed, and the drone inspection process is started. When the final risk value is between the first preset value and the second preset value, the pipeline early warning result is determined to be a medium-level early warning and is manually reviewed. When the final risk value is less than the second preset value, the pipeline early warning result is determined to be a low-level early warning, and only no warning is recorded.

[0012] In one possible implementation, after acquiring the fiber optic vibration signal of the crude oil pipeline to be detected, the method further includes: The optical fiber vibration signal is decomposed based on the adaptive wavelet packet decomposition algorithm to obtain the decomposed signal; The interference in the current environment is detected based on the current video image, and the detection result is obtained; When the detection result indicates the presence of wind and rain interference, signals in the decomposed information that are less than a preset low-frequency threshold are suppressed to obtain the processed optical fiber vibration signal.

[0013] In one possible implementation, the formula for calculating the final risk value is: Final risk value R = ( A × S + B × V + C ×W )×(1- CF ) In the formula, S For effective signal strength, V For confidence level, W To trigger the threat probability, CF This is a false alarm correction factor. A , B and C Each element is assigned a preset weight.

[0014] Secondly, the present invention also provides a multi-element fusion pipeline early warning device, comprising: The data acquisition module is used to acquire the fiber optic vibration signal, current video image, historical alarm point data, and basic fiber optic data of the crude oil pipeline to be inspected. The signal compensation module is used to compensate the optical fiber vibration signal using a polarization state compensation model and the basic data of the optical fiber to obtain an effective signal strength. The image detection module is used to detect and classify the types of construction equipment and construction actions in the current video image to obtain the confidence level of belonging to dangerous excavation behavior; The alarm judgment module is used to perform false alarm detection and threat detection based on the historical alarm point data, and obtain the false alarm correction factor and the probability of triggering the threat. The risk quantification module is used to quantify the risk of the effective signal strength, the confidence level, and the probability of triggering threat based on the false alarm correction factor, so as to obtain the pipeline early warning result.

[0015] The beneficial effects of this invention are: acquiring fiber optic vibration signals, current video images, historical alarm point data, and basic fiber optic data of the crude oil pipeline to be monitored; compensating the fiber optic vibration signals using a polarization state compensation model and basic fiber optic data to obtain effective signal strength; detecting and classifying the types of construction equipment and construction actions in the current video images to obtain the confidence level of dangerous excavation behavior; performing false alarm detection and threat detection based on historical alarm point data to obtain false alarm correction factors and trigger threat probabilities; quantifying the risk of effective signal strength, confidence level, and trigger threat probabilities based on the false alarm correction factors to obtain pipeline early warning results; and constructing a multi-dimensional risk assessment model by integrating fiber optic vibration signal compensation, video behavior recognition, and historical data analysis, effectively solving technical problems such as signal distortion, high false alarm rate, and response lag in pipeline monitoring in permafrost areas, and having the advantages of improving monitoring accuracy, reducing false alarm rate, and achieving intelligent hierarchical response. Attached Figure Description

[0016] Figure 1 A schematic flowchart of an embodiment of the pipeline early warning method with multi-element fusion provided by the present invention; Figure 2 For the present invention Figure 1 A schematic diagram of an embodiment of step S102; Figure 3 For the present invention Figure 1 A schematic flowchart of an embodiment of step S103; Figure 4 For the present invention Figure 1 A schematic flowchart of an embodiment of step S104; Figure 5 This is a schematic diagram of an embodiment of the multi-element fusion pipeline early warning device provided by the present invention. Detailed Implementation

[0017] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0018] like Figure 1 As shown, a specific embodiment of the present invention discloses a pipeline early warning method based on multi-factor fusion, comprising: S101. Acquire the fiber optic vibration signal, current video image, historical alarm point data, and basic fiber optic data of the crude oil pipeline to be inspected.

[0019] The embodiments of the present invention can be applied to pipeline early warning systems in special environments. The pipelines can be pipelines installed in industries such as petrochemicals, oil pipelines, and natural gas pipelines. The pipeline early warning system can be a software system running on a terminal device, which can be a server, tablet computer, mobile phone, or other devices. The embodiments of the present invention do not limit the specific type of terminal device.

[0020] Among them, the crude oil pipeline to be inspected is a pipeline laid in a special environment, such as silt, sand, vertical highway, and bridge. In this embodiment of the invention, the special environment is taken as an example, and the optical fiber vibration signal, current video image, historical alarm point data and optical fiber basic data of the crude oil pipeline to be inspected can be obtained.

