Drainage pipe network topology detection device and method based on multi-frequency acoustic vibration coupling

Abstract

The application provides a drainage pipe network topology detection device and method based on multi-frequency acoustic vibration coupling, wherein a detection device is installed at the bottom of one shaft and the butt joint position of one end of a horizontal pipe network, a signal receiving device is installed at the bottom of another shaft and the butt joint position of the other end of the horizontal pipe network, a signal transmission module is arranged on the detection device and the signal receiving device, and the signal transmission module is wirelessly connected with a tablet computer. Compared with manual exploration, CCTV detection, floating object method and smoke method, the application has the advantages of high efficiency, low cost, high precision and strong adaptability, provides efficient and accurate technical support for digital and intelligent operation and maintenance of municipal drainage pipe networks, and promotes the transformation of the industry to automation and intelligence.

Description

Technical Field

[0001] This invention relates to the field of drainage pipe network topology detection, and in particular to a drainage pipe network topology detection device and method based on multi-frequency acoustic-vibration coupling. Background Technology

[0002] In the field of municipal drainage network operation and maintenance, unclear pipeline topology has become a key technical bottleneck restricting the industry's development. This problem mainly stems from three factors: First, historical pipelines generally suffer from missing design drawings and inadequate record management; second, the number of existing inspection wells is large and their spatial distribution is complex, with a lack of systematic records of well-pipe connections; and third, traditional detection methods have inherent limitations in topology identification.

[0003] Current mainstream topology identification methods suffer from significant technical drawbacks: manual exploration is not only inefficient but also heavily reliant on operator experience; while CCTV camera inspection can acquire images of the pipe interior, it suffers from high equipment costs and long inspection cycles, making large-scale application difficult. Traditional methods commonly used in the industry, such as the floating object method and the smoke method, face even greater challenges: the floating object method requires multiple people to work together and has an unstable success rate; the smoke method is limited by pipe water levels and siltation conditions, making it difficult to guarantee detection effectiveness and posing certain safety hazards.

[0004] It is worth noting that existing acoustic inspection technologies primarily focus on pipeline defect diagnosis, completely lacking dedicated solutions for identifying pipeline network topology. This technological deficiency directly leads to problems such as low pipeline network operation and maintenance efficiency and insufficient decision-making basis, severely hindering the digital and intelligent transformation of municipal drainage networks. The industry urgently needs to develop new inspection technologies and equipment to achieve efficient and accurate identification of pipeline network topology relationships. Summary of the Invention

[0005] The main objective of this invention is to provide a device and method for detecting the topology of drainage pipe networks based on multi-frequency acoustic-vibration coupling, which solves the problem of unclear topology caused by the lack of design drawings, complex inspection wells and limited detection methods in traditional methods.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a drainage pipe network topology detection device based on multi-frequency acoustic-vibration coupling, wherein a detection device is installed at the bottom of one vertical shaft where it is connected to one end of the horizontal pipe network, and a signal receiving device is installed at the bottom of another vertical shaft at the adjacent location where it is connected to the other end of the horizontal pipe network. Both the detection device and the signal receiving device are equipped with a signal transmission module, and the signal transmission module is wirelessly connected to a tablet computer.

[0007] In the preferred embodiment, the detection device includes a low-frequency vibrator, an underwater acoustic transmitter, and a mid-frequency loudspeaker; The low-frequency vibrator is surface-mounted and fitted onto the inner wall of the transverse pipe network. The underwater acoustic transmitter is installed inside the horizontal pipe network and below the water surface; The mid-frequency loudspeaker is suspended in the middle of the horizontal pipe network, above the water surface; It is also equipped with a first lifting support, on which a first hoist is mounted. The hoisting rope of the first hoist is connected to the first inspection frame, and the outer ring of the first inspection frame is in contact with the surface of the vertical shaft wall. Below the first testing frame are multiple first mounting rods, on which the low-frequency vibrator, underwater acoustic transmitter, and intermediate-frequency speaker are mounted.

[0008] In the preferred embodiment, the low-frequency vibrator outputs a mechanical force ≥5N and uses an electromagnetic or piezoelectric stacked vibrator. The mid-frequency loudspeaker emits airborne sound signals omnidirectionally, with a sound pressure level ≥100dB@1m, covering the 200Hz–5kHz frequency band, and uses linear frequency modulation to enhance noise immunity; The underwater acoustic transmitter emits low-frequency underwater acoustic pulses with a pulse width of 10ms. The center frequency of the piezoelectric ceramic transducer is 5kHz, and the emission angle is compressed to ±30° by focusing through an acoustic lens.

[0009] In the preferred embodiment, the signal receiving device includes a radio receiver, a hydrophone, and an accelerometer; The accelerometer is attached to the inner wall of the transverse pipe network to detect pipe wall vibration. Its sensitivity is 90-100mV / g and its range is 40-60g. The hydrophone is installed inside the horizontal pipe network and below the water surface. The hydrophone is used to receive underwater acoustic signals. The sensitivity is -180 to 170 dBre1V / μPa, the frequency response is 10 to 10000Hz, and the sensitivity attenuation in water is ≤40dB. The sound receiving device is suspended in the middle of the horizontal pipe network and above the water surface. The sound receiving device is used to collect airborne sound signals, with a dynamic range of 30–130dB, a frequency response of 20–8000Hz, and a sensitivity attenuation of ≤60dB in air.

[0010] In the preferred embodiment, multiple second mounting rods are provided below the second testing frame. The receiver, hydrophone, and accelerometer are all mounted on the second mounting rods on the second testing frame, and the outer surface of the second testing frame is in contact with the surface of the shaft. Another shaft is also equipped with a second lifting support, on which a second hoist is mounted, and the second hoist is connected to the second inspection frame.

[0011] In a preferred embodiment, the method includes: S1. Install the detection device inside the shaft. Deploy the signal receiving device in the well to be measured with the same water level mode. Start the detection devices in the installation sequence and start them in the corresponding signal receiving device to receive data in the corresponding mode. S2. The data received by the signal receiving device is transmitted to the tablet computer through the wireless transmission module. The tablet computer transmits the data to the main server. The server processes the data and then performs multimodal feature fusion. S3. Multimodal feature fusion data uses TDOA technology and multi-well location data reception to generate a three-dimensional pipeline axis, with nodes being inspection wells and edges being pipelines. Combined with GIS base map data, it is rendered into a 3D model pipeline data. S4. Overlay the defect heatmap onto the 3D model to visualize the defect location; S5. Based on visualized defect data and combined with historical inspection data, predict pipeline performance degradation trends and potential failures; S6. Generate maintenance plans based on the forecast results and quantify funding requirements; S7. Verify the accuracy of the prediction by testing data after maintenance, and optimize the model parameters.

