Pipeline safety early warning event type identification optimization method and pipeline safety early warning system

Abstract

The invention relates to the technical field of pipeline safety monitoring, in particular to a pipeline safety early warning event type identification optimization method and a pipeline safety early warning system. By comparing the feature centers of the same type of events in the new environment and the original environment, the propagation difference of the same vibration source events in different soil environments is evaluated, and the identification model in the new environment is obtained by adjustment on the basis of the original identification model, so that the adaptability optimization of the identification model for the new soil environment is realized. The method ensures high accuracy of event type identification in a new environment, and has the advantages of simple operation and low cost.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline safety monitoring, and in particular to a pipeline safety early warning event type identification optimization method and a pipeline safety early warning system. Background Art

[0002] Pipelines are commonly used in the oil and gas industry to transmit liquids and gases. Due to the complexity of the geographical environment they pass through, they are very likely to be damaged or even leak or explode during production and operation due to manual or mechanical excavation. Therefore, it is very necessary to detect soil vibration signals along the pipeline in real time and promptly detect construction events along the pipeline to ensure the safety of the pipeline and its transmission medium.

[0003] Traditional pipeline vibration event identification usually uses manual feature extraction and machine learning classification methods. This method has the disadvantages of small event sample size, low accuracy, and easy false alarms and omissions in specific environments. When the system is deployed in a new soil geological environment, the soil type, soil hardness, moisture content, soil layer structure, and stress state will change the characteristics of vibration signal propagation, thereby changing the distribution of event signal feature space and hard identification accuracy. At the same time, conducting a large number of simulation experiments in a new environment has safety risks and cost limitations. Summary of the invention

[0004] In view of the problems that the existing systems in the prior art have low recognition accuracy when applied to new soil geological environments, and there are safety risks and cost limitations in conducting a large number of simulation experiments in the new environment, the present invention provides a pipeline safety warning event type recognition optimization method and a pipeline safety warning system. By comparing the feature centers of the same type of events in the new environment and the original environment, the propagation differences of the same vibration source events in different soil environments are evaluated, and the recognition model in the new environment is adjusted based on the original recognition model to ensure the accuracy of event type recognition in the new environment, while having the advantages of simple operation and low cost.

[0005] A pipeline safety early warning event type identification optimization method of the present invention is used for pipeline safety early warning event type identification in a new environment. The method comprises the following steps: S1. Acquire data of different types of intrusion events in the original environment, extract features from the acquired data, and determine feature centers of different types of intrusion events in the original environment based on the extracted features; S2. Acquire data of the test intrusion event in the new environment, extract features from the acquired data according to the same feature extraction method as in step S1, and determine the feature center of the test intrusion event in the new environment based on the extracted features; S3. Calculate the offset of the feature center of the test intrusion event in the new environment relative to the feature center of the corresponding intrusion event in the different types of intrusion events in the original environment; S4. The feature centers of different types of intrusion events in the original environment are shifted according to the offset amount to obtain the feature centers of different types of intrusion events in the new environment; S5. Obtain the classification decision surface of different types of intrusion events in the new environment according to the feature centers of different types of intrusion events in the new environment; S6. Use the classification decision surface of different types of intrusion events in the new environment to identify event types.

[0006] Preferably, the test intrusion event is at least one of different types of intrusion events. Further, the different types of intrusion events include manual excavation events, vehicle traffic events, and mechanical construction events. Still further, the test intrusion event is at least one of manual excavation events, vehicle traffic events, and mechanical construction events.

[0007] Preferably, step S1 comprises the following steps: S11. Acquire data of different types of intrusion events in the original environment, wherein the data of different types of intrusion events include vibration signals corresponding to different types of intrusion events and real intrusion event types; S12. Extracting features from the vibration signal to obtain a first feature class; S13. Obtaining a first sample based on the first feature class and the corresponding real intrusion event type; S14. Looping through step S11 to step S13 until a number of first samples are obtained, and then terminating the loop to obtain a first training set based on the number of first samples; S15. Use the first training set to train the classification model to obtain the feature centers and classification decision surfaces of different types of intrusion events in the original environment.

[0008] Furthermore, after step S15, the following steps are also included: S16. Use the classification decision surface of different types of intrusion events in the original environment to identify event types; S17. If the accuracy of event type identification using the classification decision surface of different types of intrusion events in the original environment is greater than the set threshold, proceed to step S2; otherwise, proceed to step S11.