[0021] S102. The optical fiber vibration signal is compensated by the polarization state compensation model and the basic data of the optical fiber to obtain the effective signal strength.

[0022] The polarization state compensation model is a mathematical model used to correct the attenuation of optical fiber signals during transmission due to soil characteristics and burial depth. Specifically, it can be implemented using an attenuation superposition formula based on fiber distance, soil density, and type coefficient. This model improves the effectiveness of vibration signals by compensating for signal distortion caused by environmental factors. In specific areas, low-light cameras can be deployed every 200 meters to capture current video images. The effective signal strength refers to the quantized value of the vibration signal after environmental attenuation compensation. It is specifically obtained by calculating the product of the total attenuation coefficient and the original signal, and is used to reflect the vibration energy caused by real threats.

[0023] S103. Detect and classify the types of construction equipment and construction actions in the current video image to obtain the confidence level of belonging to dangerous excavation behavior.

[0024] Among them, confidence level refers to the probability of matching construction equipment detected in video images with dangerous actions. Specifically, an improved lightweight target detection model can be used to classify and output the equipment type and action trajectory to assess the visual dimension risk.

[0025] S104. Based on historical alarm location data, perform false alarm detection and threat detection to obtain false alarm correction factors and threat triggering probabilities.

[0026] The false alarm correction factor is a false alarm probability adjustment coefficient calculated based on the spatiotemporal distribution patterns of historical alarms. Specifically, it can be generated by analyzing false alarm patterns in historical data using a long short-term memory network, and is used to dynamically correct current risk assessment results. The trigger threat probability refers to the probability of a real alarm event occurring within a specific time period and region. Specifically, it can be calculated by statistically analyzing the alarm effectiveness ratio in the time-region dimension of historical alarm data, and is used to quantify the impact of human activity patterns on current risks.

[0027] S105. Based on the false alarm correction factor, the effective signal strength, confidence level and trigger threat probability are quantified to obtain the pipeline early warning result.

[0028] Specifically, the system first simultaneously collects fiber optic vibration signals and real-time video data. This data, combined with fundamental information such as fiber optic distance and soil parameters, is input into a polarization state compensation model to calculate the additional attenuation caused by soil type and burial depth. The model then outputs the effective signal strength after environmental compensation. Simultaneously, an improved visual detection model identifies machine types such as excavators and bulldozers from the video stream and determines whether the machine's movement trajectory constitutes dangerous excavation behavior, outputting a confidence score. Furthermore, historical alarm data undergoes spatiotemporal distribution analysis to extract high-frequency false alarm periods and regions, generating correction factors and statistically analyzing the probability of actual threats occurring within different time periods. Finally, the effective signal strength, confidence score, and threat triggering probability are fused according to preset weights and multiplied by the false alarm correction factor to obtain a comprehensive risk value that drives tiered early warning systems.

[0029] Compared with existing technologies, this embodiment provides the following advantages: acquiring fiber optic vibration signals, current video images, historical alarm point data, and basic fiber optic data of the crude oil pipeline to be monitored; compensating the fiber optic vibration signals using a polarization state compensation model and basic fiber optic data to obtain effective signal strength; detecting and classifying the types of construction equipment and construction actions in the current video images to obtain the confidence level of dangerous excavation behavior; performing false alarm detection and threat detection based on historical alarm point data to obtain false alarm correction factors and trigger threat probabilities; quantifying the risk of effective signal strength, confidence level, and trigger threat probability based on the false alarm correction factors to obtain pipeline early warning results; and constructing a multi-dimensional risk assessment model by integrating fiber optic vibration signal compensation, video behavior recognition, and historical data analysis. This effectively solves the technical problems of signal distortion, high false alarm rate, and delayed response in pipeline monitoring in permafrost areas, and has the advantages of improving monitoring accuracy, reducing false alarm rate, and achieving intelligent hierarchical response.

[0030] In some embodiments of the present invention, after step S101, the method further includes: The optical fiber vibration signal is decomposed based on the adaptive wavelet packet decomposition algorithm to obtain the decomposed signal.

[0031] The adaptive wavelet packet decomposition algorithm is a time-frequency analysis method that automatically adjusts the number of decomposition levels and frequency band division based on signal characteristics. Specifically, it can be implemented by dynamically optimizing the wavelet basis function and decomposition depth, adapting to the non-stationary signal characteristics of complex environments in permafrost regions. The decomposed signal refers to the multiple frequency band components obtained through multi-level wavelet packet decomposition, which can be achieved through recursive filtering and reconstruction operations to separate vibration information of different frequency components. For example, high-frequency signals of 6-12Hz and low-frequency signals of 0-2Hz.

