A buried gas pipeline third-party construction damage edge early warning model

Abstract

The application discloses a buried gas pipeline third-party construction damage edge early warning model and relates to the technical field of pipeline monitoring. The model comprises a signal acquisition module, an acoustic emission sensor senses a third-party construction signal, and a data transmission device transmits the collected construction signal to a signal processing module. The signal processing module processes the construction signal uploaded by the signal acquisition module, and the signal processing includes windowing operation, band-pass filtering and feature extraction. A comparison module regularly evaluates the background noise level, compares it with the noise level and calculates a characteristic factor alpha. A calculation module determines the type of the acquired construction signal by calculating the characteristic factor threshold and the total threshold. An early warning module acquires the type of the construction signal determined by the calculation module and sends early warning information to the corresponding unit. The application screens and extracts the characteristics of the construction signal from multiple frequency bands and optimizes and selects the characteristic factor threshold and the total threshold, thereby improving the identification ability of the third-party construction damage behavior.

Description

Technical Field

[0001] This invention relates to the field of pipeline monitoring technology, specifically to an early warning model for the edge of damage caused by third-party construction of buried gas pipelines. Background Technology

[0002] In the past five years, third-party construction damage has accounted for an average of 82% of gas pipeline leakage accidents in my country, making it the primary cause of underground gas pipeline damage over the past 20 years. To reduce the damage caused by third-party construction activities to urban underground gas pipelines, this invention proposes an edge warning model for third-party construction damage to urban buried gas pipelines, based on an underground space pipeline construction monitoring technology solution and oriented towards front-end sensing equipment. The underground space pipeline construction monitoring technology solution involves fixing acoustic emission sensors to the pipeline. The sensor installation location can be in pipeline ancillary facilities such as gas valve wells or gas pressure regulating cabinets. When the pipeline is affected by construction, it will vibrate. After receiving the vibration signal, the sensor performs ADC sampling, signal amplification, and filtering before sending the data to the edge computing module. The model calculates and determines the signal type. For construction signals, the data segment is saved, and the warning information is transmitted to the cloud server via a 5G module and external antenna. For suspected construction signals, the data segment is saved, and the warning information is transmitted to the cloud server via a 5G module and external antenna for further identification and judgment. For non-construction signals, the equipment will not take any action. In addition, when the equipment detects an early warning, it will reduce its sleep time, extend its monitoring time, and increase the frequency of monitoring for construction damage.

[0003] Existing signal recognition and early warning methods can be broadly categorized into two types: one is classification models based on machine learning and deep learning algorithms such as SVM, neural networks, decision trees, and image recognition; the other is classification models based on setting multiple monitoring indicator thresholds and rules. These early warning technologies are mostly applied to third-party construction monitoring technologies using different sensing principles, such as distributed fiber optic acoustic sensing technology, fiber optic vibration detection technology, soil vibration monitoring technology based on accelerometers, video surveillance technology, and GPS positioning. However, these monitoring technologies are mostly used in long-distance oil and gas pipelines, with limited application in urban underground pipe networks. The most important reason for this is the impact of complex urban environmental noise. Frequent and diverse urban construction activities, complex traffic systems, and noise generated by human production and daily life all severely affect the identification of third-party construction signals. Furthermore, previous early warning technologies primarily relied on cloud servers for identification and judgment, typically extracting features from the signal in the time domain, frequency domain, and even images, and then using machine learning and deep learning classification models to establish corresponding early warning models. This approach usually does not require consideration of computing power and power consumption. On the equipment side, thresholds are usually set for one or more monitoring indicators. When a monitoring indicator reaches the threshold, it is identified as an abnormal construction signal. However, this method is affected by computing power and power consumption, and the number of monitoring indicators is usually very small. This may lead to insufficient prediction of abnormal construction signals, resulting in false alarms and missed alarms. Especially in urban areas with multiple noise sources, the false alarm rate and missed alarm rate are further increased. Summary of the Invention

[0004] The purpose of this invention is to provide an early warning model for third-party construction damage to buried gas pipelines. This model filters and extracts construction signal features from multiple frequency bands and optimizes the selection of feature factor thresholds and total thresholds, thereby improving the ability to identify third-party construction damage and solving the technical problem of false alarms and missed alarms caused by insufficient prediction of abnormal construction signals in existing methods.