[0009] Further, step S2 includes the following steps: S21. Acquire test intrusion event data in the new environment, the test intrusion event data includes the vibration signal corresponding to the test intrusion event and the real intrusion event type; S22. Extract features from the vibration signal using the same feature extraction method as in step S12 to obtain a second feature class; S23. Obtaining a second sample based on the second feature class and the corresponding real intrusion event type; S24. Looping through step S21 to step S23 until a number of second samples are obtained, and then terminating the loop, obtaining a second training set based on the number of second samples; S25. Input the second training set into the classification model trained in the original environment to obtain the feature center of the test intrusion event in the new environment.

[0010] Further, step S6 includes the following steps: S61. Real-time acquisition of data on intrusion events to be identified in the new environment, the data on intrusion events to be identified include vibration signals corresponding to the intrusion events to be identified; S62. Extract features from the vibration signal using the same feature extraction method as in step S12 to obtain feature classes to be identified; S63. Determine the event type of the intrusion event to be identified based on the feature class to be identified and the classification decision surface of different types of intrusion events in the new environment.

[0011] Furthermore, after step S63, the following steps are also included: S64. If the real type of the intrusion event to be identified is inconsistent with the identification type in step S63, the classification decision surface of different types of intrusion events in the new environment is corrected using the feature class to be identified and the real type of the intrusion event to be identified.

[0012] The present invention also provides a pipeline safety early warning system for pipeline safety early warning in a new environment, the system comprising an identification module for executing the above pipeline safety early warning event type identification optimization method.

[0013] Compared with the prior art, the pipeline safety warning event type identification optimization method and pipeline safety warning system of the present invention take the characteristic centers and classification decision surfaces of different types of intrusion events in the original environment as the basis, compare the characteristic centers of the same type of events in the new environment and the original environment, evaluate the propagation differences of the same vibration source events in different soil environments, and adjust the recognition model in the new environment based on the original recognition model, thereby realizing the adaptive optimization of the recognition model for the new soil environment, ensuring the high accuracy of event type recognition in the new environment, and having the advantages of simple operation and low cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0015] Figure 1 It is a flow chart of the pipeline safety early warning event type identification optimization method of the present invention.

[0016] Figure 2 It is a schematic diagram of the distribution of features extracted from vibration signals of different types of intrusion events in the original environment in an embodiment of the present invention and the classification decision surface.

[0017] Figure 3 The vibration signal of the intrusion event is tested in a new environment according to an embodiment of the present invention. Figure 2 Schematic diagram of the distribution of features extracted by the same feature extraction method in the feature space.

[0018] Figure 4 This is a schematic diagram of the feature centers of different types of intrusion events in the new environment obtained by shifting the feature centers of different types of intrusion events in the original environment based on the offset of the feature centers of the intrusion events tested in the new environment relative to the feature centers of the same intrusion events in the original environment.

[0019] Figure 5 It is the classification decision surface of different types of intrusion events in the new environment obtained based on the feature centers of different types of intrusion events in the new environment. DETAILED DESCRIPTION

[0020] In order to provide a further understanding of the purpose, structure, features, and functions of the present invention, the following detailed description is given in conjunction with the embodiments.

[0021] See also Figure 1 The present invention provides a pipeline safety early warning event type identification optimization method, which is used for pipeline safety early warning event type identification in a new environment. The method comprises the following steps: S1. Acquire data of different types of intrusion events in the original environment, extract features from the acquired data, and determine feature centers of different types of intrusion events in the original environment based on the extracted features; S2. Acquire data of the test intrusion event in the new environment, extract features from the acquired data according to the same feature extraction method as in step S1, and determine the feature center of the test intrusion event in the new environment based on the extracted features; S3. Calculate the offset of the feature center of the test intrusion event in the new environment relative to the feature center of the corresponding intrusion event in the different types of intrusion events in the original environment. The offset here is a vector with both magnitude and direction. S4. The feature centers of different types of intrusion events in the original environment are shifted according to the offset amount to obtain the feature centers of different types of intrusion events in the new environment; S5. Obtain the classification decision surface of different types of intrusion events in the new environment according to the feature centers of different types of intrusion events in the new environment; S6. Use the classification decision surface of different types of intrusion events in the new environment to identify event types.