[0032] The interference in the current environment is detected based on the current video image, and the detection results are obtained.

[0033] Interference detection refers to identifying whether there are natural interference factors such as wind and rain in the current environment through video images. Specifically, it can be implemented using motion detection algorithms based on optical flow or background modeling to determine the source of environmental noise.

[0034] When the detection result indicates the presence of wind and rain interference, signals in the decomposed information that are less than a preset low-frequency threshold are suppressed to obtain the processed fiber optic vibration signal.

[0035] The preset low-frequency threshold refers to a pre-defined frequency boundary value, such as a value within the range of 0.6Hz to 2Hz, used to distinguish between low-frequency vibrations caused by wind and rain and high-frequency vibrations generated by mechanical excavation. When the detection result indicates the presence of wind and rain interference, the low-frequency signal of 0-2Hz can be considered low-frequency vibration caused by wind and rain and can be automatically suppressed, while the high-frequency signal of 6-12Hz, which is high-frequency vibration generated by mechanical excavation, needs to be retained.

[0036] Specifically, after acquiring the fiber optic vibration signal, it is first decomposed into multiple frequency bands using an adaptive wavelet packet decomposition algorithm. This algorithm automatically selects the optimal decomposition level based on the signal energy distribution; for example, it increases the decomposition depth to capture transient features when sudden temperature changes cause signal abrupt shifts in permafrost regions. Subsequently, a video analysis module detects the current environment; for example, it identifies wind and rain interference when trees are violently swaying or heavy rain is detected. At this point, the low-frequency signals after decomposition are suppressed, for example, by attenuating the energy of frequency bands below 1Hz by 50%, retaining the high-frequency components for subsequent analysis. This eliminates the interference of periodic low-frequency noise caused by wind and rain on the effective signal.

[0037] Through the above technical solutions, this application effectively reduces the false alarm rate caused by natural disturbances such as wind and rain. In scenarios with large diurnal temperature differences and variable weather conditions in permafrost regions, it can accurately distinguish between environmental noise and real threat signals, improving the robustness of vibration signal analysis. Simultaneously, by suppressing low-frequency signals in the frequency domain rather than directly eliminating them, it avoids the problem of missing detection of real excavation activities due to over-filtering.

[0038] In some embodiments of the present invention, such as Figure 2 As shown, the basic data of the optical fiber includes the inherent attenuation value of the optical fiber; step S102 includes: S201. Determine the additional attenuation value caused by environmental factors based on the environmental data in the basic optical fiber data. S202. Based on the polarization state compensation model, the total attenuation coefficient is obtained by considering the additional attenuation value and the inherent attenuation value of the optical fiber.

[0039] The inherent attenuation of optical fiber refers to the inherent loss of optical signals caused by the optical fiber material itself. This loss can be calibrated using an optical power meter under standard testing conditions and serves as a fundamental parameter for calculating the total attenuation coefficient. The specific total attenuation coefficient of the optical fiber signal is determined by the inherent attenuation (…). α_fiber ) and additional attenuation introduced by environmental factors ( α_env Composed of, i.e., total attenuation coefficient α_total = α_fiber + α_env .

[0040] S203. Calculate the total attenuation coefficient and the fiber vibration signal to obtain the effective signal strength.

[0041] In a specific embodiment of the present invention, when the optical fiber vibration signal is obtained, the original intensity can be determined. For example, the original intensity can be 0.6, then the effective signal intensity S = original intensity + α_total.

[0042] Specifically, in the complex geological conditions of permafrost regions, the attenuation of the environment and pipelines is closely related. Therefore, calculating the environmental data yields an additional attenuation value. The polarization state compensation model constructs a full-parameter attenuation model that conforms to the special geological conditions of permafrost regions by superimposing the inherent attenuation of the optical fiber with the additional attenuation. Finally, this model is used to compensate for the original vibration signal, effectively eliminating signal distortion caused by environmental factors.

[0043] In some embodiments of the present invention, the basic data of the optical fiber includes the optical fiber distance, initial burial depth, soil type coefficient, and soil density; step S201 includes: Soil type is obtained by multiplying soil density and soil type coefficient.

[0044] Different soils have different conductivity, and specific soil types... α_soil = k s × ρ_soil , k s Soil type coefficient, ρ_soil Soil density.

