A pipeline remote monitoring and early warning method and system based on the Internet of Things

Abstract

The present invention discloses a pipeline remote monitoring and early warning method and system based on the Internet of Things, which relates to the technical field of pipeline leakage monitoring. By collecting pipeline acoustic signal data in real time and inputting the denoised acoustic signal into a pre-trained pipeline leakage recognition model, an identification result is obtained; if the identification result is a leak, the leak is located based on the subsequent acoustic signal to obtain the location information of the leak point; and an early warning report is generated based on the identification result and location information. By combining the Internet of Things technology with acoustic signal collection and intelligent recognition models, real-time, remote monitoring and precise positioning of pipeline leaks are achieved, especially with higher recognition sensitivity for tiny cracks and chronic leaks, making up for the shortcomings of traditional flow and pressure detection methods, improving operation and maintenance efficiency, and reducing safety hazards.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline leakage detection, and in particular to a pipeline remote monitoring and early warning method and system based on the Internet of Things. Background Art

[0002] With the acceleration of industrialization and urbanization, pipelines have become an indispensable infrastructure in modern society, widely used in various sectors such as oil, natural gas, water supply, and heating. The safe operation of pipelines is directly related to energy supply, environmental protection, and the safety of people's lives and property. However, because pipelines are often distributed over vast geographical areas, and some are buried underground, traditional monitoring and maintenance methods often cannot fully and in real time understand the operational status of pipelines. Pipeline failures such as leaks and ruptures are difficult to detect and repair in a timely manner, which can easily lead to serious environmental pollution, safety accidents, and economic losses.

[0003] Flow and pressure detection are common techniques used in traditional pipeline monitoring. Leaks are typically detected by monitoring changes in flow rate and pressure fluctuations within the pipeline. For large-scale leaks (such as pipeline ruptures or other large-scale leaks), flow and pressure changes are often significant, allowing monitoring systems to easily detect anomalies. However, for smaller leaks, particularly tiny cracks or chronic leaks, flow and pressure changes are often subtle or absent, making it difficult for the system to detect leaks in a timely manner. This can hinder early warnings and hinder repair efficiency. Summary of the Invention

[0004] The purpose of the present invention is to solve the problem of difficulty in timely detecting leakage mentioned in the above background technology, and to propose a pipeline remote monitoring and early warning method and system based on the Internet of Things.

[0005] The first aspect of the present invention provides a pipeline remote monitoring and early warning method based on the Internet of Things, the method comprising:

[0006] Acquiring acoustic signal data of a target section of the pipeline in real time through a plurality of sensors, and extracting a first acoustic signal of a first target time period;

[0007] preprocessing the first acoustic signal to obtain a second acoustic signal;

[0008] Inputting the second acoustic signal into a pre-trained pipeline leakage recognition model to obtain a recognition result; the recognition result includes normal and leakage;

[0009] If the identification result is a leak, obtaining an acoustic signal of a second target period, and locating the leak based on the acoustic signal of the second target period to obtain the location information of the leak point; the second target period is a period of preset length after the first target period;

[0010] Generate an early warning report based on the recognition result and the location information.

[0011] Optionally, the preprocessing the first acoustic signal to obtain the second acoustic signal includes:

[0012] performing variational modal decomposition on the first acoustic signal data to obtain a plurality of modal components;

[0013] The envelope entropy of each modal component is calculated, and the modal component whose envelope entropy is less than a preset threshold is recorded as a valid component;

[0014] Perform wavelet threshold denoising on each effective component to obtain the denoised effective component;

[0015] Signal reconstruction is performed based on the multiple denoised effective components to obtain a second acoustic signal.

[0016] Optionally, performing wavelet threshold denoising on each effective component to obtain the denoised effective component includes:

[0017] Using wavelet transform to decompose the target modal component into detail coefficients and approximate coefficients at multiple scales; the target modal component is any valid component;

[0018] A preset threshold function is used to perform threshold processing on detail coefficients of multiple scales to obtain multiple target detail coefficients; the threshold function is:

[0019] ;

[0020] in, is the target detail coefficient after threshold processing; W j,k is the detail coefficient of the j-th level, λ j is the corresponding wavelet threshold; sgn() is the sign function; p is a constant;

[0021] The approximation coefficient and a plurality of target detail coefficients are subjected to inverse wavelet transform to obtain effective components for denoising.