[0012] In the preferred embodiment, in step S1: the low-frequency vibrator, the underwater acoustic transmitter, and the intermediate-frequency loudspeaker of the detection device are installed, the water inside the transverse pipe network is half full, the intermediate-frequency loudspeaker is suspended in the middle of the air layer, the low-frequency vibrator is attached to the pipe wall, and the underwater acoustic transmitter is immersed below the water surface. The signal receiving device is deployed in the same water level pattern for the well to be measured. The accelerometer and the transmitter vibrator are axially aligned. The hydrophone and the hydroacoustic head are at the same height. The microphone and the speaker are at the same height. Among them, the outer surfaces of the first and second testing frames are bonded to the shaft wall by rubber sealing rings; The low-frequency vibrator, underwater acoustic transmitter, and mid-frequency loudspeaker are activated in sequence; The accelerometer, hydrophone, and microphone receive data in sequence.

[0013] In the preferred scheme, in steps S2-S4, First, the low-frequency vibrator, intermediate-frequency loudspeaker, and underwater acoustic transmitter at the transmitting end transmit 20-800Hz pipe wall vibration signals, 200Hz-5kHz air acoustic signals, and 1-10kHz underwater acoustic signals respectively according to the time-division multiplexing sequence. Among them, the vibrator and loudspeaker are activated first in the empty pipe / low water level condition, and the vibrator and underwater acoustic transmitter are activated first in the full pipe condition. Secondly, the accelerometer, hydrophone and microphone at the receiving end synchronously collect vibration, underwater sound and air sound signals, which are wirelessly transmitted to the server. Single-channel preprocessing is performed through wavelet packet denoising and FIR filtering to extract multimodal feature vectors containing resonant frequency, sound pressure level attenuation rate and signal arrival time. Principal component analysis is used to reduce the dimensionality to 3D space, and the topological connectivity confidence is output through logistic regression model. Then, based on the multi-well location received data, the TDOA triangulation algorithm is used to solve the three-dimensional coordinates of the pipeline network nodes. The pipeline axis is smoothed by the moving least squares method. Combined with GIS base map coordinate transformation and WebGL rendering technology, a 3D pipeline network model containing material and pipe diameter attributes is generated. Finally, Gaussian kernel density estimation is applied to the fused feature vectors to generate a defect heatmap. Defect locations are visualized using HSV color mapping and Otsu thresholding. Defect type classification is implemented based on a random forest or CNN model, and a maintenance decision report containing repair suggestions and cost estimates is output. The Gaussian kernel density function of the defect heatmap is... ,bandwidth , The standard deviation of the defect location. This represents the number of defects.

[0014] In the preferred scheme, the 3D model rendering uses the Three.js framework to load the GIS OpenStreetMap base map, the pipe model adopts cylindrical geometry, the radius is mapped from the pipe diameter data, and the material is set with metal roughness and environmental reflection. The 3D space in the 3D model data is divided into a grid, and the center of each grid is... Calculate density value Generate density matrix ; Add a volume rendering layer to the Three.js scene, convert the density matrix F into Texture3D, render it as a semi-transparent heatmap through a shader, and overlay it on top of the pipeline model; The heat map area displays density values ​​and defect types; clicking it generates a repair report that includes location and severity.

[0015] In the preferred embodiment, steps S5-S7 include: The 3D coordinates of pipeline nodes are calculated based on the TDOA algorithm, and a 3D pipeline model is generated by combining it with a GIS base map. Gaussian kernel density estimation is applied to the fused features to generate a defect heatmap, and defects are visualized by HSV color mapping and Otsu threshold segmentation. Based on the visualized defect data and historical detection data, an LSTM network is used to construct a spatiotemporal feature vector. The prediction model, combined with the risk index formula Predicting pipeline performance degradation trends, among which To predict severity, m is the material fragility index. This is the critical distance threshold. Distance between defects and inspection wells; Based on the prediction results, a maintenance plan is matched using a rule engine, and a cost formula is applied. Quantify funding needs, among which For material costs, This is the material loss coefficient. For labor costs, The variable is an artificial floating coefficient; the prediction accuracy is verified by calculating MAE and RMSE using post-repair inspection data, and the LSTM model parameters are adjusted using the backpropagation algorithm, where the MAE formula is... The RMSE formula is: , This represents the severity value after repair.

[0016] This invention provides a device and method for detecting the topology of drainage pipe networks based on multi-frequency acoustic-vibration coupling. Utilizing multi-frequency acoustic-vibration coupling technology, complex drainage pipe network topologies can be effectively detected and identified, solving the problem of traditional detection methods' difficulty in accurately determining pipe network layouts. Secondly, the combined application of the detection device and signal receiving device not only improves detection accuracy but also simplifies the operation process, making on-site work more convenient and efficient. Furthermore, real-time data transmission to a tablet computer via wireless connection enables instant data analysis and visualization, greatly improving work efficiency and decision-making speed. In addition, this method reduces interference with the environment and the pipe network structure itself, helping to protect the integrity of existing facilities. Finally, the entire system design considers various variables in practical application scenarios, possessing strong adaptability and stability, providing reliable technical support for the maintenance and management of drainage pipe networks.

[0017] By employing multi-frequency acoustic coupling through a low-frequency vibrator, a mid-frequency loudspeaker, and a hydroacoustic transmitter, combined with multi-modal signal acquisition using an accelerometer, hydrophone, and receiver, this method can cover all operating conditions, including empty pipes, low water levels, and full pipes, as well as pipe networks made of various materials such as concrete and PE. This solves the problem of unclear topological structures caused by traditional methods due to missing design drawings, complex inspection wells, and limited detection capabilities. Compared to manual inspection, CCTV detection, floating object methods, and smoke methods, it offers advantages such as high efficiency, low cost, high accuracy, and strong adaptability. It provides efficient and precise technical support for the digital and intelligent operation and maintenance of municipal drainage pipe networks, driving the industry's transformation towards automation and intelligence. Attached Figure Description

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a schematic diagram of the layout of the drainage pipe network topology detection device of the present invention; Figure 2 This is a flowchart of the drainage pipe network topology detection method of the present invention. Detailed Implementation

[0019] Example 1 like Figure 1-2As shown, a drainage pipe network topology detection device based on multi-frequency acoustic-vibration coupling is provided. The detection device is installed at the bottom of one vertical shaft 1, which is connected to one end of the horizontal pipe network 2. The signal receiving device is installed at the bottom of another vertical shaft at the adjacent position, which is connected to the other end of the horizontal pipe network 2. Both the detection device and the signal receiving device are equipped with a signal transmission module, which is wirelessly connected to a tablet computer 5.