[0045] The target burial depth is obtained based on the initial burial depth, the preset initial attenuation coefficient, and the preset attenuation rate constant.

[0046] The initial burial depth refers to the original burial depth data recorded during pipeline installation, which can be obtained from as-built drawings or ground-penetrating radar detection. This initial depth, along with soil parameters, determines the impact of the underground environment on signal attenuation. The preset attenuation rate constant is an empirical coefficient describing the relationship between burial depth changes and signal attenuation. It can be determined through regression analysis of historical data and is used to dynamically correct the impact of burial depth changes caused by frost heave or thawing on the signal. The soil type coefficient, preset initial attenuation coefficient, and preset attenuation rate constant can be obtained by fitting measured data using the least squares method. Specifically, the deeper the burial depth, the less noticeable the vibration transmission; the target burial depth value... , k 1 and k 2 represents the preset initial attenuation coefficient and the preset attenuation rate constant, respectively. e It is a natural constant in mathematics. h This is the initial burial depth value.

[0047] Based on the target burial depth, soil type, and fiber optic distance, the additional attenuation value caused by environmental factors is obtained.

[0048] Among these, fiber optic distance refers to the physical interval between two adjacent monitoring points in a distributed fiber optic sensor, specifically measured using an optical time-domain reflectometer (OTDR). Its value directly affects the spatial distribution characteristics of signal attenuation. Soil type coefficient is a dimensionless parameter characterizing the ability of different soil types to impede the propagation of vibration waves. It can be calibrated through a correlation between laboratory soil sample compression tests and acoustic wave transmission tests. Multiplying it by soil density quantifies the signal attenuation characteristics of permafrost regions due to their unique geological features. Additional attenuation (i.e., additional attenuation value) is caused by specific environmental factors. , L This refers to the fiber optic distance.

[0049] Specifically, the polarization state compensation model constructed by combining all factors is shown in formula (1): (1) In the formula, α_total The total attenuation coefficient is... α_fiber For the target burial depth value, k 1. k 2 and k s These are the preset initial attenuation coefficient, the preset attenuation rate constant, and the soil type coefficient, respectively. e It is a natural constant in mathematics. h This is the initial burial depth value. ρ_ soil Soil type L This refers to the fiber optic distance.

[0050] Specifically, under the complex geological conditions of permafrost regions, the product of soil density and soil type coefficient can transform the mechanical impedance characteristics of the permafrost layer into quantifiable soil type parameters. An exponential function model consisting of a preset initial attenuation coefficient and a preset attenuation rate constant can dynamically reflect the nonlinear influence of permafrost depth changes on signal attenuation; for example, the reduction in depth caused by frost heave will increase the attenuation through the exponential term. The calculation of additional attenuation values, by integrating three dimensions—target burial depth, soil type, and fiber optic distance—can accurately characterize the propagation loss characteristics of vibration signals in the permafrost medium.

[0051] Through the above technical solution, this application effectively solves the technical problem of fiber optic vibration signal distortion caused by complex geological conditions in permafrost regions. The joint calculation of soil type coefficient and soil density accurately reflects the absorption characteristics of the permafrost layer on vibration waves. The introduction of dynamic burial depth parameters can adapt to the common frost heave and thaw settlement phenomena in permafrost regions. The multi-dimensional calculation model with additional attenuation values ​​makes the signal compensation process more closely resemble the actual geological environment. This multi-parameter fusion compensation mechanism significantly improves the calculation accuracy of effective signal strength, providing a reliable data foundation for subsequent risk quantification.

[0052] In some embodiments of the present invention, such as Figure 3 As shown, step S103 includes: S301. Obtain historical video images.

[0053] The historical video images can be image data collected from crude oil pipelines in special environments (such as permafrost areas), or image data obtained from other databases under special environments. Specific embodiments of the present invention are not limited here.

[0054] S302. Train the YOLOv7-tiny model based on historical video images to obtain the trained YOLOv7-tiny model. The input of the trained YOLOv7-tiny model is video images, and the output includes the type of construction equipment and construction actions.

[0055] The YOLOv7-tiny model is a lightweight object detection algorithm that utilizes pre-trained weights combined with transfer learning. Its purpose is to optimize computational efficiency and achieve real-time detection in resource-constrained environments such as permafrost regions. The pipeline construction equipment feature library is a dataset storing historical construction equipment types and actions (e.g., vibration pattern recognition for directional drilling rigs). This can be implemented using a structured database or feature vector matrix, providing diverse training samples to enhance classification accuracy. Adding the pipeline construction equipment feature library to the YOLOv7-tiny model enables it to identify and output the types of construction equipment and their actions in images.

