Intelligent flaw detection and leakage positioning system for building pipeline

Abstract

The invention discloses a building pipeline intelligent flaw detection and leakage positioning system, which relates to the technical field of building pipeline detection and comprises a data acquisition module, a multi-modal fusion module, a defect identification module, a leakage positioning module, a digital twinborn module, a health prediction module and a decision control module. The data acquisition module is used for deploying multiple types of sensors along a building pipeline, acquiring acoustic emission signals, infrared thermal imaging images, pressure fluctuation data and fluxgate detection signals in a running state of the pipeline, and outputting original sensing data; the multi-modal fusion module is connected with the data acquisition module and is used for performing feature extraction on the original sensing data, constructing a multi-modal feature vector, dynamically adjusting each modal weight based on environmental parameters and outputting a weighted fusion feature vector; and the defect identification module is connected with the multi-modal fusion module and is used for inputting the weighted fusion feature vector into a deep neural network model.

Description

Technical Field

[0001] This invention relates to the field of building pipeline detection technology, and in particular to an intelligent flaw detection and leak location system for building pipelines. Background Technology

[0002] Building pipeline detection technology refers to a general term for various technical means used to detect and assess the operational status and structural integrity of various pipelines within buildings. Therefore, how to utilize advanced technologies to improve the intelligence and safety of building pipeline detection has become one of the most pressing issues to be addressed.

[0003] In the field of building pipeline detection, existing leak location methods are greatly affected by noise interference and have irregular propagation paths in complex pipe networks, resulting in high location errors. Furthermore, the performance of sensors fluctuates greatly under different operating conditions, leading to unstable detection results. At the same time, traditional methods focus on post-event detection and cannot predict potential faults in advance, resulting in delayed maintenance. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an intelligent flaw detection and leak location system for building pipelines to solve the problems of existing leak location methods being greatly affected by noise interference and having irregular propagation paths in complex pipeline networks, resulting in high location errors. Furthermore, the sensor performance fluctuates greatly under different operating conditions, leading to unstable detection results. At the same time, traditional methods focus on post-event detection and cannot predict potential faults in advance, resulting in delayed maintenance.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides an intelligent flaw detection and leak location system for building pipelines, comprising:

[0008] Data acquisition module, multimodal fusion module, defect identification module, leak location module, digital twin module, health prediction module, and decision control module;

[0009] The data acquisition module is used to deploy multiple types of sensors along the building pipeline to collect acoustic emission signals, infrared thermal imaging images, pressure fluctuation data and fluxgate detection signals during pipeline operation, and output raw sensor data.

[0010] The multimodal fusion module is connected to the data acquisition module and is used to extract features from the original sensing data, construct multimodal feature vectors, dynamically adjust the weights of each modality based on environmental parameters, and output a weighted fusion feature vector.

[0011] The defect identification module is connected to the multimodal fusion module and is used to input the weighted fusion feature vector into the deep neural network model to identify whether there are cracks, corrosion or perforation defects in the pipeline, and output defect type labels and confidence information.

[0012] The leak location module is connected to the data acquisition module. After identifying a leak-type defect, it uses the cross-correlation function method to calculate the signal arrival time difference between multiple sensors, and combines the nonlinear propagation model and the weighted least squares optimization method to estimate the spatial coordinates of the leak point.

[0013] The digital twin module is connected to the data acquisition module and is used to establish a three-dimensional virtual model based on the pipeline's physical structure and material properties, receive real-time monitoring data and synchronously update the model status to form a digital twin that is consistent with the actual pipeline.

[0014] The health prediction module is connected to the digital twin module and is used to train a prediction model based on historical flaw detection data and real-time status data, and output the pipeline health index and the predicted value of remaining service life.

[0015] The decision control module is connected to the defect identification module, the leak location module, and the health prediction module. It is used to generate maintenance suggestions and maintenance instructions by comprehensively considering the defect type, leak location, and health status prediction results.

[0016] As a preferred embodiment of the intelligent flaw detection and leakage location system for building pipelines described in this invention, the process of feature extraction from the original sensor data in the multimodal fusion module includes:

[0017] Wavelet packet transform is used to perform frequency band energy decomposition on acoustic emission signals, and the energy entropy values ​​of each frequency band are extracted.