[0020] A detection device is installed at the bottom of a vertical shaft 1, where it connects to one end of the horizontal pipe network 2. Simultaneously, a signal receiving device is installed at the bottom of another adjacent vertical shaft, where it connects to the other end of the horizontal pipe network 2. Both the detection device and the signal receiving device are equipped with signal transmission modules, and these modules are capable of wirelessly connecting to a tablet computer 5. After the detection device is activated, it emits multi-frequency acoustic vibration signals that propagate through the pipe network. When the signals encounter different parts of the pipe network, they are reflected or scattered. The signal receiving device captures these altered signals and transmits them wirelessly to the tablet computer 5 for preliminary analysis.

[0021] First, multi-frequency acoustic-vibration coupling technology can effectively detect and identify complex drainage pipe network topologies, solving the problem of traditional detection methods' difficulty in accurately determining pipe network layout. Second, the combined application of detection and signal receiving devices not only improves detection accuracy but also simplifies the operation process, making on-site work more convenient and efficient. Third, real-time data transmission to a tablet PC via wireless connection enables instant data analysis and visualization, greatly improving work efficiency and decision-making speed. Furthermore, this method reduces interference with the environment and the pipe network structure itself, helping to protect the integrity of existing facilities. Finally, the entire system design considers various variables in practical application scenarios, possessing strong adaptability and stability, providing reliable technical support for the maintenance and management of drainage pipe networks.

[0022] In the preferred embodiment, the detection device includes a low-frequency vibrator 8, an underwater acoustic transmitter 9, and a mid-frequency loudspeaker 10; The low-frequency vibrator 8 is surface-mounted and installed on the inner wall of the transverse pipe network 2; The underwater acoustic transmitter 9 is installed inside the horizontal pipe network 2 and below the water surface; The intermediate frequency loudspeaker 10 is suspended in the middle of the horizontal pipe network 2, above the water surface; The transmitter is lowered to the bottom of the target shaft using a counterweight device and the first hoist 4. Based on the real-time water level (0-100%) fed back by the water level sensor, the deployment mode is automatically selected. Empty pipe / low water level ≤20%: intermediate frequency loudspeaker 10 is suspended in the middle of the air layer, low frequency vibrator 8 is attached to the bottom of the pipe wall, and underwater acoustic transmitter 9 is in a dormant state; Full tube ≥80%: underwater acoustic transmitter 9 is submerged below the water surface, low-frequency vibrator 8 remains in contact with the tube wall, and mid-frequency speaker 10 is in sleep mode; Half-full tube 20%–80%: All three components are fully activated, and crosstalk is reduced to <-50 dB through a silicone vibration isolation layer.

[0023] The receiver is deployed in the same water level mode in the well to be measured, ensuring that the axial alignment error between the accelerometer 17 and the transmitter vibrator 8 is ≤5 cm, the hydrophone 16 is at the same height as the hydrophone head 9, and the microphone 15 is at the same height as the speaker 10.

[0024] The second detection frame 13 at the receiving end is attached to the shaft wall by a rubber sealing ring to suppress environmental noise interference.

[0025] It is also provided with a first lifting support 3, on which a first hoist 4 is provided. The hoisting rope of the first hoist 4 is connected to the first inspection frame 6, and the outer ring of the first inspection frame 6 is in contact with the surface of the shaft wall 1. Multiple first mounting rods 7 are provided below the first testing frame 6, and the low-frequency vibrator 8, the underwater acoustic transmitter 9 and the intermediate frequency speaker 10 are mounted on the mounting rods.

[0026] In the preferred embodiment, the low-frequency vibrator outputs a mechanical force of ≥5N with 8 vibrations, and uses an electromagnetic or piezoelectric stacked vibrator. The intermediate frequency loudspeaker 10 emits airborne sound signals omnidirectionally, with a sound pressure level ≥100dB@1m, covering the 200Hz–5kHz frequency band, and uses linear frequency modulation to enhance noise immunity; The underwater acoustic transmitter 9 emits low-frequency underwater acoustic pulses with a pulse width of 10ms. The center frequency of the piezoelectric ceramic transducer is 5kHz, and the emission angle is compressed to ±30° by focusing through an acoustic lens.

[0027] The intermediate frequency loudspeaker emits a 200 Hz–5 kHz linear frequency modulated chirp signal, using pseudo-random code PRN modulation, with a sound pressure level ≥100 dB@1m, and excites air layer standing waves such as 200 Hz standing waves with a node spacing of 0.85 m in a DN800 duct; The underwater acoustic transmitter emits a BPSK modulated pulse with a pulse width of 10 ms at a center frequency of 5 kHz. The single reflection loss due to pipe wall reflection is reduced by ≥6 dB by focusing the emission angle by ±30° through an acoustic lens. The low-frequency vibrator scans the 20–800 Hz frequency band in 10 Hz increments. Concrete pipes are prioritized for scanning 20–200 Hz, while PE pipes are scanned for 200–800 Hz. The output mechanical force is programmable from 5–20 N.

[0028] In a preferred embodiment, the signal receiving device includes a radio receiver 15, a hydrophone 16, and an accelerometer 17. Accelerometer 17 is attached to the inner wall of transverse pipe network 2 to detect pipe wall vibration. Its sensitivity is 90-100mV / g and its range is 40-60g. Hydrophone 16 is installed inside the transverse pipe network 2 and below the water surface. Hydrophone 16 is used to receive underwater acoustic signals, with a sensitivity of –180--170dBre1V / μPa, a frequency response of 10-10000Hz, and a sensitivity attenuation of ≤40dB in water. The sound receiving device 15 is installed in the middle of the horizontal pipe network 2 and above the water surface. The sound receiving device 15 is used to collect airborne sound signals with a dynamic range of 30–130dB, a frequency response of 20–8000Hz, and a sensitivity attenuation of ≤60dB in the air.

[0029] The receiver synchronously acquires data from the accelerometer (20 Hz–1 kHz), hydrophone (1–10 kHz), and microphone (200 Hz–5 kHz) at a sampling rate of 20 kHz.