[0056] S303. The YOLOv7-tiny model is used to detect and classify the types of construction equipment and construction actions in the current video image to obtain the confidence level of dangerous excavation behavior.

[0057] The confidence level of dangerous excavation behavior refers to the probability value output by the model, which can be calculated using the softmax function or classifier threshold. Its function is to quantitatively assess the risk level of construction behavior and provide input for subsequent risk quantification.

[0058] Specifically, historical video images are input into the YOLOv7-tiny model for object detection, extracting the type and motion features of construction equipment, such as the bucket trajectory of an excavator or the direction of travel of a bulldozer. This feature data is standardized and used to construct a pipeline construction equipment feature library, which is then embedded into the model as prior knowledge. The improved model, by fusing the historical feature library, can more accurately identify the type of construction equipment in the current video, such as distinguishing between ordinary engineering vehicles and high-risk excavation equipment, while also analyzing their motion patterns, such as determining whether it is continuous excavation or a short-term stop. Finally, the model outputs a probability value indicating that the current scene represents dangerous excavation behavior, which is normalized to a confidence level within the range of 0 to 1.

[0059] In some embodiments of the present invention, such as Figure 4 As shown, step S104 includes: S401. False alarm detection is performed on historical alarm point data based on long short-term memory network to obtain false alarm correction factor.

[0060] Long Short-Term Memory (LSTM) networks are recurrent neural networks with memory units and gating mechanisms. Specifically, they can be implemented using a network structure that includes input gates, forget gates, and output gates. LSTM networks are used to capture temporal dependencies in historical alarm data and identify false alarm patterns. By learning the characteristics of historical alarm data, such as event type, level, duration, signal strength, and false alarm identifiers in the alarm response, the probability of a current alarm being a false alarm (i.e., a false alarm correction factor) can be identified, thereby optimizing the accuracy of false alarm filtering.

[0061] S402. Divide time and region according to the patterns of human activities to obtain the division results.

[0062] Among them, human activity patterns refer to periodic behavioral patterns derived from historical data statistics. Specifically, this can be achieved by dividing the day into work periods and rest periods, and dividing the region into industrial areas, agricultural areas, and residential areas, to establish a threat probability calculation framework in terms of time and space. For example, the farmland harvesting season: September 1st to October 15th, 6:00-18:00 daily.

[0063] S403. Based on historical alarm point data, calculate the actual number of alarms and the total number of alarms for each area in each time period in the division results to obtain the trigger threat probability for each area in each time period.

[0064] The triggering threat probability refers to the proportion of actual alarm events to the total number of alarms within a specific time period and area. It is calculated by dividing the number of actual alarms by the total number of alarms and is used to quantify the threat level posed by human activities to pipeline safety. The triggering threat probability W = number of actual alarms in that time period / total number of alarms in that time period, reflecting the probability that human activities will trigger a real threat during that time period.

[0065] Specifically, by modeling the temporal characteristics of historical alarm data using a Long Short-Term Memory (LSTM) network, false alarms caused by equipment noise or environmental interference can be identified, generating correction factors that reflect the patterns of false alarm occurrence. Furthermore, the monitoring area is divided into multiple spatiotemporal units based on human activity patterns; for example, the working hours in industrial areas are defined as 8:00-18:00, and the busy farming season in agricultural areas is defined as May-October. Within each spatiotemporal unit, a dynamically adjusted threat probability assessment model is constructed by statistically analyzing the ratio of actual threat events to the total number of alarms in historical alarm data. For example, the threat probability in residential areas during nighttime hours can be set to 0.05, while it may reach 0.3 during working hours in industrial areas.

[0066] Through the above technical solution, this application effectively solves the problem of false alarms caused by the lack of periodic human activity prediction in pipeline monitoring in permafrost areas. It reduces the impact of environmental noise interference by using dynamic correction factors, and improves the accuracy of identifying real threat events by utilizing the threat probability in the spatiotemporal dimension. For example, it realizes real-time updates of threat probability during working hours in high-incidence construction areas, thereby reducing the risk of response lag.

[0067] In some embodiments of the present invention, step S105 includes: The risk is quantified by using preset weights and false alarm correction factors to assess the strength of the effective signal, confidence level, and probability of triggering a threat, resulting in a final risk value.