[0005] The objective of this invention can be achieved through the following technical solutions: A third-party construction damage edge early warning model for buried gas pipelines includes: The signal acquisition module uses an acoustic emission sensor to detect third-party construction signals and transmits the collected construction signals to the signal processing module via a data transmission device. The signal processing module is used to process the construction signals uploaded by the signal acquisition module. Signal processing includes windowing, bandpass filtering, and feature extraction. The comparison module periodically evaluates the background noise level, compares it with the noise level, calculates the feature factor α, and then performs threshold and total threshold judgments on α to determine the signal type. The calculation module determines the type of construction signal acquired by calculating the feature factor threshold and the total threshold. The early warning module acquires the construction signal type determined by the calculation module and sends early warning information to the corresponding unit.

[0006] As a further aspect of the present invention: the windowing operation truncates the original signal as needed to maintain a unified data structure feature. Specifically, it uses a rectangular window to process the signal in stages. The rectangular window size is divided into two types: a narrow window of 0.2s and a wide window of 1s. The overlap area between two adjacent rectangular windows is 50%.

[0007] As a further aspect of the present invention: the bandpass filtering uses a Chebyshev II type bandpass filter to filter the signal. The filtering frequency band is divided into low-frequency bandpass filtered signal with a frequency range of 1Hz-1.5kHz, mid-frequency bandpass filtered signal with a frequency range of 0.75kHz-2.25kHz, and high-frequency bandpass filtered signal with a frequency range of 1.5kHz-3kHz, based on the frequency distribution characteristics of the construction signal.

[0008] As a further aspect of the present invention: the feature extraction is used to extract and filter feature values ​​that are highly correlated with the signal type from the time domain and frequency domain respectively, and to perform feature extraction on the three frequency band signals after bandpass filtering, wherein the feature extraction range of each signal is the signal after windowing operation.

[0009] As a further aspect of the present invention: the specific features extracted are as follows: Low frequency band: peak value A 11 Peak factor A 12 Center frequency A 13 Signal-to-noise ratio A 14 ; Mid-frequency band: Crest factor A 21 Power bandwidth A 22 Center frequency A 23 Signal-to-noise ratio A 24 ; High frequency band: power bandwidth A 31 Center frequency A 32 .

[0010] As a further aspect of the present invention: the method of periodically evaluating the background noise level, comparing it with the noise level, calculating the feature factor α, and then determining the signal type by thresholding α and the total threshold is as follows: The signal acquisition module periodically acquires background noise signals during four time periods: morning peak (7:00-10:00), noon peak (11:00-14:00), evening peak (17:00-19:00), and midnight (00:00). Features are extracted from the acquired background noise signals during these time periods using a feature extraction method to obtain the specific feature data for each time period. Low frequency band: peak value a 11 Peak factor a 12 Center frequency a 13 Signal-to-noise ratio (a) 14 ; Mid-frequency band: Crest factor a 21 Power bandwidth a 22 Center frequency a 23 Signal-to-noise ratio (a) 24 ; High frequency band: power bandwidth a 31 Center frequency a 32 ; A 11 / a 11 =α 11 A 21 / a 21 =α 21 ...and so on, we can obtain a new eigenvector T={α} 11 α 12 , ..., α 31 α 32 If each feature factor α in the feature vector T ij A value of ≈1 indicates that there is no construction activity.

[0011] As a further aspect of the present invention: the calculation method for the feature factor threshold and the total threshold is specifically as follows: First, determine the weight of each feature factor. , It can be obtained from the normalized result of the correlation coefficient of each feature value, and the specific formula is as follows: (1) In the above formula (1), This represents the weight of the j-th feature factor in frequency band i, where i = 1, 2, and 3 represent low, medium, and high frequencies, respectively. Represents the correlation coefficient of the j-th feature in frequency band i; The number of features in frequency band i; feature factor The range of values ​​is 1≤ ≤ ,set up Within this range, it is considered a construction signal; outside this range, it is considered not a construction signal. Therefore, use... To represent satisfying the feature factor threshold The state is given by the following formula: (2) In equation (2) above, the subscript k represents the number of signals to be calculated; using and The product of these factors serves as the validity of the feature factor, and summing over all features yields the result. ,when A signal greater than the total threshold θ is considered a construction signal; otherwise, it is considered another signal. The specific formula is as follows: , (3) In the above formula (3), The result indicates the judgment: 1 represents a construction signal, and 0 represents a non-construction signal. After determining the calculation formula, the values ​​of the feature factor threshold and the total threshold are calculated by substituting several collected construction signal and noise samples.