[0022] Optionally, the pipeline leakage identification model includes an initial convolutional network, a dense convolutional network, and a fully connected network; wherein:

[0023] The initial convolutional network uses three consecutive one-dimensional convolutional layers to extract local features of the input acoustic signal, and then uses a maximum pooling layer to perform spatial downsampling to obtain a first feature map;

[0024] The dense convolutional network uses four consecutive dense blocks to enhance the feature representation of the first feature map, and then uses a maximum pooling layer to perform spatial downsampling to obtain a second feature map;

[0025] The fully connected network uses three fully connected layers to map the second feature map to a leakage probability score, and if the leakage probability score is greater than a preset threshold, the output recognition result is leakage.

[0026] Optionally, acquiring the acoustic signal of the second target time period, and locating the leak according to the acoustic signal of the second target time period to obtain the position information of the leak point includes:

[0027] Marking the sensor corresponding to the acoustic signal data identified as a leak as a first target sensor, and acquiring acoustic signals collected by multiple first target sensors during a second target period;

[0028] According to the position information of each first target sensor, two adjacent sensors are recorded as a sensor group, and the time difference threshold is determined according to the distance between them;

[0029] Calculate the propagation time difference of the acoustic signal in each sensor group;

[0030] The pipeline section corresponding to the sensor group whose propagation time difference is less than the time difference threshold is taken as the leakage section; the sensors of this sensor group are recorded as upstream sensors and downstream sensors;

[0031] The position of the leakage point is calculated based on the distance between the upstream sensor and the downstream sensor, the propagation time difference and the speed of the sound wave.

[0032] A second aspect of the present invention provides a pipeline remote monitoring and early warning system based on the Internet of Things, the system comprising:

[0033] A data acquisition module is used to collect acoustic signal data of a target section of the pipeline in real time through multiple sensors, and extract a first acoustic signal of a first target time period;

[0034] a preprocessing module, configured to preprocess the first acoustic signal to obtain a second acoustic signal;

[0035] a leakage identification module, configured to input the second acoustic signal into a pre-trained pipeline leakage identification model to obtain an identification result; the identification result includes normal and leakage;

[0036] a leakage locating module configured to, if the identification result is a leakage, obtain acoustic signals of a second target period, and perform leakage locating based on the acoustic signals of the second target period to obtain location information of the leakage point; the second target period is a period of preset duration after the first target period;

[0037] The early warning module is used to generate an early warning report based on the recognition result and the location information.

[0038] Optionally, the preprocessing module includes:

[0039] a modal decomposition module, configured to perform variational modal decomposition on the first acoustic signal data to obtain a plurality of modal components;

[0040] The noise discrimination module is used to calculate the envelope entropy of each modal component and record the modal component with an envelope entropy less than a preset threshold as a valid component;

[0041] The wavelet denoising module is used to perform wavelet threshold denoising on each effective component to obtain the denoised effective component;

[0042] The signal reconstruction module is used to reconstruct the signal according to the multiple effective components of the denoising to obtain the second sound signal.

[0043] Optionally, the wavelet denoising module includes:

[0044] A wavelet transform module, configured to decompose a target modal component into detail coefficients and approximate coefficients of multiple scales using wavelet transform; the target modal component is any valid component;

[0045] The threshold processing module is used to perform threshold processing on detail coefficients of multiple scales using a preset threshold function to obtain multiple target detail coefficients; the threshold function is:

[0046] ;

[0047] in, is the target detail coefficient after threshold processing; W j,k is the detail coefficient of the j-th level, λ j is the corresponding wavelet threshold; sgn() is the sign function; p is a constant;

[0048] The wavelet inverse transform module is used to perform wavelet inverse transform on the approximate coefficients and a plurality of target detail coefficients to obtain effective components for denoising.