[0022] Preferably, the test intrusion event is at least one of the different types of intrusion events to ensure that the feature center corresponding to the test intrusion event exists in the feature center of the different types of intrusion events in the original environment, so that the offset of the test intrusion event in the new environment relative to the feature center in the original environment can be calculated. Different types of intrusion events are several of all possible pipeline intrusion events. The number of event types contained in different types of intrusion events determines how many classification problems the recognition optimization method involves and how many types of events the recognition optimization method ultimately recognizes. For example, if the number of event types contained in different types of intrusion events is 3, then the type of event ultimately recognized by the recognition optimization method is 3; if the number of event types contained in different types of intrusion events is 4, then the type of event ultimately recognized by the recognition optimization method is 4.

[0023] Furthermore, different types of intrusion events include common and easy-to-operate manual excavation events, vehicle traffic events, and mechanical construction events.

[0024] Further, the test intrusion event is at least one of a manual excavation event, a vehicle passage event, and a mechanical construction event, so as to facilitate the experimental operation. Further, the test intrusion event is a manual excavation event, so as to reduce the risk of destructive experiments in a new environment and reduce the experimental cost.

[0025] Preferably, step S1 comprises the following steps: S11. Acquire data of different types of intrusion events in the original environment, where the data of different types of intrusion events include vibration signals corresponding to different types of intrusion events and real intrusion event types; S12. Extracting features from the vibration signal to obtain a first feature class; S13. Obtaining a first sample based on the first feature class and the corresponding real intrusion event type; S14. Loop through step S11 to step S13 until a number of first samples are obtained, and then terminate the loop to obtain a training set based on the first samples; S15. Use the training set to train the classification model to obtain the feature centers and classification decision surfaces of different types of intrusion events in the original environment.

[0026] In order to ensure the accuracy of event type recognition in the original environment, and then ensure the accuracy of event type recognition in the new environment, the following steps are also included after step S15: S16. Use the classification decision surface of different types of intrusion events in the original environment to identify event types; S17. If the accuracy of event type identification using the classification decision surface of different types of intrusion events in the original environment is greater than the set threshold, proceed to step S2; otherwise, proceed to step S11.

[0027] In actual use, the pipeline safety warning event type recognition model in the original environment, which serves as a relatively basic one, usually undergoes a lot of training and has a sufficiently high recognition accuracy for event type recognition.

[0028] Preferably, step S2 comprises the following steps: S21. Acquire test intrusion event data in the new environment, the test intrusion event data includes the vibration signal corresponding to the test intrusion event and the real intrusion event type; S22. Extract features from the vibration signal using the same feature extraction method as in step S12 to obtain a second feature class; S23. Obtaining a second sample based on the second feature class and the corresponding real intrusion event type; S24. Looping through step S21 to step S23 until a number of second samples are obtained, and then terminating the loop, obtaining a second training set based on the number of second samples; S25. Input the second training set into the classification model trained in the original environment to obtain the feature center of the test intrusion event in the new environment.

[0029] Preferably, step S6 comprises the following steps: S61. Real-time acquisition of data on intrusion events to be identified in the new environment, the data on intrusion events to be identified include vibration signals corresponding to the intrusion events to be identified; S62. Extract features from the vibration signal using the same feature extraction method as in step S12 to obtain feature classes to be identified; S63. Determine the event type of the intrusion event to be identified based on the feature class to be identified and the classification decision surface of different types of intrusion events in the new environment.

[0030] For the purpose of reducing the experimental cost, the number of samples in the second training set in step S2 is usually not large, and the feature center of the test intrusion event in the new environment may not be accurate. The classification decision surface of different types of intrusion events in the new environment may be biased. In order to improve the accuracy of pipeline safety warning event type recognition in the new environment, the following steps are also included after step S63: S64. If the real type of the intrusion event to be identified is inconsistent with the identification type in step S63, the classification decision surface of different types of intrusion events in the new environment is corrected using the feature class to be identified and the real type of the intrusion event to be identified.

[0031] In actual use, the first feature class, the second feature class, and the feature class to be identified may include at least two of spatial features, temporal features, intensity features, and frequency features; the classification model may be a neural network model, a support vector machine, and the like.

[0032] Another aspect of the present invention is a pipeline safety early warning system for pipeline safety early warning in a new environment. The system includes an identification module, and the identification module is used to execute the above pipeline safety early warning event type identification optimization method.