[0018] The gray-level co-occurrence matrix algorithm is used to extract texture features from infrared thermal imaging images to obtain contrast, correlation and energy parameters.

[0019] Fourier transform was used to perform frequency domain analysis on the pressure fluctuation data to extract the main frequency amplitude and bandwidth information;

[0020] The sliding window root mean square algorithm is used to quantify the fluctuation intensity of the fluxgate detection signal to obtain the trend index of metal wall thickness variation.

[0021] Based on the above features, a multi-dimensional feature vector is constructed as the basic input for subsequent weighted fusion.

[0022] As a preferred embodiment of the intelligent flaw detection and leakage location system for building pipelines described in this invention, the process of dynamically adjusting the weights of each modality includes:

[0023] The current temperature, humidity, and fluid medium type are obtained using an environmental sensing unit;

[0024] Input the environmental parameters into the preset weight allocation function, the expression of which is:

[0025]

[0026] Among them, w i Let f represent the weighting coefficient corresponding to the i-th type of sensor, E represent the current set of environmental parameters, and f i (E) represents the sensitivity function of the i-th type of sensor under the current environment, β i Empirical moderating factor;

[0027] Based on the calculation results, various feature vectors are weighted and fused to generate a weighted fused feature vector.

[0028] As a preferred embodiment of the intelligent flaw detection and leak location system for building pipelines described in this invention, the deep neural network model in the defect identification module adopts the following structure and process:

[0029] The input layer receives a weighted fusion feature vector from the output of the multimodal fusion module;

[0030] The hidden layers sequentially consist of convolutional layers, pooling layers, and long short-term memory network layers, used to extract local temporal features and model long-term dependencies;

[0031] The output layer uses the Softmax function to map the final features to four classification label spaces, with the labels being: normal, crack, erosion, and perforation.

[0032] The classification result is determined by the following formula:

[0033]

[0034] Where C represents the final defect type, z k This represents the output score for the k-th class;

[0035] If the output label is crack, corrosion, or perforation, it is determined that there is a leakage defect, triggering the leakage location process.

[0036] As a preferred embodiment of the intelligent flaw detection and leak location system for building pipelines described in this invention, the process of estimating the spatial coordinates of the leak point in the leak location module includes:

[0037] After identifying the presence of a leakage defect, acoustic signals received by multiple sensors were collected.

[0038] The time difference of arrival between any two sensors is calculated using the cross-correlation function method. The cross-correlation function is defined as follows:

[0039]

[0040] Among them, s i (t), s j (t) represent the signals collected by the i-th and j-th sensors, respectively, and τ represents the time delay variable;

[0041] Take the τ corresponding to the maximum value of the cross-correlation function ij As the time difference of signal arrival between the two sensors;

[0042] Construct the time delay matrix τ = [τ] between all sensors ij ];

[0043] Combined with known sensor coordinates (x) i ,y i Given the set propagation speed v, establish a set of nonlinear positioning equations:

[0044] ||(x l ,y l )-(x i ,y i )||=v·t i i = 1, 2, ..., N;

[0045] Among them, (x l ,y l ) represents the coordinates of the leak point to be solved, t i This indicates the arrival time of the signal at the i-th sensor;

[0046] We introduce a weighted least squares optimization method to solve for the optimal leak point coordinates. The objective function is:

[0047]

[0048] Where, ω i The confidence weights for the sensors are derived from historical error statistics.

[0049] The final output is the spatial coordinates of the leak point (x). l ,y l ).

[0050] As a preferred embodiment of the intelligent flaw detection and leakage location system for building pipelines described in this invention, the specific implementation of synchronously updating the model state in the digital twin module includes:

[0051] A three-dimensional topology model is established based on the actual pipe's geometric dimensions, material properties, and connection methods.

[0052] Receive real-time monitoring data from the data acquisition module, including temperature, pressure, flow rate, and defect location information;

[0053] The finite element method was used to simulate the stress distribution and damage propagation trend inside the pipeline.