[0068] Among them, the preset allocation weight refers to the contribution ratio of different risk elements in the quantification process, which can be determined by empirical values ​​or machine learning optimization methods. For example, the weights of effective signal strength, confidence level, and trigger threat probability are set to 0.4, 0.4, and 0.2 respectively, so as to balance the impact of different elements on risk. The false alarm correction factor refers to the correction coefficient generated based on historical false alarm data. Specifically, it can be dynamically calculated by using a long short-term memory network to reduce the interference of false alarms on risk quantification. The effective signal strength refers to the intensity of fiber optic vibration signal after being processed by the polarization state compensation model, reflecting the actual threat level of environmental vibration. The confidence level refers to the probability of dangerous excavation behavior detected by the YOLOv7-tiny model, representing the risk level of construction equipment type and action. The trigger threat probability refers to the alarm trigger probability calculated after dividing the time and region according to the historical alarm point data, reflecting the potential threat of periodic human activities to the pipeline. The final risk value refers to the comprehensive risk index after weighted calculation and false alarm correction, used to judge the warning level. The calculation of the final risk value is shown in formula (2): Final risk value R = ( A × S + B × V + C × W )×(1- CF (2) In the formula, S For effective signal strength, V For confidence level, W To trigger the threat probability, CF This is a false alarm correction factor. A , B and C The preset weights for the corresponding elements can be 0.4, 0.4, and 0.2 respectively.

[0069] The final risk value is determined based on the preset graded response mechanism, and the pipeline early warning result is obtained.

[0070] Among them, the preset graded response mechanism refers to the different early warning levels and corresponding measures based on the final risk value, such as setting high, medium and low early warning thresholds and associating them with response actions such as valve control, drone inspection and manual review.

[0071] Specifically, the risk quantification process first multiplies the effective signal strength, confidence level, and threat triggering probability by their respective preset weights to obtain weighted values ​​for each element. Then, a false alarm correction factor is used to adjust the weighted results overall, ultimately yielding a comprehensive risk value. For example, a high false alarm correction factor indicates a higher likelihood of false alarms in the current environment; in this case, the final risk value will be dynamically lowered to reduce the probability of false alarms. Subsequently, the final risk value is assessed based on thresholds in a preset tiered response mechanism.

[0072] In some embodiments of the present invention, the final risk value is determined according to a preset graded response mechanism to obtain a pipeline early warning result, including: When the final risk value exceeds the first preset value, the pipeline warning result is determined to be a high-level warning, and the valve is automatically closed, initiating the drone inspection process.

[0073] The preset graded response mechanism refers to pre-setting early warning response strategies corresponding to different risk threshold ranges. This can be implemented using piecewise functions or conditional judgment logic, such as dividing risk values ​​into high, medium, and low intervals, each with a different processing flow. The first and second preset values ​​are two threshold parameters used to classify risk levels, which can be determined through historical data analysis or experimental calibration. For example, the first preset value could be set to 0.8, and the second preset value to 0.5. Automatic valve closure refers to triggering the closing action of valve actuators along the pipeline through control commands, which can be implemented using industrial IoT communication protocols. The drone inspection process refers to calling upon drone equipment to conduct aerial inspections of the alarm area, which can be implemented using preset flight path planning algorithms and image transmission modules.

[0074] When the final risk value is between the first and second preset values, the pipeline warning result is determined to be a medium-level warning and is manually reviewed.

[0075] Manual verification refers to pushing alarm information to the monitoring platform for confirmation by operators, which can be achieved through a work order system or message notification interface.

[0076] When the final risk value is less than the second preset value, the pipeline early warning result is determined to be a low-level early warning, and only no warning is recorded.

[0077] Specifically, in pipeline monitoring scenarios in permafrost regions, the final risk value is calculated by fusing fiber optic signals, construction behavior confidence levels, and threat probabilities. When the final risk value exceeds a first preset value, it indicates an extremely high risk of damage. In this case, automatically closing valves can prevent the leakage from spreading, and drone inspections can quickly acquire real-time images of the site. When the risk value is in the intermediate range, the system prioritizes pushing alarms to the manual review stage to avoid misoperation caused by environmental noise or occasional vibrations. When the risk value is below a second preset value, the data is only recorded in the log to avoid frequent alarms interfering with normal monitoring work. For example, during the busy farming season, periodic human activities may generate low-frequency vibration signals. Through a classification mechanism, these can be categorized as low-risk events, recorded without triggering alarms. This process achieves accurate threat assessment in complex environments by dynamically adjusting risk weights and correcting the impact of false alarms.