[0030] In the preferred embodiment, multiple second mounting rods 14 are provided below the second testing frame 13. The receiver 15, hydrophone 16 and accelerometer 17 are all mounted on the second mounting rods 14 on the second testing frame 13. The outer surface of the second testing frame 13 is in contact with the surface of the shaft. Another shaft is also equipped with a second lifting support 11, on which a second hoist 12 is mounted, and the second hoist 12 is connected to the second inspection frame 13.

[0031] In the municipal drainage network, at the bottom of a vertical shaft 1 where it connects with one end of a horizontal pipe network 2, a first detection frame 6 is lowered using a first lifting bracket 3 and a first lifting machine 4. This allows the low-frequency vibrator 8 to be in contact with the inner wall of the horizontal pipe network 2, the underwater acoustic transmitter 9 to be placed below the water surface, and the intermediate frequency speaker 10 to be suspended in the middle position above the water surface. At the bottom of another vertical shaft at an adjacent location where it connects with the other end of the horizontal pipe network 2, a second detection frame 13 is lowered using a second lifting bracket 11 and a second lifting machine 12. This allows the accelerometer 17 to be in contact with the inner wall of the pipe network, the hydrophone 16 to be placed below the water surface, and the sound receiving device 15 to be suspended in the middle position above the water surface. The detection device and the signal receiving device are wirelessly connected to a tablet computer 5 via a signal transmission module. After activation, the low-frequency vibrator 8 of the detection device outputs a mechanical force ≥5N to excite pipe wall vibration at 20–800Hz, the intermediate frequency speaker 10 emits omnidirectional signals at 200Hz–5kHz with an airborne sound pressure level ≥100dB@1m, linear frequency modulation enhances noise reduction, and the underwater acoustic transmitter 9 emits a pulse width of 10ms. The system includes a 5kHz low-frequency underwater acoustic pulse lens with a focusing emission angle of ±30°; an accelerometer 17 for the signal receiving device with a sensitivity of 90-100mV / g and a range of 40-60g to detect pipe wall vibration; a hydrophone 16 with a sensitivity of -180~-170dB re 1V / μPa, a frequency response of 10–10000Hz, and an attenuation of ≤40dB in water to receive underwater acoustic signals; and a receiver 15 with a dynamic range of 30–130dB, a frequency response of 20–8000Hz, and an attenuation of ≤60dB in air to collect airborne acoustic signals. The data is transmitted to a tablet computer 5 via a signal transmission module for processing and analysis.

[0032] By employing multi-frequency acoustic coupling through low-frequency vibrators, mid-frequency loudspeakers, and underwater acoustic transmitters, combined with multi-modal signal acquisition using accelerometers, hydrophones, and receivers, this method can cover all operating conditions, including empty pipes, low water levels, and full pipes, as well as pipe networks made of various materials such as concrete and PE. It solves the problem of unclear topology caused by missing design drawings, complex manholes, and limited detection methods in traditional methods. Compared to manual exploration, CCTV inspection, floating object methods, and smoke methods, it offers advantages such as high efficiency (single-person operation, short single-inspection time), low cost (reducing manpower and equipment investment), high accuracy (achieving ±0.5 meter topology positioning accuracy based on TDOA technology and signal attenuation models), strong adaptability (intelligent switch control adapts to complex environments), and a 95% success rate. This provides efficient and accurate technical support for the digital and intelligent operation and maintenance of municipal drainage pipe networks, driving the industry towards automation and intelligence.

[0033] Example 2 Further explanation in conjunction with Example 1, such as Figure 1-2As shown, the method includes: S1, installing a detection device inside the shaft 1, deploying a signal receiving device in the well to be measured with the same water level mode, starting the detection devices in sequence and starting them in the order of installation, and the corresponding signal receiving device receiving data in the corresponding manner. S2. The data received by the signal receiving device is transmitted to the tablet computer 5 through the wireless transmission module. The tablet computer 5 transmits the data to the main server. The server processes the data and then performs multimodal feature fusion on the data. S3. Multimodal feature fusion data uses TDOA technology and multi-well location data reception to generate a three-dimensional pipeline axis, with nodes being inspection wells and edges being pipelines. Combined with GIS base map data, it is rendered into a 3D model pipeline data. S4. Overlay the defect heatmap onto the 3D model to visualize the defect location; S5. Based on visualized defect data and combined with historical inspection data, predict pipeline performance degradation trends and potential failures; S6. Generate maintenance plans based on the forecast results and quantify funding requirements; S7. Verify the accuracy of the prediction by testing data after maintenance, and optimize the model parameters.

[0034] A detection device is installed inside shaft 1, and signal receiving devices are deployed in the wells to be tested using the same water level pattern. The detection devices are activated sequentially according to the installation order, and the corresponding signal receiving devices receive data in the corresponding manner. The signal receiving devices transmit the collected data to tablet computer 5 via a wireless transmission module, and tablet computer 5 then uploads the data to the central server for processing. The server performs fusion analysis on the multimodal data and, combined with TDOA time difference of arrival technology and multi-well location received data, generates a three-dimensional axial pipe network model with inspection wells as nodes and pipelines as edges. This model is further rendered into 3D pipe network data by combining GIS base map data.

[0035] Subsequently, a defect heatmap is overlaid on the constructed 3D model to visualize the location of pipeline defects. Based on the visualized defect data and combined with historical inspection data, the system can predict the degradation trend of pipeline performance and potential failure risks. Based on the prediction results, a targeted maintenance plan is generated and the required financial investment is quantified. After maintenance is completed, the device is used again to acquire inspection data to verify the accuracy of the prediction and further optimize model parameters to improve the accuracy of subsequent identification and prediction.

[0036] First, it achieves efficient and accurate identification of drainage pipe network topology, solving the problems of low efficiency, strong dependence, and poor adaptability of traditional methods. Second, through TDOA technology and multi-well location collaborative reception, it can more accurately reconstruct the three-dimensional spatial layout of the pipe network, providing reliable spatial data support for subsequent operation and maintenance. Third, the system supports the overlay display of defect heatmaps, intuitively presenting the pipeline problem areas and improving the efficiency of problem location. Fourth, by combining historical data and real-time detection information, it has the ability to predict pipeline performance degradation trends, which helps to realize the transformation from passive maintenance to proactive prevention. Fifth, the functions of maintenance decision-making and funding requirement quantification enhance the scientific nature and controllability of project management. Finally, by continuously transmitting new data and optimizing model parameters, the system has the ability to continuously learn and improve itself, which is conducive to promoting the long-term development of intelligent operation and maintenance systems for drainage pipe networks.