[0012] As a further aspect of the present invention: in the process of calculating the feature factor threshold and the total threshold, an optimization algorithm is used to optimize the total threshold θ and the feature factor threshold. The optimization objective function is as follows: (4) In the formula, False alarm rate is calculated based on the true category and the predicted category; σ is the false alarm rate; σ is the balance factor between the false alarm rate and the false alarm rate, which is set to 0.7.

[0013] As a further aspect of the present invention: after being identified as a construction signal, when it is identified as a construction signal 5-20 times consecutively within 1 second, i.e. at a frequency of 5-20Hz, it can be considered as a road breaking construction signal.

[0014] As a further aspect of the present invention: before the early warning module sends the early warning information to the corresponding unit, it sends the original signal data and judgment result to the cloud server for further confirmation, and sends the early warning information after confirmation.

[0015] The beneficial effects of this invention are: (1) The proposed multi-band construction signal feature extraction method can still achieve feature extraction and feature fusion using other frequency band signals when some frequency band signals are submerged by noise, providing more possibilities for construction signal recognition; (2) By filtering the signal features, the computational load in the feature extraction process is reduced.

[0016] (3) A periodic background noise level estimation strategy is proposed, which ensures the accuracy of construction signal detection by calculating the characteristic factor α.

[0017] (4) The method for determining the weights and thresholds of feature factors is more scientific and reasonable, and is updated and iterated remotely by the background program. As the amount of data increases, the accuracy of the model will be further improved.

[0018] (5) A peak detection module was proposed, which improved the detection capability of construction signals. Attached Figure Description

[0019] The invention will now be further described with reference to the accompanying drawings.

[0020] Figure 1 This is a structural block diagram of the edge warning model of the present invention; Figure 2 This is a flowchart of the early warning process of the edge early warning model of this invention. Detailed Implementation

[0021] 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 only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please see Figures 1-2 As shown, this invention is an early warning model for the edge of damage caused by third-party construction of buried gas pipelines, comprising: The signal acquisition module uses an acoustic emission sensor to detect third-party construction signals and transmits the collected construction signals to the signal processing module via a data transmission device. Specifically, the collected data is processed, including ADC sampling, signal amplification, and low-pass filtering (1Hz-3kHz).

[0023] Furthermore, the ADC sampling frequency is 20kHz, the signal amplification is 75 times, the low-pass filter is a fourth-order Butterworth low-pass filter with a cutoff frequency of 3kHz, a passband gain of 0dB, and a stopband attenuation of 80dB.

[0024] The signal processing module is used to process the construction signals uploaded by the signal acquisition module. Signal processing includes windowing, bandpass filtering, and feature extraction. The comparison module periodically evaluates the background noise level, compares it with the noise level, calculates the feature factor α, and then performs threshold and total threshold judgments on α to determine the signal type. The calculation module determines the type of construction signal acquired by calculating the feature factor threshold and the total threshold. The early warning module acquires the construction signal type determined by the calculation module and sends early warning information to the corresponding unit.

[0025] The successful application of the edge warning model of this invention is beneficial for effective early warning of third-party construction damage during the maintenance and upgrading of underground pipelines such as municipal road construction, gas, and water supply, including construction activities such as excavation, drilling, and piling. It can also monitor and warn of ongoing construction activities that may damage the pipeline itself.

[0026] The windowing operation truncates the original signal as needed while maintaining a unified data structure. Specifically, it uses a rectangular window to process the signal in stages. The rectangular window size is divided into two types: a narrow window of 0.2s and a wide window of 1s. The overlap area between two adjacent rectangular windows is 50%.

[0027] The bandpass filtering uses a Chebyshev II bandpass filter to filter the signal. The filtering frequency bands are divided into low-frequency bandpass filtered signals with a frequency range of 1Hz-1.5kHz, mid-frequency bandpass filtered signals with a frequency range of 0.75kHz-2.25kHz, and high-frequency bandpass filtered signals with a frequency range of 1.5kHz-3kHz, based on the frequency distribution characteristics of the construction signal.