[0049] Optionally, the pipeline leakage identification model includes an initial convolutional network, a dense convolutional network, and a fully connected network; wherein:

[0050] The initial convolutional network uses three consecutive one-dimensional convolutional layers to extract local features of the input acoustic signal, and then uses a maximum pooling layer to perform spatial downsampling to obtain a first feature map;

[0051] The dense convolutional network uses four consecutive dense blocks to enhance the feature representation of the first feature map, and then uses a maximum pooling layer to perform spatial downsampling to obtain a second feature map;

[0052] The fully connected network uses three fully connected layers to map the second feature map to a leakage probability score, and if the leakage probability score is greater than a preset threshold, the output recognition result is leakage.

[0053] Optionally, the leakage locating module includes:

[0054] an area determination module, configured to mark a sensor corresponding to acoustic signal data identified as a leak as a first target sensor, and obtain acoustic signals collected by multiple first target sensors during a second target period;

[0055] a longest time difference determination module, configured to record two adjacent sensors as a sensor group based on the position information of each first target sensor, and determine a time difference threshold based on the distance between the two sensors;

[0056] A time difference calculation module, used to calculate the propagation time difference of the acoustic signal in each sensor group;

[0057] A pipeline segment determination module is used to select the pipeline segment corresponding to the sensor group whose propagation time difference is less than the time difference threshold as the leakage segment; and record the sensors of the sensor group as upstream sensors and downstream sensors;

[0058] The position determination module is used to calculate the position of the leakage point according to the distance between the upstream sensor and the downstream sensor, the propagation time difference and the speed of the sound wave.

[0059] Beneficial effects of the present invention:

[0060] The present invention proposes a pipeline remote monitoring and early warning method based on the Internet of Things. The method includes: collecting acoustic signal data of a target section of the pipeline in real time through multiple sensors, and extracting a first acoustic signal of a first target time period; preprocessing the first acoustic signal to obtain a second acoustic signal; inputting the second acoustic signal into a pre-trained pipeline leakage recognition model to obtain a recognition result; the recognition result includes normal and leakage; if the recognition result is leakage, obtaining an acoustic signal of a second target time period, and locating the leakage based on the acoustic signal of the second target time period to obtain location information of the leakage point; the second target time period is a preset time period after the first target time period; and generating an early warning report based on the recognition result and the location information.

[0061] By combining Internet of Things technology with acoustic signal acquisition and intelligent recognition models, real-time, remote monitoring and precise positioning of pipeline leaks can be achieved, especially with higher recognition sensitivity for tiny cracks and chronic leaks, making up for the shortcomings of traditional flow and pressure detection methods, improving operation and maintenance efficiency, reducing safety hazards, and significantly enhancing the safety and intelligence level of pipeline operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1The present invention provides a flowchart of a pipeline remote monitoring and early warning method based on the Internet of Things.

[0063] Figure 2 A network structure diagram of a leakage identification model is provided for an embodiment of the present invention;

[0064] Figure 3 The present invention provides an architecture diagram of a pipeline remote monitoring and early warning system based on the Internet of Things. DETAILED DESCRIPTION

[0065] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0066] The embodiment of the present invention provides a pipeline remote monitoring and early warning method based on the Internet of Things. Figure 1 , Figure 1 The present invention provides a flowchart of a pipeline remote monitoring and early warning method based on the Internet of Things. The method includes the following steps:

[0067] S101 , collecting acoustic signal data of a target section of a pipeline in real time through multiple sensors, and extracting a first acoustic signal of a first target time period.

[0068] S102: Preprocess the first sound signal to obtain a second sound signal.

[0069] S103: Input the second acoustic signal into a pre-trained pipeline leakage recognition model to obtain a recognition result.

[0070] S104: If the identification result is leakage, obtain the acoustic signal of the second target period, and locate the leakage according to the acoustic signal of the second target period to obtain the position information of the leakage point.

[0071] S105: Generate an early warning report based on the recognition result and location information.