[0037] In the preferred embodiment, in step S1: the low-frequency vibrator 8, the underwater acoustic transmitter 9, and the intermediate-frequency speaker 10 of the detection device are installed, the water inside the transverse pipe network 2 is half full, the intermediate-frequency speaker 10 is suspended in the middle of the air layer, the low-frequency vibrator 8 is attached to the pipe wall, and the underwater acoustic transmitter 9 is immersed below the water surface. The signal receiving device is deployed in the same water level mode for the well to be measured. The accelerometer 17 is axially aligned with the transmitter vibrator 8, the hydrophone 16 is at the same height as the hydrophone head 9, and the microphone 15 is at the same height as the speaker 10. Among them, the outer surfaces of the first testing frame 6 and the second testing frame 13 are attached to the shaft wall by rubber sealing rings; The low-frequency vibrator 8, the underwater acoustic transmitter 9, and the intermediate-frequency loudspeaker 10 are activated in sequence; The accelerometer 17, hydrophone 16, and microphone 15 receive data in sequence.

[0038] Example 3 Further explanation in conjunction with Example 1, such as Figure 1-2 As shown, the method includes: In the preferred embodiment, in steps S2-S4, firstly, through the low-frequency vibrator, intermediate-frequency loudspeaker and underwater acoustic transmitter at the transmitting end, 20-800Hz pipe wall vibration signal, 200Hz-5kHz air acoustic signal and 1-10kHz underwater acoustic signal are transmitted respectively according to the time-division multiplexing sequence, wherein the vibrator and loudspeaker are activated preferentially in the empty pipe / low water level condition, and the vibrator and underwater acoustic transmitter are activated preferentially in the full pipe condition; Secondly, the accelerometer, hydrophone and microphone at the receiving end synchronously collect vibration, underwater sound and air sound signals, which are wirelessly transmitted to the server. Single-channel preprocessing is performed through wavelet packet denoising and FIR filtering to extract multimodal feature vectors containing resonant frequency, sound pressure level attenuation rate and signal arrival time. Principal component analysis is used to reduce the dimensionality to 3D space, and the topological connectivity confidence is output through logistic regression model. Then, based on the multi-well location received data, the TDOA triangulation algorithm is used to solve the three-dimensional coordinates of the pipeline network nodes. The pipeline axis is smoothed by the moving least squares method. Combined with GIS base map coordinate transformation and WebGL rendering technology, a 3D pipeline network model containing material and pipe diameter attributes is generated. Finally, Gaussian kernel density estimation is applied to the fused feature vectors to generate a defect heatmap. Defect locations are visualized using HSV color mapping and Otsu thresholding. Defect type classification is implemented based on a random forest or CNN model, and a maintenance decision report containing repair suggestions and cost estimates is output. The Gaussian kernel density function of the defect heatmap is... ,bandwidth , The standard deviation of the defect location. This represents the number of defects.

[0039] In the preferred scheme, the 3D model rendering uses the Three.js framework to load the GIS OpenStreetMap base map, the pipe model adopts cylindrical geometry, the radius is mapped from the pipe diameter data, and the material is set with metal roughness and environmental reflection. The 3D space in the 3D model data is divided into a grid, and the center of each grid is... Calculate density value Generate density matrix ; Add a volume rendering layer to the Three.js scene, convert the density matrix F into Texture3D, render it as a semi-transparent heatmap through a shader, and overlay it on top of the pipeline model; The heat map area displays density values ​​and defect types; clicking it generates a repair report that includes location and severity.

[0040] The following are the specific methods for steps S2-S4: S2. Multimodal feature fusion algorithm: It fuses heterogeneous data from accelerometers, hydrophones, and microphones into a unified feature space, eliminates physical differences between modes, and extracts key information.

[0041] Wavelet packet denoising decomposes the signal into different frequency bands and performs thresholding on the noise-dominant high-frequency components. Taking the db4 wavelet basis as an example, a 3-level decomposition can divide the vibration signal (20Hz–1kHz) into 8 frequency bands (such as 20–50Hz, 50–100Hz, etc.).

[0042] Decomposition: Analysis of vibration signals Perform 3-level wavelet packet decomposition to obtain the frequency band coefficient matrix. Each row corresponds to a frequency band. This represents the number of sampling points.

[0043] Threshold calculation: for each frequency band Calculate the standard deviation of noise threshold .

[0044] Soft thresholding: ; Soft thresholding avoids phase distortion caused by hard thresholding by preserving symbol information.

[0045] Reconstruction: Using the denoised coefficients Reconstructed signal The signal-to-noise ratio is improved by approximately 10–15 dB.

[0046] It effectively suppresses power frequency noise (50Hz) and environmental vibration interference, highlighting the vibration characteristics related to defects (such as high-frequency resonance of cracks).

[0047] Filter Design: An FIR bandpass filter is designed using the window function method, taking an airborne acoustic filter as an example. Passband range: 200Hz–5kHz, stopband attenuation ≥60dB Impulse response: ; in This is the Hanning window function, used to suppress the Gibbs effect.

[0048] Real-time processing: microphone signal and hydrophone signal Each output is passed through its corresponding filter. and .

[0049] Ensure that airborne and underwater acoustic signals contain only the target frequency band energy and avoid cross-modal contamination (such as inefficient radiation signals from underwater acoustic transducers in the airborne acoustic channel).

[0050] Multimodal feature fusion: (1) Construction of joint feature vectors Vibration feature extraction: resonant frequency Perform an FFT on the denoised vibration signal and take the frequency corresponding to the maximum peak value. The formula is as follows: ; Effective value (RMS): Measures vibrational energy; the formula is: ; Kurtosis: Used to detect impact defects; the formula is: ; The kurtosis of a normal signal is approximately 3, while the kurtosis of a defective signal is significantly increased.

[0051] Acoustic Feature Extraction: Peak Frequency of Airborne Sound Standing Waves Perform an FFT on the filtered airborne sound signal and take the maximum peak frequency within the range of 200–5kHz.

[0052] Attenuation rate The formula for calculating the ratio of the sound pressure level difference (SPL) between the transmitter and receiver to the distance is: ; Underwater acoustic arrival time (ToA): The time delay between transmitted and received signals is calculated using a cross-correlation algorithm, with the following formula: ; Bit Error Rate (BER): The ratio of the number of erroneous bits to the total number of bits after demodulating a BPSK modulated underwater acoustic signal.