[0078] Through the aforementioned technical solution, this application can dynamically adjust response strategies based on risk levels, ensuring rapid handling of high-risk events while reducing false alarm rates caused by environmental noise and periodic human activities in permafrost regions. For example, when heavy rain causes abnormally strong fiber optic signals, if the threat probability and the confidence level of construction activities do not increase synchronously, the system can classify it as a medium-risk event through a grading mechanism. After manual verification to eliminate environmental interference, a decision can be made on whether to activate the emergency procedure, thereby avoiding economic losses caused by mistakenly closing valves. In addition, the silent handling of low-risk events reduces the workload of monitoring personnel and improves overall monitoring efficiency.

[0079] In one embodiment, taking a crude oil pipeline in a permafrost region as an example: A 9Hz vibration signal (original intensity 0.6) was detected. After compensation for attenuation, the strength is 0.65. The video identified an excavator (confidence level 0.95). The area is a non-sensitive zone (W=0); The false alarm correction factor for this location is CF=0.1; The calculated value is R = (0.4 × 0.65 + 0.4 × 0.95 + 0) × (1 - 0.1) = 0.54 → triggering a medium-level warning.

[0080] Furthermore, different colors can be set according to different warning levels, such as red for high level, yellow for medium level, and blue for low level. The specific settings can be made according to the needs, and this embodiment of the invention does not impose any limitations.

[0081] To better implement the multi-element fusion pipeline early warning method in the embodiments of the present invention, based on the multi-element fusion pipeline early warning method, the embodiments of the present invention also provide a multi-element fusion pipeline early warning device, such as... Figure 5As shown, the multi-element integrated pipeline early warning device 500 includes: The data acquisition module 501 is used to acquire the fiber optic vibration signal, current video image, historical alarm point data and basic fiber optic data of the crude oil pipeline to be inspected. The signal compensation module 502 is used to compensate for the fiber vibration signal by using the polarization state compensation model and the basic data of the fiber to obtain an effective signal strength; The image detection module 503 is used to detect and classify the types of construction equipment and construction actions in the current video image, and obtain the confidence level of belonging to dangerous excavation behavior; The alarm judgment module 504 is used to perform false alarm detection and threat detection based on historical alarm point data, and obtain the false alarm correction factor and the probability of triggering threat. The risk quantification module 505 is used to quantify the risk of effective signal strength, confidence level and trigger threat probability based on the false alarm correction factor, and obtain pipeline early warning results.

[0082] The multi-element fusion pipeline early warning device 500 provided in the above embodiments can realize the technical solutions described in the above multi-element fusion pipeline early warning method embodiments. The specific implementation principles of each module or unit can be found in the corresponding content in the above multi-element fusion pipeline early warning method embodiments, which will not be repeated here.

[0083] The multi-element fusion pipeline early warning method and device provided by the present invention have been described in detail above. Specific examples have been used to illustrate the principle and implementation of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation and application scope based on the idea of ​​the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A pipeline early warning method integrating multiple elements, characterized in that, include: Acquire the fiber optic vibration signal, current video image, historical alarm point data, and basic fiber optic data of the crude oil pipeline to be inspected; The vibration signal of the optical fiber is compensated using a polarization state compensation model and the basic data of the optical fiber to obtain an effective signal strength. The types of construction equipment and construction actions in the current video image are detected and classified to obtain the confidence level of belonging to dangerous excavation behavior; Based on the historical alarm point data, false alarm detection and threat detection are performed to obtain the false alarm correction factor and the probability of triggering the threat. The pipeline early warning result is obtained by quantifying the effective signal strength, the confidence level, and the trigger threat probability based on the false alarm correction factor.

2. The pipeline early warning method based on multi-factor fusion according to claim 1, characterized in that, The basic data of the optical fiber includes the inherent attenuation value of the optical fiber; the compensation of the optical fiber vibration signal using the polarization state compensation model and the basic data of the optical fiber to obtain the effective signal strength includes: Based on the environmental data in the fiber optic basic data, determine the additional attenuation value caused by environmental factors; The total attenuation coefficient is obtained by applying the additional attenuation value and the inherent attenuation value of the optical fiber to the polarization state compensation model. The effective signal strength is obtained by calculating the total attenuation coefficient and the optical fiber vibration signal.