[0072] The target pipe section is a straight pipe; the identification results include normal and leaking; and the second target period is a preset time period after the first target period.

[0073] An IoT-based pipeline remote monitoring and early warning method provided by an embodiment of the present invention utilizes IoT technology in conjunction with acoustic signal acquisition and intelligent recognition models to achieve real-time, remote monitoring and precise positioning of pipeline leaks. This method, with higher recognition sensitivity for microcracks and chronic leaks, compensates for the shortcomings of traditional flow and pressure detection methods, improves operation and maintenance efficiency, reduces safety hazards, and significantly enhances the safety and intelligence level of pipeline operations.

[0074] In one implementation, the first target time period is the period before the current time, for example, from 1.2 seconds before the current time to 1 second before the current time. The first acoustic signal is the acoustic signal collected by any sensor during this time period. By delaying the analysis by one second for near-real-time analysis, it is possible to avoid situations where only the near-range sensor detects a leak while the far-range sensor has not yet detected it, due to delays in signal transmission and processing between sensors.

[0075] In one embodiment, step S102 includes:

[0076] Step 1: Perform variational modal decomposition on the first acoustic signal data to obtain multiple modal components.

[0077] Step 2: Calculate the envelope entropy of each modal component, and record the modal component whose envelope entropy is less than a preset threshold as a valid component.

[0078] Step 3: Perform wavelet threshold denoising on each effective component to obtain the denoised effective component.

[0079] Step 4: reconstruct the signal based on the multiple denoised effective components to obtain a second sound signal.

[0080] This embodiment uses variational modal decomposition to decompose the acoustic signal into multiple modal components, effectively extracting signal features. By calculating envelope entropy, significant components are screened, irrelevant noise is removed, and leak-related information is enhanced. Subsequently, wavelet threshold denoising is used to remove residual noise, further improving signal clarity and stability. Finally, signal reconstruction is performed based on the denoised significant components to produce a second acoustic signal. This process significantly improves signal quality, provides more accurate input for subsequent leak identification, and enhances the reliability and sensitivity of the detection system.

[0081] In one implementation, the preset threshold can be determined using statistical methods. For example, the mean and standard deviation of the envelope entropy of all modal components can be calculated, and the threshold can then be set to the mean plus three times the standard deviation. This statistically based threshold setting method can effectively filter out noise signals, reduce false positives, while retaining true leakage signals and improving the overall robustness of the system.

[0082] In one embodiment, wavelet threshold denoising is performed on each effective component, and the denoised effective components obtained include:

[0083] Wavelet transform is used to decompose the target modal component into detail coefficients and approximate coefficients at multiple scales; the target modal component is any valid component;

[0084] The threshold function is used to perform threshold processing on the detail coefficients of multiple scales to obtain multiple target detail coefficients; the threshold function is:

[0085] ;

[0086] in, is the target detail coefficient after threshold processing; W j,k is the detail coefficient of the j-th level, λ j is the corresponding wavelet threshold; sgn() is the sign function; p is a constant;

[0087] The approximation coefficients and multiple target detail coefficients are subjected to inverse wavelet transform to obtain the effective denoising components.

[0088] In one implementation, the wavelet transform decomposes the signal into detail coefficients and approximation coefficients at multiple scales, allowing for analysis at different scales. This effectively removes noise from various frequency bands while preserving key leakage signature information. By applying thresholding to the detail coefficients, fine-tuning can be performed for noise at different scales, optimizing signal quality and minimizing the impact of interfering noise.

[0089] In one implementation, the db8 wavelet basis function can be used for wavelet transform, and the number of decomposition layers is 4. The threshold calculation formula is: ; Among them, λ j is the wavelet threshold of the jth layer; MED j is the median of the detail coefficients of the jth layer; N j is the length of the j-th layer detail coefficient.