[0033] Example 1: The original environment is a soil geological environment with high soil moisture content such as the Jiangnan water network in the middle and lower reaches of the Yangtze River in my country. A pipeline safety early warning system is deployed in this environmental area, and various event type tests such as manual excavation, vehicle rolling and mechanical construction are carried out at a certain position along the sensor optical cable. The system automatically records and stores soil vibration signals caused by various events.

[0034] Based on the classification model within the system, the characteristics, distribution and event center of each event are extracted, and the position of each center in the multidimensional feature space is recorded. Figure 2 , Figure 2 This is a schematic diagram of the distribution of features extracted from vibration signals of different types of intrusion events in the original environment of this embodiment in the feature space and the classification decision surface. The two dimensions of feature j and feature m in the multidimensional feature are selected to display the feature distribution and event classification decision boundary. The multidimensional feature can be spatial feature, time feature, intensity feature, frequency feature, etc.

[0035] When the system is deployed on the Loess Plateau in northwest my country, a simple, easy-to-implement, and low-cost artificial excavation event is selected for testing in view of a new typical soil type, soil hardness, moisture content, soil layer structure, and stress state. The system automatically records and stores soil vibration signals caused by artificial excavation events, extracts the characteristics, distribution, and event center of artificial excavation events based on the classification model inside the system, and evaluates the signal propagation characteristics of the new soil address environment based on the new distribution of event centers in the feature space. See Figure 3 , Figure 3 The vibration signal of the intrusion event is tested in the new environment of this embodiment according to Figure 2 Schematic diagram of the distribution of features extracted by the same feature extraction method in the feature space. Figure 3 As shown, it can be found that the feature distribution of the manual mining events in the new environment represented by the triangle has shifted to a certain extent compared with the feature distribution of the manual mining events in the original environment, which affects the classification decision surface of the manual mining events.

[0036] The feature center of the manually mined event in the new environment is re-evaluated by using the feature of the manually mined event in the new environment, and the offset of the feature center of the manually mined event in the new environment relative to the feature center of the manually mined event in the original environment is calculated, such as Figure 4 As shown, the characteristic center position of the artificial excavation event shifts to the upper right, thereby obtaining the quantitative change in the vibration signal propagation characteristics of the new geological environment relative to the original environment.

[0037] The change in the characteristic center of the artificial excavation event is transferred to the characteristic center of other types of events, that is, the characteristic centers of other intrusion events in the original environment are shifted according to the offset of the characteristic center of the artificial excavation event, and the characteristic center positions of other types of events in the new geological environment are obtained. Figure 4 The characteristic centers of each invasion event under the original environment are marked as 1, and the characteristic centers of each invasion event under the new geological environment are marked as 2.

[0038] Based on the characteristic center of each intrusion event under the new geological environment, the classification decision surface of each event is re-determined, such as Figure 5 As shown by the middle dotted line, the adaptability optimization of the pipeline safety early warning system to the new soil geological environment is completed.

[0039] Example 2: The original environment is a test site for development and research. A pipeline safety early warning system is deployed in the test site area. Different soil geological environments such as sandy soil and clay soil are designed in sections along the sensor optical cable. Various event types such as manual excavation, vehicle rolling and mechanical construction are tested at a certain location. The system automatically records and stores soil vibration signals caused by various events.

[0040] The typical geological environment data obtained from the experimental field is used as the basis for system identification. When the pipeline safety early warning system is deployed in a real production environment, a small number of sample simulation experiments are carried out in the production environment to calculate the offset relative to the experimental field area in the production environment, and the corresponding event center and classification decision surface are adaptively adjusted.

[0041] The pipeline safety early warning event type identification optimization method and pipeline safety early warning system of the present invention take the characteristic centers and classification decision surfaces of different types of intrusion events in the original environment as the basis, compare the characteristic centers of the same type of events in the new environment and the original environment, evaluate the propagation differences of the same vibration source events in different soil environments, and adjust the identification model in the new environment based on the original identification model, thereby realizing the adaptive optimization of the identification model for the new soil environment, ensuring the high accuracy of event type identification in the new environment, and having the advantages of simple operation and low cost.

[0042] The present invention has been described by the above-mentioned relevant embodiments, but the above-mentioned embodiments are only examples for implementing the present invention. In addition, the technical features involved in the different embodiments of the present invention described above can be combined with each other as long as they do not conflict with each other. It must be pointed out that the disclosed embodiments do not limit the scope of the present invention. On the contrary, changes and modifications made without departing from the spirit and scope of the present invention are all within the scope of patent protection of the present invention.