[0028] The feature extraction is used to extract and filter feature values ​​that are highly correlated with the signal type from the time domain and frequency domain respectively. Feature extraction is performed on the three frequency band signals after bandpass filtering. The feature extraction range of each signal is the signal after windowing operation, that is, the signal within the rectangular window after the signal is truncated by a rectangular window.

[0029] The specific features extracted are: Low frequency band: peak value A 11 Peak factor A 12 Center frequency A 13 Signal-to-noise ratio A 14 ; Mid-frequency band: Crest factor A 21 Power bandwidth A 22 Center frequency A 23 Signal-to-noise ratio A 24 ; High frequency band: power bandwidth A 31 Center frequency A 32 .

[0030] Specifically, the selection criteria are based on the correlation between the above features and construction signals and noise signals. Features with a correlation coefficient greater than 0.6 are selected as the final features. The purpose of feature selection is to reduce the amount of computation and improve the battery life.

[0031] The method for periodically assessing the background noise level, comparing it with the noise level, calculating the feature factor α, and then applying threshold and total threshold judgments to α to determine the signal type is as follows: The signal acquisition module periodically acquires background noise signals during four time periods: morning peak (7:00-10:00), noon peak (11:00-14:00), evening peak (17:00-19:00), and midnight (00:00). The signal acquisition time is 5 minutes. Feature extraction is then used to extract features from the acquired background noise signals within these time periods, yielding the specific feature data for each time period: Low frequency band: peak value a 11 Peak factor a 12 Center frequency a 13 Signal-to-noise ratio (a) 14 ; Mid-frequency band: Crest factor a 21 Power bandwidth a 22 Center frequency a 23 Signal-to-noise ratio (a) 24 ; High frequency band: power bandwidth a 31 Center frequency a 32 ; A 11 / a 11 =α 11 A 21 / a 21 =α 21 ...and so on, we can obtain a new eigenvector T={α} 11 α 12 , ..., α 31 α 32 If each feature factor α in the feature vector T ij A value of ≈1 indicates that there is no construction activity.

[0032] The specific methods for calculating the feature factor threshold and the total threshold are as follows: Each feature factor contributes differently to the result judgment; some features are relatively important, while others are relatively unimportant. Therefore, it is necessary to first determine the weight of each feature factor. , It can be obtained from the normalized result of the correlation coefficient of each feature value, and the specific formula is as follows: (1) In the above formula (1), This represents the weight of the j-th feature factor in frequency band i, where i = 1, 2, and 3 represent low, medium, and high frequencies, respectively. Represents the correlation coefficient of the j-th feature in frequency band i; The number of features in frequency band i; feature factor The range of values ​​is 1≤ ≤ ,set up Within this range, it is considered a construction signal; outside this range, it is considered not a construction signal. Therefore, use... To represent satisfying the feature factor threshold The state is given by the following formula: (2) In equation (2) above, the subscript k represents the number of signals to be calculated; using and The product of these factors serves as the validity of the feature factor, and summing over all features yields the result. ,when A signal greater than the total threshold θ is considered a construction signal; otherwise, it is considered another signal. The specific formula is as follows: , (3) In the above formula (3), The result indicates the judgment: 1 represents a construction signal, and 0 represents a non-construction signal. After determining the calculation formula, the values ​​of the feature factor threshold and the total threshold are calculated by substituting several collected construction signal and noise samples.

[0033] In the calculation of the feature factor threshold and the total threshold, an optimization algorithm is used to optimize the total threshold θ and the feature factor threshold. The optimization objective function is as follows: (4) In the formula, False alarm rate is calculated based on the true category and the predicted category; σ is the false alarm rate; σ is the balance factor between the false alarm rate and the false alarm rate, which is set to 0.7.

[0034] Specifically, through optimization using 301 construction signal segments and 280 non-construction signals, and 655 iterations, the objective function Objective = 0.2178 was finally obtained. ={3.5, 2.1, 2.81, 3.48, 2.35, 1.48, 5.14, 6.47, 1.68, 1.45}, θ=1.28. That is, when the characteristic factors of the acquired signal... Substituting into equations (2) and (3), when If the value is greater than 1.28, then the signal type is a construction signal; otherwise, it is a non-construction signal.