[0054] Within each update cycle, the following state update formula is used:

[0055] S t =F(S) t-1 D t );

[0056] Among them, S t S represents the state of the digital twin at time t. t-1 D represents the state at the previous moment. t Let F(.) represent the actual input data at time t, and let F(.) represent the state transition function, which is jointly modeled by the physical model and data-driven approach.

[0057] The updated digital twin status is used as data input for the health prediction module.

[0058] As a preferred embodiment of the intelligent flaw detection and leakage location system for building pipelines described in this invention, the process of training the prediction model in the health prediction module includes:

[0059] Collect historical flaw detection records and operational data to form a training sample set;

[0060] Extract the input feature vector X for each sample, including the current health index, cumulative running time, historical defect type, and number of leaks;

[0061] Define the target output Y as the health index and remaining lifespan prediction for the next k steps;

[0062] The sequence data is modeled using a Long Short-Term Memory (LSTM) network, where the hidden state update formula is:

[0063] h t =LSTM(h t-1 ,X t );

[0064] Y t =W o ·h t +b o ;

[0065] Among them, h t W represents the hidden state at time t. o b o These are the weight matrix and bias term of the output layer, respectively;

[0066] After the model is trained, it periodically receives updated status data from the digital twin module and outputs a health index H.i Compared with the predicted remaining useful life.

[0067] As a preferred embodiment of the intelligent flaw detection and leakage location system for building pipelines described in this invention, the process by which the decision control module generates maintenance suggestions and instructions includes:

[0068] The defect type output by the integrated defect identification module, the leak coordinates output by the leak location module, and the health index and remaining service life prediction value output by the health prediction module are combined.

[0069] A three-tiered early warning mechanism has been established, including:

[0070] When the health index is below the threshold T1 and no leakage has occurred, a "continuous monitoring" instruction is generated.

[0071] When the health index falls below the threshold T2 or a minor leak is detected, a "Schedule Inspection" instruction is generated.

[0072] When the health index falls below the threshold T3 or a serious leak is detected, an "immediate repair" command is generated, and valves in the designated area are shut down in conjunction with the command.

[0073] The maintenance instructions are sent to the remote control terminal through the communication interface, and the operation log is saved for traceability.

[0074] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements any step of the intelligent flaw detection and leakage location system for building pipelines as described in the first aspect of the present invention.

[0075] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the intelligent flaw detection and leakage location system for building pipelines as described in the first aspect of the present invention.

[0076] The beneficial effects of this invention are as follows: By inputting the weighted and fused feature vector into a deep neural network model containing convolutional layers, pooling layers, and LSTM layers, the output includes four types of defect labels—normal, crack, corrosion, and perforation—along with their confidence information. This achieves automated identification and classification of building pipeline defect types, improving detection efficiency and reducing errors caused by manual intervention. Simultaneously, by outputting confidence information, it provides a quantitative basis for subsequent decision-making, enhancing the system's intelligent diagnostic capabilities and interpretability. After identifying a leak-type defect, it collects acoustic signals from multiple sensors, calculates the time difference of arrival, constructs a time delay matrix, and estimates the leak point coordinates using a nonlinear propagation model and weighted least squares method. This effectively solves the positioning deviation problem caused by noise interference and irregular propagation paths in traditional TDOA positioning methods, achieving high-precision and rapid positioning of the leak point, providing accurate coordinate support for on-site maintenance, and significantly shortening fault diagnosis time. Attached Figure Description

[0077] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0078] Figure 1 This is a schematic diagram of the connection of the intelligent flaw detection and leakage location system for building pipelines in Example 1.

[0079] Figure 2 This is an architecture diagram of the intelligent flaw detection and leakage location system for building pipelines in Example 1.

[0080] Figure 3 This is a neural network structure diagram of the defect identification module in Example 1. Detailed Implementation

[0081] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0082] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0083] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0084] Example, refer to Figure 1 , Figure 2 and Figure 3 This embodiment of the invention provides an intelligent flaw detection and leak location system for building pipelines, comprising:

[0085] Data acquisition module, multimodal fusion module, defect identification module, leak location module, digital twin module, health prediction module, and decision control module;

[0086] The data acquisition module is used to deploy various types of sensors along the building pipeline to collect acoustic emission signals, infrared thermal imaging images, pressure fluctuation data and fluxgate detection signals during pipeline operation, and output raw sensor data.