[0053] (2) Principal Component Analysis (PCA) Dimensionality Reduction Standardization: For the 7-dimensional feature matrix Perform zero-mean, unit-variance standardization: ; Solving for the covariance matrix: ; Eigenvalue decomposition: Solve ,according to Sort in descending order and select the first 3 principal components.

[0054] Projection transformation: ; Highly relevant features are compressed into a low-dimensional space to reduce the computational cost of subsequent modeling while retaining key information.

[0055] (3) Logistic Regression Fusion (Decision Level) Model construction: using the dimensionality-reduced features Using the input as input, construct a binary classification logistic regression model and output the topological connectivity probability. : ; in As weight, As a bias, training is performed using stochastic gradient descent, with the loss function being cross-entropy: ; This indicates that a connection exists. This indicates no connection.

[0056] Threshold determination: Set a threshold ,like It is determined to be a valid topology connection.

[0057] S3. 3D Pipeline Network Modeling Algorithm (TDOA and GIS Integration): A three-dimensional geometric model of the pipeline network is constructed based on multi-well TDOA data and integrated with the geographic information system (GIS) base map to achieve spatial positioning and visualization.

[0058] 1. Triangulation Algorithm (TDOA) No. The coordinates of the receiver are The coordinates of the transmitter are The signal travels through the air for 100 minutes. In water ,but: ; Take the first receiving end as the reference point ( After eliminating the radical sign, we obtain the TDOA equation: ; Least squares solution, defining the error function: ; The minimum value is found using the Newton-Gaussian iteration method, with the initial value taken as the coordinates of the first receiver end. Iteration continues until the error change is less than [the minimum value is missing from the original text]. m.

[0059] Accuracy Improvement: Joint localization is achieved by combining multimodal signals (airborne sound + underwater sound), and weights are adjusted using weighted least squares. ; 2. Moving Least Squares (MLS) Smoothing of Axis Local weighted fitting: for discrete points on the pipeline axis At the query point Select nearby Using the nearest neighbors, construct a weighted multinomial fit: ; The weighting function uses a Gaussian kernel: ; Coefficient solution: by minimizing the weighted sum of squared errors: ; Solving for coefficients ,in For the basis function matrix, It is a diagonal weight matrix. This is a vector of data points.

[0060] Output: Generates a smoothed 3D axis, with nodes representing manholes and edges representing smooth pipes, with an error ≤ 0.3m.

[0061] 3. GIS base map fusion and rendering Coordinate system transformation: Converting TDOA to Cartesian coordinates Convert to latitude and longitude coordinates The seven-parameter Burshawf model is adopted: ; in For rotation parameters, For translation parameters, This is the scale parameter.

[0062] 3D model rendering: The Three.js framework is used to load GIS base maps (such as OpenStreetMap). The pipe model adopts cylindrical geometry, and the radius is mapped from the pipe diameter data (such as DN800 corresponding to a radius of 0.4m). The material settings include metal roughness and environmental reflection to enhance the realism.

[0063] Interactive features: Clicking on a pipe node displays topology information (such as connection probability 92% and pipe length 31.4m), and right-clicking on a defect area triggers the generation of a maintenance work order.

[0064] S4. Defect Heatmap Generation Algorithm Visualize the defect distribution in the 3D pipeline network model, highlight high-risk areas through density estimation, and assist in operation and maintenance decision-making.

[0065] 1. Defect location kernel density estimation (KDE) Continuous density function: Let the set of detected defect locations be... The three-dimensional kernel density function is: ; in For the number of defects, bandwidth Using Silverman's empirical formula: ; Discretization: Divide the 3D space into a grid (e.g., 0.5m × 0.5m × 0.5m), and for each grid center... Calculate density value Generate density matrix .

[0066] 2. Color Mapping and Thresholding Density normalization: ; HSV color mapping: Low risk ( : Green (H=120°, S=0.8, V=0.9); Medium risk (0.3≤ Yellow (H=60°, S=0.8, V=0.9); High risk ( : Red (H=0°, S=0.8, V=0.9); Color values ​​are converted to RGB using the following formula: ; Otsu thresholding: Calculating the inter-class variance of the density histogram. : ; Traversing thresholds Select to make The largest As a threshold for defect region segmentation, it automatically distinguishes between normal and abnormal regions.

[0067] 3. Heatmap Overlay and Interaction 3D visualization: Add a volume rendering layer to the Three.js scene, convert the density matrix $F$ into Texture3D, render it as a semi-transparent heatmap through a shader, and overlay it on top of the pipeline model.

[0068] Dynamic query: Hovering the mouse over the heat map area displays the density value and defect type; clicking generates a repair report including location and severity.

[0069] Example 4 Further explanation in conjunction with Example 1, such as Figure 1-2 As shown, the method includes: In the preferred embodiment, steps S2-S4, steps S5-S7 include: The 3D coordinates of pipeline nodes are calculated based on the TDOA algorithm, and a 3D pipeline model is generated by combining it with a GIS base map. Gaussian kernel density estimation is applied to the fused features to generate a defect heatmap, and defects are visualized by HSV color mapping and Otsu threshold segmentation. Based on the visualized defect data and historical detection data, an LSTM network is used to construct a spatiotemporal feature vector. The prediction model, combined with the risk index formula Predicting pipeline performance degradation trends, among which To predict severity, m is the material fragility index. This is the critical distance threshold. Distance between defects and inspection wells; Based on the prediction results, a maintenance plan is matched using a rule engine, and a cost formula is applied. Quantify funding needs, among which For material costs, This is the material loss coefficient. For labor costs, The variable is an artificial floating coefficient; the prediction accuracy is verified by calculating MAE and RMSE using post-repair inspection data, and the LSTM model parameters are adjusted using the backpropagation algorithm, where the MAE formula is... The RMSE formula is: , This represents the severity value after repair.

[0070] The specific methods for S5-S7 are as follows: S5. Pipeline Performance Degradation Prediction Algorithm Based on the visualization of defects in 3D models, and combined with historical inspection data, the trend of pipeline performance degradation is predicted, and potential faults are identified.

[0071] 1. Time-series characteristic engineering Data preprocessing: Extract defect location, type (corrosion / crack / leak), severity score (110 points), and detection timestamp from historical inspection data to construct a time-series dataset. ,in For the detection time, For the defect coordinates, Severity level.

[0072] Time alignment: Data from different detection periods are uniformly resampled into a monthly sequence, and cubic spline interpolation is used to fill in missing values.