[0090] In one implementation, the traditional threshold function is divided into a hard threshold function and a soft threshold function. The hard threshold function can obtain a higher peak signal-to-noise ratio by setting the wavelet coefficients below the threshold to zero, but due to the discontinuity of the threshold function, the processed signal may oscillate or be discontinuous at the edge of the signal. In contrast, the soft threshold function has a continuous structure, which can make the processed signal smoother. However, soft threshold processing also introduces a constant deviation, resulting in a certain degree of distortion of the high-frequency information of the signal, especially at the level of detail of the signal. The threshold function introduced in this embodiment, the parameter p can be set to 0.5. The function is an odd function, taking the positive half axis as an example. When W j,kWhen it is greater than the threshold, the larger it is, the closer the corrected value is to the original value, and it is continuous at the threshold, avoiding the reconstruction oscillation problem caused by the hard threshold function and the reconstruction deviation problem caused by the soft threshold function.

[0091] In one implementation, 30 datasets were randomly selected from the experimental data and denoised using different threshold functions (hard threshold function, soft threshold function, and improved threshold function). The signal-to-noise ratio (SNR) and root mean square error (RMSE) of each dataset after denoising were calculated. Finally, the denoising effects of different methods were evaluated by taking the average value. The comparison results are shown in Table 1.

[0092] Table 1:

[0093]

[0094] The improvement in signal-to-noise ratio indicates that this method can more effectively increase the proportion of effective components in the signal and reduce the impact of noise. The reduction in root mean square error indicates that the denoised signal deviates less from the original signal, resulting in a better denoising effect.

[0095] In one embodiment, see Figure 2 , Figure 2 This is a network structure diagram of a leakage identification model provided by an embodiment of the present invention. The pipeline leakage identification model includes an initial convolutional network, a dense convolutional network, and a fully connected network; wherein:

[0096] The initial convolutional network uses three consecutive one-dimensional convolutional layers to extract local features from the input acoustic signal. This is followed by a maximum pooling layer for spatial downsampling to produce the first feature map. Specifically, each convolutional layer uses the ReLU activation function; the maximum pooling layer has a pooling window size of 2 and a stride of 2.

[0097] The dense convolutional network uses four consecutive dense blocks to enhance the feature representation of the first feature map, followed by a max pooling layer for spatial downsampling to produce the second feature map. Specifically, each dense block consists of four cascaded one-dimensional convolutional layers, with the last convolutional layer receiving the output of all preceding convolutional layers as input. Each convolutional layer uses the ReLU activation function.

[0098] The fully connected network uses three fully connected layers to map the second feature map to a leakage probability score. If the leakage probability score exceeds a preset threshold, the identification result is a leakage. Specifically, the second feature map is flattened into a one-dimensional vector and input into the first fully connected layer. The first two fully connected layers use the ReLU activation function. The final fully connected layer uses the Sigmoid function to obtain the probability score.

[0099] In one implementation, a dense convolutional network enhances feature flow and reuse through multiple layers of dense connections, improving the model's ability to express and generalize leakage features and mitigating the vanishing gradient problem. The overall model structure efficiently integrates shallow and deep features, significantly improving the accuracy of identifying small leaks and the robustness of the model.

[0100] In one implementation, to verify the effectiveness of the pipeline leak detection method proposed in this paper, the performance of the proposed improved model was compared with that of mainstream 1D CNN and LSTM models. Evaluation metrics included accuracy, precision, recall, and F1 score. The specific comparison results are shown in Table 2.

[0101] Table 2:

[0102]

[0103] As shown in Table 2, the improved model outperforms the comparison model across all evaluation metrics. Its higher accuracy and precision indicate that the overall prediction results are closest to the actual situation, reducing both false positives and false negatives. Its higher recall rate indicates a stronger ability to detect actual leaks, a lower false negative rate, and improved security for the system.

[0104] In one embodiment, step S104 includes:

[0105] Marking the sensor corresponding to the acoustic signal data identified as a leak as a first target sensor, and acquiring acoustic signals collected by multiple first target sensors during a second target period;

[0106] According to the position information of each first target sensor, two adjacent sensors are recorded as a sensor group, and the time difference threshold is determined according to the distance between them;

[0107] Calculate the propagation time difference of the acoustic signal in each sensor group;

[0108] The pipeline section corresponding to the sensor group whose propagation time difference is less than the time difference threshold is taken as the leakage section; the sensors of this sensor group are recorded as upstream sensors and downstream sensors;

[0109] The location of the leak is calculated based on the distance between the upstream sensor and the downstream sensor, the propagation time difference and the speed of the sound wave.