3. The pipeline early warning method based on multi-factor fusion according to claim 2, characterized in that, The basic data of the optical fiber includes the optical fiber distance, initial burial depth, soil type coefficient, and soil density; The step of determining the additional attenuation value caused by environmental factors based on the environmental data in the fiber optic basic data includes: The soil type is obtained by multiplying the soil density and the soil type coefficient. The target burial depth is obtained based on the initial burial depth, the preset initial attenuation coefficient, and the preset attenuation rate constant. Based on the target burial depth, the soil type, and the fiber optic distance, the additional attenuation value caused by environmental factors is obtained.

4. The pipeline early warning method based on multi-factor fusion according to claim 1, characterized in that, The process of detecting and classifying the types of construction equipment and construction actions in the current video image to obtain the confidence level of belonging to dangerous excavation behavior includes: Retrieve historical video images; The YOLOv7-tiny model is trained based on the historical video images to obtain the trained YOLOv7-tiny model; the input of the trained YOLOv7-tiny model is the video images, and the output includes the type of construction equipment and the construction actions. The YOLOv7-tiny model is used to detect and classify the types of construction equipment and construction actions in the current video image to obtain the confidence level of dangerous excavation behavior.

5. The pipeline early warning method based on multi-factor fusion according to claim 1, characterized in that, The step of performing false alarm detection and threat detection based on the historical alarm point data to obtain false alarm correction factors and threat triggering probabilities includes: False alarm detection is performed on the historical alarm point data based on a long short-term memory network to obtain a false alarm correction factor; The time and region are divided according to the patterns of human activity, resulting in the division; Based on the historical alarm location data, the actual number of alarms and the total number of alarms for each region in each time period in the division results are calculated to obtain the trigger threat probability for each region in each time period.

6. The pipeline early warning method based on multi-factor fusion according to claim 1, characterized in that, The step of quantifying the risk of the effective signal strength, the confidence level, and the trigger threat probability based on the false alarm correction factor to obtain the pipeline early warning result includes: Based on the preset allocation weights and the false alarm correction factor, the effective signal strength, the confidence level, and the trigger threat probability are risk-quantified to obtain the final risk value; The final risk value is determined based on a preset graded response mechanism to obtain a pipeline early warning result.

7. The pipeline early warning method based on multi-factor fusion according to claim 6, characterized in that, The step of judging the final risk value according to the preset graded response mechanism to obtain the pipeline early warning result includes: When the final risk value is greater than the first preset value, the pipeline early warning result is determined to be a high-level early warning, and the valve is automatically closed, and the drone inspection process is started. When the final risk value is between the first preset value and the second preset value, the pipeline early warning result is determined to be a medium-level early warning and is manually reviewed. When the final risk value is less than the second preset value, the pipeline early warning result is determined to be a low-level early warning, and only no warning is recorded.

8. The pipeline early warning method based on multi-factor fusion according to claim 1, characterized in that, After acquiring the fiber optic vibration signal of the crude oil pipeline to be detected, the process further includes: The optical fiber vibration signal is decomposed based on the adaptive wavelet packet decomposition algorithm to obtain the decomposed signal; The interference in the current environment is detected based on the current video image, and the detection result is obtained; When the detection result indicates the presence of wind and rain interference, signals in the decomposed information that are less than a preset low-frequency threshold are suppressed to obtain the processed optical fiber vibration signal.

9. The pipeline early warning method based on multi-factor fusion according to claim 6, characterized in that, The formula for calculating the final risk value is as follows: Final risk value R =( A × S + B × V + C × W )×(1- CF ) In the formula, S For effective signal strength, V For confidence level, W To trigger the threat probability, CF This is a false alarm correction factor. A , B and C Each element is assigned a preset weight.

10. A multi-element integrated pipeline early warning device, characterized in that, include: The data acquisition module is used to acquire the fiber optic vibration signal, current video image, historical alarm point data, and basic fiber optic data of the crude oil pipeline to be inspected. The signal compensation module is used to compensate the optical fiber vibration signal using a polarization state compensation model and the basic data of the optical fiber to obtain an effective signal strength. The image detection module is used to detect and classify the types of construction equipment and construction actions in the current video image to obtain the confidence level of belonging to dangerous excavation behavior; The alarm judgment module is used to perform false alarm detection and threat detection based on the historical alarm point data, and obtain the false alarm correction factor and the probability of triggering the threat. The risk quantification module is used to quantify the risk of the effective signal strength, the confidence level, and the probability of triggering threat based on the false alarm correction factor, so as to obtain the pipeline early warning result.

Images