[0073] Outlier Handling: The IQR (Interquartile Range) method is used to detect the severity of outliers. The formula is as follows: ; Values ​​outside the interval are smoothed and replaced with a weighted average of the nearest valid values.

[0074] 2. Long Short-Term Memory (LSTM) Network Prediction Model Network architecture: The input layer is a spatiotemporal feature vector. Temporal dependencies are extracted using two LSTM layers, and the fully connected layer outputs the severity prediction values ​​for the next 12 months. .

[0075] Gating mechanism: LSTM uses forget gates Input gate and output gate Control the flow of information. Attention mechanism: Add attention weights to the LSTM output layer, as shown in the formula: ; in For time step The attention weight is adjusted to enhance the focus on key historical information.

[0076] 3. Potential fault warning Risk scoring: Combining predicted severity Pipe material fragility index (Take 0.8 for concrete and 0.6 for PE), calculate the risk index. : ; in This is the critical distance threshold; the closer the distance, the higher the risk.

[0077] Spatiotemporal clustering analysis: DBSCAN clustering is performed on the risk index to identify high-risk areas. The parameter is set as: neighborhood radius. Minimum number of samples .

[0078] Density function: ; in For indicator functions, when the distance is less than It is 1 if it is true, otherwise it is 0.

[0079] S6. Maintenance Plan Generation and Cost Quantification Algorithm: Automatically generates maintenance plans based on prediction results, and quantifies the costs of materials, labor, equipment, etc.

[0080] 1. Defect-based solution matching algorithm Rule Engine: Establishes a mapping table of defect type remediation solutions, based on expert experience and a historical case library, and uses production rule representation: ; Fuzzy matching: For complex defects (such as those with both corrosion and cracks), fuzzy logic is used to calculate the matching degree. ; in , , These are the weighting coefficients.

[0081] 2. Cost Quantification Model Material costs: ; Pipe diameter correction factor: ; in This is the actual pipe diameter. This is the reference pipe diameter.

[0082] Labor costs: ; in To speed up construction, This refers to the curing time.

[0083] Environmental correction factor: ; The burial depth is measured in meters, and the traffic impact is rated on a scale of 15.

[0084] 3. Multi-objective optimization Pareto optimal solution: Solved using the NSGAII algorithm, the objective function is: ; Fitness function: ; Elite retention strategy: The top 10% of the best individuals in each generation are retained and directly enter the next generation to ensure that the quality of the solution does not degrade.

[0085] S7. Model Validation and Parameter Optimization Algorithm The accuracy of the predictions was verified by testing data after maintenance, and the parameters of the LSTM model were iteratively optimized.

[0086] 1. Prediction Error Analysis Error decomposition: The total error is decomposed into bias and variance. ; in , .

[0087] Spatiotemporal error heatmap: The prediction error for each spatial location is calculated, and a continuous error distribution is generated through kriging interpolation. ; in For the weighting coefficients, satisfying .

[0088] 2. Model parameter update Meta-learning optimization: Employing the MAML (Model Agnostic Meta-Learning) algorithm to quickly adapt to new data, the formula is as follows: ; ; in For the inner loop learning rate, The learning rate is the outer loop rate.

[0089] Bayesian hyperparameter optimization: Constructing a Gaussian process surrogate model and maximizing the expected improvement (EI) to obtain the optimal hyperparameters: ; in This is the current optimal solution. This is the true objective function.

[0090] 3. Closed-loop feedback process Dynamic weight adjustment: The prediction weights are dynamically adjusted based on the historical errors at different locations. The formula is as follows: ; in For position The mean absolute error, This represents the global average error.

[0091] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A drainage pipe network topology detection device based on multi-frequency acoustic-vibration coupling, characterized in that: one of them A detection device is installed at the bottom of the vertical shaft (1) where it connects with one end of the horizontal pipe network (2). A signal receiving device is installed at the bottom of another vertical shaft at the adjacent location where it connects with the other end of the horizontal pipe network (2). Both the detection device and the signal receiving device are equipped with a signal transmission module. The signal transmission module is wirelessly connected to the tablet computer (5). The detection device includes a low-frequency vibrator (8), an underwater acoustic transmitter (9), and a mid-frequency loudspeaker (10). The low-frequency vibrator (8) is surface-mounted and installed on the inner wall of the transverse pipe network (2); The underwater acoustic transmitter (9) is located inside the transverse pipe network (2) and below the water surface; The mid-frequency loudspeaker (10) is suspended in the middle of the horizontal pipe network (2) and above the water surface; It is also provided with a first lifting support (3), a first hoist (4) is provided on the first lifting support (3), the hoisting rope of the first hoist (4) is connected to the first inspection frame (6), and the outer ring of the first inspection frame (6) is in contact with the surface of the shaft wall (1); Multiple first mounting rods (7) are provided below the first testing frame (6), and a low-frequency vibrator (8), a water acoustic transmitter (9) and a medium-frequency loudspeaker (10) are mounted on the mounting rods; The signal receiving device includes a radio (15), a hydrophone (16), and an accelerometer (17). An accelerometer (17) is attached to the inner wall of the transverse pipe network (2) to detect pipe wall vibration. The sensitivity is 90-100mV / g and the range is 40-60g. The hydrophone (16) is installed inside the transverse pipe network (2) and below the water surface. The hydrophone (16) is used to receive underwater acoustic signals. The sensitivity is -180 to 170 dBre1V / μPa, the frequency response is 10 to 10000Hz, and the sensitivity attenuation in water is ≤40dB. The receiver (15) is suspended in the middle of the horizontal pipe network (2) and above the water surface. The receiver (15) is used to collect airborne sound signals with a dynamic range of 30–130dB, a frequency response of 20–8000Hz, and a sensitivity attenuation of ≤60dB in the air.

2. The drainage pipe network topology detection device based on multi-frequency acoustic-vibration coupling according to claim 1, characterized in that: The low-frequency vibrator (8) outputs a mechanical force ≥5N and uses an electromagnetic or piezoelectric stack exciter. The intermediate frequency loudspeaker (10) emits airborne sound signals in all directions with a sound pressure level ≥100dB@1m, covering the 200Hz–5kHz frequency band, and uses linear frequency modulation to enhance noise immunity; The underwater acoustic transmitter (9) emits low-frequency underwater acoustic pulses with a pulse width of 10ms and a piezoelectric ceramic transducer center frequency of 5kHz. The emission angle is compressed to ±30° by focusing through an acoustic lens.