[0110] In one implementation, the time difference threshold TM = D / v, where D is the distance between the two sensors and v is the speed of sound waves. This ensures that the propagation time difference is within a reasonable range, effectively eliminating abnormal data and improving positioning accuracy.

[0111] In one implementation, a cross-correlation method can be used to calculate the propagation time difference within a sensor group. By analyzing the propagation time difference of the acoustic signals between adjacent sensors, the leak point can be accurately located.

[0112] In one implementation, the calculation formula for the leak point location is: ; Where x is the distance from the upstream sensor, that is, the location of the leak; d is the distance between the upstream sensor and the downstream sensor; v is the speed of sound; is the time difference. When the upstream sensor is closer to the leak point, the time difference is positive.

[0113] The embodiment of the present invention provides a pipeline remote monitoring and early warning system based on the Internet of Things. Figure 3 , Figure 3 The present invention provides an architecture diagram of a remote pipeline monitoring and early warning system based on the Internet of Things. The system includes:

[0114] The data acquisition module is used to collect the acoustic signal data of the target section pipeline in real time through multiple sensors, and extract the first acoustic signal of the first target time period.

[0115] The preprocessing module is used to preprocess the first acoustic signal to obtain a second acoustic signal.

[0116] The leakage identification module is used to input the second acoustic signal into the pre-trained pipeline leakage identification model to obtain an identification result.

[0117] The leakage locating module is used to obtain the acoustic signal of the second target period if the identification result is leakage, and to locate the leakage according to the acoustic signal of the second target period to obtain the position information of the leakage point.

[0118] The early warning module is used to generate early warning reports based on the recognition results and location information.

[0119] The target pipe section is a straight pipe; the identification results include normal and leaking; and the second target period is a preset time period after the first target period.

[0120] An IoT-based pipeline remote monitoring and early warning system provided by an embodiment of the present invention utilizes IoT technology in conjunction with acoustic signal acquisition and intelligent recognition models to achieve real-time, remote monitoring and precise positioning of pipeline leaks. The system offers increased recognition sensitivity for microcracks and chronic leaks, compensating for the shortcomings of traditional flow and pressure detection methods, improving operation and maintenance efficiency, reducing safety hazards, and significantly enhancing the safety and intelligence of pipeline operations.

[0121] In one embodiment, the pre-processing module includes:

[0122] The modal decomposition module is used to perform variational modal decomposition on the first acoustic signal data to obtain multiple modal components.

[0123] The noise discrimination module is used to calculate the envelope entropy of each modal component and record the modal component with an envelope entropy less than a preset threshold as a valid component.

[0124] The wavelet denoising module is used to perform wavelet threshold denoising on each effective component to obtain the denoised effective component.

[0125] The signal reconstruction module is used to reconstruct the signal according to the multiple effective components of the denoising to obtain the second sound signal.

[0126] In one embodiment, the wavelet denoising module includes:

[0127] The wavelet transform module is used to decompose the target modal component into detail coefficients and approximate coefficients of multiple scales by using wavelet transform; the target modal component is any valid component.

[0128] The threshold processing module is used to perform threshold processing on detail coefficients of multiple scales using a preset threshold function to obtain multiple target detail coefficients; the threshold function is:

[0129] ;

[0130] in, is the target detail coefficient after threshold processing; W j,k is the detail coefficient of the j-th level, λ j is the corresponding wavelet threshold; sgn() is the sign function; p is a constant.

[0131] The inverse wavelet transform module is used to perform inverse wavelet transform on the approximate coefficients and multiple target detail coefficients to obtain the effective components for denoising.