Claims

1. A pipeline safety early warning event type identification optimization method, used for pipeline safety early warning event type identification under new environment, characterized in that: The method comprises the following steps: S1. Acquire data of different types of intrusion events in the original environment, extract features from the acquired data, and determine feature centers of different types of intrusion events in the original environment based on the extracted features; S2. Acquire data of the test intrusion event in the new environment, extract features from the acquired data according to the same feature extraction method as in step S1, and determine the feature center of the test intrusion event in the new environment based on the extracted features; S3. Calculate the offset of the feature center of the test intrusion event in the new environment relative to the feature center of the corresponding intrusion event in the different types of intrusion events in the original environment; S4. The feature centers of different types of intrusion events in the original environment are shifted according to the offset amount to obtain the feature centers of different types of intrusion events in the new environment; S5. Obtain the classification decision surface of different types of intrusion events in the new environment according to the feature centers of different types of intrusion events in the new environment; S6. Use the classification decision surface of different types of intrusion events in the new environment to identify event types.

2. The pipeline safety early warning event type identification and optimization method according to claim 1, characterized in that: The test intrusion event is at least one of the different types of intrusion events.

3. The pipeline safety early warning event type identification optimization method according to claim 2, characterized in that: The different types of intrusion events include manual excavation events, vehicle traffic events and mechanical construction events.

4. The pipeline safety early warning event type identification and optimization method according to claim 3, characterized in that: The test intrusion event is at least one of a manual excavation event, a vehicle passage event, and a mechanical construction event.

5. The pipeline safety early warning event type identification and optimization method according to claim 1, characterized in that: Step S1 includes the following steps: S11. Acquire data of different types of intrusion events in the original environment, wherein the data of different types of intrusion events include vibration signals corresponding to different types of intrusion events and real intrusion event types; S12. Extracting features from the vibration signal to obtain a first feature class; S13. Obtaining a first sample based on the first feature class and the corresponding real intrusion event type; S14. Looping through step S11 to step S13 until a number of first samples are obtained, and then terminating the loop to obtain a first training set based on the number of first samples; S15. Use the first training set to train the classification model to obtain the feature centers and classification decision surfaces of different types of intrusion events in the original environment.

6. The pipeline safety early warning event type identification and optimization method according to claim 5, characterized in that: After step S15, the following steps are also included: S16. Use the classification decision surface of different types of intrusion events in the original environment to identify event types; S17. If the accuracy of event type identification using the classification decision surface of different types of intrusion events in the original environment is greater than the set threshold, proceed to step S2; otherwise, proceed to step S11.

7. The pipeline safety early warning event type identification and optimization method according to claim 5, characterized in that: Step S2 includes the following steps: S21. Acquire test intrusion event data in the new environment, wherein the test intrusion event data includes a vibration signal corresponding to the test intrusion event and a real intrusion event type; S22. Extract features from the vibration signal using the same feature extraction method as in step S12 to obtain a second feature class; S23. Obtaining a second sample based on the second feature class and the corresponding real intrusion event type; S24. Looping through step S21 to step S23 until a number of second samples are obtained, and then terminating the loop, obtaining a second training set based on the number of second samples; S25. Input the second training set into the classification model trained in the original environment to obtain the feature center of the test intrusion event in the new environment.

8. The pipeline safety early warning event type identification and optimization method according to claim 5, characterized in that: Step S6 includes the following steps: S61. Real-time acquisition of data of intrusion events to be identified in the new environment, wherein the data of intrusion events to be identified include vibration signals corresponding to the intrusion events to be identified; S62. Extract features from the vibration signal using the same feature extraction method as in step S12 to obtain feature classes to be identified; S63. Determine the event type of the intrusion event to be identified based on the feature class to be identified and the classification decision surface of different types of intrusion events in the new environment.

9. The pipeline safety early warning event type identification and optimization method according to claim 8, characterized in that: The following steps are also included after step S63: S64. If the real type of the intrusion event to be identified is inconsistent with the identification type in step S63, the classification decision surface of different types of intrusion events in the new environment is corrected using the feature class to be identified and the real type of the intrusion event to be identified.

10. A pipeline safety early warning system, used for pipeline safety early warning in a new environment, characterized in that: The system comprises: An identification module is used to execute the pipeline safety warning event type identification optimization method described in any one of claims 1 to 9.

Images