3. The drainage pipe network topology detection device based on multi-frequency acoustic-vibration coupling according to claim 1, characterized in that: Multiple second mounting rods (14) are provided below the second testing frame (13). The receiver (15), hydrophone (16) and accelerometer (17) are all mounted on the second mounting rods (14) on the second testing frame (13). The outer surface of the second testing frame (13) is in contact with the surface of the shaft. Another shaft is provided with a second lifting support (11) on the outside, and a second hoist (12) is provided on the second lifting support (11). The second hoist (12) is connected to the second inspection frame (13).

4. The detection method of the drainage pipe network topology detection device based on multi-frequency acoustic-vibration coupling according to any one of claims 1-3, characterized in that: The method includes: S1. Install detection devices inside the vertical shaft (1), deploy signal receiving devices in the well to be tested with the same water level mode, start the detection devices in sequence, and receive data in the corresponding manner. S2. The data received by the signal receiving device is transmitted to the tablet computer (5) through the wireless transmission module. The tablet computer (5) transmits the data to the main server. The server processes the data and then performs multimodal feature fusion on the data. S3. Multimodal feature fusion data uses TDOA technology and multi-well location data reception to generate a three-dimensional pipeline axis, with nodes being inspection wells and edges being pipelines. Combined with GIS base map data, it is rendered into a 3D model pipeline data. S4. Overlay the defect heatmap onto the 3D model to visualize the defect location; S5. Based on visualized defect data and combined with historical inspection data, predict pipeline performance degradation trends and potential failures; S6. Generate a maintenance plan based on the forecast results and quantify the funding requirements; S7. Verify the accuracy of the prediction by testing data after maintenance, and optimize the model parameters.

5. The detection method of the drainage pipe network topology detection device based on multi-frequency acoustic-vibration coupling according to claim 4, characterized in that: in step S1: the low-frequency vibrator (8), the underwater acoustic transmitter (9) and the intermediate frequency speaker (10) of the detection device are installed, the water inside the transverse pipe network (2) is half full, the intermediate frequency speaker (10) is suspended in the middle of the air layer, the low-frequency vibrator (8) is attached to the pipe wall, and the underwater acoustic transmitter (9) is immersed below the water surface; The signal receiving device is deployed in the same water level mode for the well to be measured. The accelerometer (17) is axially aligned with the low-frequency vibrator (8) of the transmitting end, the hydrophone (16) is at the same height as the underwater acoustic transmitter (9), and the sound receiving device (15) is at the same height as the loudspeaker (10). in, The outer surfaces of the first testing frame (6) and the second testing frame (13) are bonded to the shaft wall by rubber sealing rings; The low-frequency vibrator (8), the underwater acoustic transmitter (9), and the mid-frequency loudspeaker (10) are activated in sequence; The accelerometer (17), hydrophone (16), and receiver (15) receive data in sequence.

6. The detection method of the drainage pipe network topology detection device based on multi-frequency acoustic-vibration coupling according to claim 4, characterized in that: In steps S2-S4, First, the low-frequency vibrator, intermediate-frequency loudspeaker, and underwater acoustic transmitter at the transmitting end transmit 20-800Hz pipe wall vibration signals, 200Hz-5kHz air acoustic signals, and 1-10kHz underwater acoustic signals respectively according to the time-division multiplexing sequence. Among them, the vibrator and loudspeaker are activated first in the empty pipe / low water level condition, and the vibrator and underwater acoustic transmitter are activated first in the full pipe condition. Secondly, the accelerometer, hydrophone and radio receiver (15) at the receiving end synchronously collect vibration, underwater sound and air sound signals, which are wirelessly transmitted to the server. Single-channel preprocessing is performed by wavelet packet denoising and FIR filtering to extract multimodal feature vectors containing resonance frequency, sound pressure level attenuation rate and signal arrival time. Principal component analysis is used to reduce the dimension to 3D space, and the topological connectivity confidence is output through logistic regression model. Then, based on the multi-well location received data, the TDOA triangulation algorithm is used to solve the three-dimensional coordinates of the pipeline network nodes. The pipeline axis is smoothed by the moving least squares method. Combined with GIS base map coordinate transformation and WebGL rendering technology, a 3D pipeline network model containing material and pipe diameter attributes is generated. Finally, Gaussian kernel density estimation is applied to the fused feature vectors to generate a defect heatmap. Defect locations are visualized using HSV color mapping and Otsu thresholding. Defect type classification is implemented based on a random forest or CNN model, and a maintenance decision report containing repair suggestions and cost estimates is output. The Gaussian kernel density function of the defect heatmap is... ,bandwidth , The standard deviation of the defect location. This represents the number of defects.

7. The detection method of the drainage pipe network topology detection device based on multi-frequency acoustic-vibration coupling according to claim 6, characterized in that: The 3D model rendering uses the Three.js framework to load the GIS OpenStreetMap base map. The pipe model adopts cylindrical geometry, with the radius mapped from the pipe diameter data. The material settings include metal roughness and environmental reflection. The 3D space in the 3D model data is divided into a grid, and the center of each grid is... Calculate density value Generate density matrix ; Add a volume rendering layer to the Three.js scene, convert the density matrix F into Texture3D, render it as a semi-transparent heatmap through a shader, and overlay it on top of the pipeline model; The heat map area displays density values ​​and defect types; clicking it generates a repair report that includes location and severity.

8. The detection method of the drainage pipe network topology detection device based on multi-frequency acoustic-vibration coupling according to claim 7, characterized in that: Steps S5-S7 include: The 3D coordinates of pipeline nodes are calculated based on the TDOA algorithm, and a 3D pipeline model is generated by combining it with a GIS base map. Gaussian kernel density estimation is applied to the fused features to generate a defect heatmap, and defects are visualized by HSV color mapping and Otsu threshold segmentation. Based on the visualized defect data and historical detection data, an LSTM network is used to construct a spatiotemporal feature vector. The prediction model, combined with the risk index formula Predicting pipeline performance degradation trends, among which To predict severity, m is the material fragility index. This is the critical distance threshold. Distance between defects and inspection wells; Based on the prediction results, a maintenance plan is matched using a rule engine, and a cost formula is applied. Quantify funding needs, among which For material costs, This is the material loss coefficient. For labor costs, The variable is an artificial floating coefficient; the prediction accuracy is verified by calculating MAE and RMSE using post-repair inspection data, and the LSTM model parameters are adjusted using the backpropagation algorithm, where the MAE formula is... The RMSE formula is: , This represents the severity value after repair.

Images