[0132] In one embodiment, the leakage locating module includes:

[0133] an area determination module, configured to mark a sensor corresponding to acoustic signal data identified as a leak as a first target sensor, and obtain acoustic signals collected by multiple first target sensors during a second target period;

[0134] a longest time difference determination module, configured to record two adjacent sensors as a sensor group based on the position information of each first target sensor, and determine a time difference threshold based on the distance between the two sensors;

[0135] A time difference calculation module, used to calculate the propagation time difference of the acoustic signal in each sensor group;

[0136] A pipeline segment determination module is used to select the pipeline segment corresponding to the sensor group whose propagation time difference is less than the time difference threshold as the leakage segment; and record the sensors of the sensor group as upstream sensors and downstream sensors;

[0137] The location determination module is used to calculate the location of the leakage point based on the distance between the upstream sensor and the downstream sensor, the propagation time difference and the speed of the sound wave.

[0138] It should be noted that, in this document, terms such as "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements that are inherent to such process, method, article or apparatus.

[0139] The embodiments of the present invention are described in detail above, but the contents described are only preferred embodiments of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A pipeline remote monitoring and early warning method based on the Internet of Things, characterized in that: The method comprises: Acquiring acoustic signal data of a target section of the pipeline in real time through a plurality of sensors, and extracting a first acoustic signal of a first target time period; preprocessing the first acoustic signal to obtain a second acoustic signal; Inputting the second acoustic signal into a pre-trained pipeline leakage recognition model to obtain a recognition result; the recognition result includes normal and leakage; If the identification result is a leak, obtaining an acoustic signal of a second target period, and locating the leak based on the acoustic signal of the second target period to obtain the location information of the leak point; the second target period is a period of preset length after the first target period; generating an early warning report based on the identification result and the location information; The preprocessing of the first acoustic signal to obtain the second acoustic signal includes: performing variational modal decomposition on the first acoustic signal data to obtain a plurality of modal components; The envelope entropy of each modal component is calculated, and the modal component whose envelope entropy is less than a preset threshold is recorded as a valid component; Perform wavelet threshold denoising on each effective component to obtain the denoised effective component; Reconstructing the signal according to the multiple effective components of denoising to obtain a second sound signal; The wavelet threshold denoising is performed on each effective component to obtain the denoised effective component, which includes: Using wavelet transform to decompose the target modal component into detail coefficients and approximate coefficients at multiple scales; the target modal component is any valid component; A preset threshold function is used to perform threshold processing on detail coefficients of multiple scales to obtain multiple target detail coefficients; the threshold function is: ； in, is the target detail coefficient after threshold processing; W j,k is the detail coefficient of the j-th level, λ j is the corresponding wavelet threshold; sgn( ) is the sign function; p is a constant; The approximation coefficient and a plurality of target detail coefficients are subjected to inverse wavelet transform to obtain effective components for denoising.

2. The method for remote monitoring and early warning of pipelines based on the Internet of Things according to claim 1 is characterized in that: The pipeline leakage identification model includes an initial convolutional network, a dense convolutional network and a fully connected network; wherein: The initial convolutional network uses three consecutive one-dimensional convolutional layers to extract local features of the input acoustic signal, and then uses a maximum pooling layer to perform spatial downsampling to obtain a first feature map; The dense convolutional network uses four consecutive dense blocks to enhance the feature representation of the first feature map, and then uses a maximum pooling layer to perform spatial downsampling to obtain a second feature map; The fully connected network uses three fully connected layers to map the second feature map to a leakage probability score, and if the leakage probability score is greater than a preset threshold, the output recognition result is leakage.

3. The method for remote monitoring and early warning of pipelines based on the Internet of Things according to claim 1 is characterized in that: The acquiring of the acoustic signal of the second target period, and performing leakage location according to the acoustic signal of the second target period to obtain the position information of the leakage point includes: Marking the sensor corresponding to the acoustic signal data identified as a leak as a first target sensor, and acquiring acoustic signals collected by multiple first target sensors during a second target period; According to the position information of each first target sensor, two adjacent sensors are recorded as a sensor group, and the time difference threshold is determined according to the distance between them; Calculate the propagation time difference of the acoustic signal in each sensor group; The pipeline section corresponding to the sensor group whose propagation time difference is less than the time difference threshold is taken as the leakage section; the sensors of this sensor group are recorded as upstream sensors and downstream sensors; The position of the leakage point is calculated based on the distance between the upstream sensor and the downstream sensor, the propagation time difference and the speed of the sound wave.