[0035] The above construction signal identification model mainly targets the identification of single impact elastic waves generated during construction on the ground or pipelines. For multiple impact elastic waves, such as those generated by a hydraulic breaker striking the road surface, a striking frequency of 5 to 20 times per second can occur. Therefore, for the latter, to more accurately determine the construction behavior, it is necessary to judge the frequency of the generated impact elastic waves to improve the accuracy of the determination of construction damage behavior. In this case, a rectangular wide window in the windowing operation is used to truncate the signal, with the window overlap area set to 0. After identifying it as a construction signal, if it is continuously identified as a construction signal 5-20 times within 1 second, i.e., the frequency is 5-20Hz, it can be considered a road breaking construction signal.

[0036] Before sending an early warning message to the corresponding unit, the early warning module sends the original signal data and judgment result to the cloud server for further confirmation. After confirmation, the early warning message is sent.

[0037] The working principle of this invention is as follows: Signal acquisition: Construction signals are acquired, specifically through acoustic emission sensors sensing third-party construction signals; Windowing operation is performed on the acquired signals: The original signals are truncated as needed to maintain a uniform data structure; Bandpass filtering is performed on the windowed signals: Bandpass filtering is performed on signals of different frequency bands. On the one hand, higher-frequency noise sources are filtered out; on the other hand, frequency band filtering can prevent the construction signal frequencies from being submerged by other noise sources, resulting in the inability to extract effective features or insignificant features. That is, the construction signal is actually a wide-band distribution, meaning it has a large bandwidth. When some frequencies are contaminated by noise, it can be judged by extracting features from other uncontaminated frequency bands; After the signal is bandpass filtered, feature values ​​are obtained through feature extraction: from the time domain... Feature values ​​with high correlation to signal type are extracted and filtered in both the frequency domain and the background noise level estimation: the background noise level is evaluated periodically for signal type comparison and feature factor α is calculated; weights and thresholds are used to determine the feature factor threshold and total threshold, and the construction type to which the signal belongs is determined by comparing the relationship between the feature factor and the threshold; peak detection is used to further confirm the construction activity behavior, mainly because the characteristics of broken construction signals are very typical, and the construction activity behavior can be determined by determining the impact pulse frequency. This step is mainly to improve the accuracy of identification. Warning information is issued. When the signal is determined to be a construction signal, the original signal data and the judgment result are sent to the cloud server for further confirmation. Finally, a warning information is issued to the gas pipeline ownership unit for on-site supervision and management.

[0038] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A pre-warning model for the edge of damage during third-party construction of buried gas pipelines, characterized in that, include: The signal acquisition module uses an acoustic emission sensor to detect third-party construction signals and transmits the collected construction signals to the signal processing module via a data transmission device. The signal processing module is used to process the construction signals uploaded by the signal acquisition module. Signal processing includes windowing, bandpass filtering, and feature extraction. The comparison module periodically evaluates the background noise level, compares it with the noise level, calculates the feature factor α, and then performs threshold and total threshold judgments on α to determine the signal type. The calculation module determines the type of construction signal acquired by calculating the feature factor threshold and the total threshold. The early warning module acquires the construction signal type determined by the calculation module and sends early warning information to the corresponding unit. The specific methods for calculating the feature factor threshold and the total threshold are as follows: First, determine the weight of each feature factor. , It can be obtained from the normalized result of the correlation coefficient of each feature value, and the specific formula is as follows: （1） In the above formula (1), This represents the weight of the j-th feature factor in frequency band i, where i = 1, 2, and 3 represent low, medium, and high frequencies, respectively. Represents the correlation coefficient of the j-th feature in frequency band i; Indicates the number of features in frequency band i; Feature factors The range of values ​​is 1≤ ≤ ,set up Within this range, it is considered a construction signal; outside this range, it is considered not a construction signal. Therefore, use... To represent satisfying the feature factor threshold The state is given by the following formula: （2） In equation (2) above, the subscript k represents the number of signals to be calculated; using and The product of these factors serves as the validity of the feature factor, and summing over all features yields the result. ,when A signal greater than the total threshold θ is considered a construction signal; otherwise, it is considered another signal. The specific formula is as follows: , （3） In the above formula (3), The result indicates the judgment: 1 represents a construction signal, and 0 represents a non-construction signal. After determining the calculation formula, the values ​​of the feature factor threshold and the total threshold are calculated by substituting several collected construction signal and noise samples.

2. The early warning model for the edge of third-party construction damage to buried gas pipelines according to claim 1, characterized in that, The windowing operation truncates the original signal as needed while maintaining a unified data structure. Specifically, it uses a rectangular window to process the signal in stages. The rectangular window size is divided into two types: a narrow window of 0.2s and a wide window of 1s. The overlap area between two adjacent rectangular windows is 50%.

3. The early warning model for the edge of third-party construction damage to buried gas pipelines according to claim 1, characterized in that, The bandpass filtering uses a Chebyshev II bandpass filter to filter the signal. The filtering frequency bands are divided into low-frequency bandpass filtered signals with a frequency range of 1Hz-1.5kHz, mid-frequency bandpass filtered signals with a frequency range of 0.75kHz-2.25kHz, and high-frequency bandpass filtered signals with a frequency range of 1.5kHz-3kHz, based on the frequency distribution characteristics of the construction signal.

4. The early warning model for the edge of third-party construction damage to buried gas pipelines according to claim 3, characterized in that, The feature extraction is used to extract and filter feature values ​​that are highly correlated with the signal type from the time domain and frequency domain, respectively. Feature extraction is performed on the three frequency band signals after bandpass filtering, and the feature extraction range of each signal is the signal after windowing operation.

5. The early warning model for the edge of third-party construction damage to buried gas pipelines according to claim 4, characterized in that, The specific features extracted are: Low frequency band: peak value A 11 Peak factor A 12 Center frequency A 13 Signal-to-noise ratio A 14 ; Mid-frequency band: Crest factor A 21 Power bandwidth A 22 Center frequency A 23 Signal-to-noise ratio A 24 ; High frequency band: power bandwidth A 31 Center frequency A 32 .

6. The early warning model for the edge of third-party construction damage to buried gas pipelines according to claim 5, characterized in that, The method for periodically assessing the background noise level, comparing it with the noise level, calculating the feature factor α, and then applying threshold and total threshold judgments to α to determine the signal type is as follows: The signal acquisition module periodically acquires background noise signals during four time periods: morning peak (7:00-10:00), noon peak (11:00-14:00), evening peak (17:00-19:00), and midnight (00:00). Features are extracted from the acquired background noise signals during these time periods using a feature extraction method to obtain the specific feature data for each time period. Low frequency band: peak value a 11 Peak factor a 12 Center frequency a 13 Signal-to-noise ratio (a) 14 ; Mid-frequency band: Crest factor a 21 Power bandwidth a 22 Center frequency a 23 Signal-to-noise ratio (a) 24 ; High frequency band: power bandwidth a 31 Center frequency a 32 ; A 11 / a 11 =α 11 A 21 / a 21 =α 21 ...and so on, we can obtain a new eigenvector T={α} 11 α 12 , ..., α 31 α 32 If each feature factor α in the feature vector T ij A value of ≈1 indicates that there is no construction activity.

7. The early warning model for the edge of third-party construction damage to buried gas pipelines according to claim 1, characterized in that, In the calculation of the feature factor threshold and the total threshold, an optimization algorithm is used to optimize the total threshold θ and the feature factor threshold. The optimization objective function is as follows: （4） In equation (4), False alarm rate is calculated based on the true category and the predicted category; σ is the false alarm rate; σ is the balance factor between the false alarm rate and the false alarm rate, which is set to 0.

7.

8. The early warning model for the edge of third-party construction damage to buried gas pipelines according to claim 1, characterized in that, Once a construction signal is identified, if it is identified as a construction signal 5-20 times consecutively within 1 second (i.e., at a frequency of 5-20Hz), it can be considered a road breaking construction signal.

9. A third-party construction damage edge early warning model for buried gas pipelines according to any one of claims 1-8, characterized in that, Before sending an early warning message to the corresponding unit, the early warning module sends the original signal data and judgment result to the cloud server for further confirmation. After confirmation, the early warning message is sent.

Images