4. A pipeline remote monitoring and early warning system based on the Internet of Things, characterized in that: The system comprises: A data acquisition module is used to collect acoustic signal data of a target section of the pipeline in real time through multiple sensors, and extract a first acoustic signal of a first target time period; a preprocessing module, configured to preprocess the first acoustic signal to obtain a second acoustic signal; a leakage identification module, configured to input the second acoustic signal into a pre-trained pipeline leakage identification model to obtain an identification result; the identification result includes normal and leakage; a leakage locating module configured to, if the identification result is a leakage, obtain acoustic signals of a second target period, and perform leakage locating based on the acoustic signals of the second target period to obtain location information of the leakage point; the second target period is a period of preset duration after the first target period; An early warning module, configured to generate an early warning report based on the recognition result and the location information; The pre-processing module comprises: a modal decomposition module, configured to perform variational modal decomposition on the first acoustic signal data to obtain a plurality of modal components; The noise discrimination module is used to calculate the envelope entropy of each modal component and record the modal component with an envelope entropy less than a preset threshold as a valid component; The wavelet denoising module is used to perform wavelet threshold denoising on each effective component to obtain the denoised effective component; A signal reconstruction module, configured to reconstruct a signal based on a plurality of effective components removed from noise to obtain a second acoustic signal; The wavelet denoising module includes: A wavelet transform module, configured to decompose a target modal component into detail coefficients and approximate coefficients of multiple scales using wavelet transform; the target modal component is any valid component; The threshold processing module is used to perform threshold processing on detail coefficients of multiple scales using a preset threshold function to obtain multiple target detail coefficients; the threshold function is: ； in, is the target detail coefficient after threshold processing; W j,k is the detail coefficient of the j-th level, λ j is the corresponding wavelet threshold; sgn( ) is the sign function; p is a constant; The wavelet inverse transform module is used to perform wavelet inverse transform on the approximate coefficients and a plurality of target detail coefficients to obtain effective components for denoising.

5. The pipeline remote monitoring and early warning system based on the Internet of Things according to claim 4 is characterized in that: The pipeline leakage identification model includes an initial convolutional network, a dense convolutional network and a fully connected network; wherein: The initial convolutional network uses three consecutive one-dimensional convolutional layers to extract local features of the input acoustic signal, and then uses a maximum pooling layer to perform spatial downsampling to obtain a first feature map; The dense convolutional network uses four consecutive dense blocks to enhance the feature representation of the first feature map, and then uses a maximum pooling layer to perform spatial downsampling to obtain a second feature map; The fully connected network uses three fully connected layers to map the second feature map to a leakage probability score, and if the leakage probability score is greater than a preset threshold, the output recognition result is leakage.

6. The pipeline remote monitoring and early warning system based on the Internet of Things according to claim 4 is characterized in that: The leakage positioning module includes: an area determination module, configured to mark a sensor corresponding to acoustic signal data identified as a leak as a first target sensor, and obtain acoustic signals collected by multiple first target sensors during a second target period; a longest time difference determination module, configured to record two adjacent sensors as a sensor group based on the position information of each first target sensor, and determine a time difference threshold based on the distance between the two sensors; A time difference calculation module, used to calculate the propagation time difference of the acoustic signal in each sensor group; A pipeline segment determination module is used to select the pipeline segment corresponding to the sensor group whose propagation time difference is less than the time difference threshold as the leakage segment; and record the sensors of the sensor group as upstream sensors and downstream sensors; The position determination module is used to calculate the position of the leakage point according to the distance between the upstream sensor and the downstream sensor, the propagation time difference and the speed of the sound wave.

Images