[0087] Furthermore, the process of feature extraction from the raw sensor data in the multimodal fusion module includes:

[0088] Wavelet packet transform is used to perform frequency band energy decomposition on acoustic emission signals, and the energy entropy values ​​of each frequency band are extracted.

[0089] The gray-level co-occurrence matrix algorithm is used to extract texture features from infrared thermal imaging images to obtain contrast, correlation and energy parameters.

[0090] Fourier transform was used to perform frequency domain analysis on the pressure fluctuation data to extract the main frequency amplitude and bandwidth information;

[0091] The sliding window root mean square algorithm is used to quantify the fluctuation intensity of the fluxgate detection signal to obtain the trend index of metal wall thickness variation.

[0092] Based on the above features, a multi-dimensional feature vector is constructed as the basic input for subsequent weighted fusion.

[0093] It should be noted that the feature extraction process of the original sensor data in the multimodal fusion module uses wavelet packet transform, gray-level co-occurrence matrix, Fourier transform and sliding window root mean square algorithm to extract key features with physical and statistical significance from acoustic emission signals, infrared images, pressure fluctuations and fluxgate detection signals respectively. The process unifies different types of sensor data into feature vectors that can be fused and analyzed, which solves the problem that multi-source heterogeneous data is difficult to directly compare and integrate, and provides a structured input basis for subsequent weighted fusion.

[0094] The multimodal fusion module, connected to the data acquisition module, is used to extract features from the raw sensor data, construct multimodal feature vectors, dynamically adjust the weights of each modality based on environmental parameters, and output a weighted fusion feature vector.

[0095] Furthermore, the process of dynamically adjusting the weights of each modality includes:

[0096] The current temperature, humidity, and fluid medium type are obtained using an environmental sensing unit;

[0097] Input the environmental parameters into the preset weight allocation function, the expression of which is:

[0098]

[0099] Among them, w i Let f represent the weighting coefficient corresponding to the i-th type of sensor, E represent the current set of environmental parameters, and f i (E) represents the sensitivity function of the i-th type of sensor under the current environment, β i Empirical moderating factor;

[0100] Based on the calculation results, the various feature vectors are weighted and fused to generate a weighted fused feature vector;

[0101] It should be noted that the process of dynamically adjusting the weights of each mode is based on automatically adjusting the contribution ratio of each sensor mode according to the current environmental parameters. The design of the weight allocation function takes into account the sensitivity differences of different sensors under different operating conditions. When the infrared image may be severely distorted in a high-temperature environment, the system automatically reduces its weight and increases the participation ratio of other modes. The mechanism effectively improves the system's adaptability in complex environments and the stability of defect identification.

[0102] The defect identification module, connected to the multimodal fusion module, is used to input the weighted fusion feature vector into the deep neural network model to identify whether there are cracks, corrosion or perforation defects in the pipeline, and output defect type labels and confidence information.

[0103] Furthermore, the deep neural network model in the defect identification module adopts the following structure and process:

[0104] The input layer receives a weighted fusion feature vector from the output of the multimodal fusion module;

[0105] The hidden layers sequentially consist of convolutional layers, pooling layers, and long short-term memory network layers, used to extract local temporal features and model long-term dependencies;

[0106] The output layer uses the Softmax function to map the final features to four classification label spaces, with the labels being: normal, crack, erosion, and perforation.

[0107] The classification result is determined by the following formula:

[0108]

[0109] Where C represents the final defect type, z k This represents the output score for the k-th class;

[0110] If the output label is crack, corrosion or perforation, it is determined that there is a leakage defect and the leakage location process is triggered.

[0111] It should be noted that the deep neural network model combines the advantages of convolutional layers, pooling layers, and long short-term memory network layers, and can simultaneously extract local spatial features and temporal series dependencies. It is suitable for processing multimodal fusion data with spatiotemporal characteristics generated during the operation of building pipelines. The Softmax classifier outputs four types of defect labels and their confidence information, which enables the system to have high classification accuracy and interpretability, providing a reliable basis for judgment in subsequent processes.

[0112] The leak location module, connected to the data acquisition module, is used to calculate the signal arrival time difference between multiple sensors using the cross-correlation function method after a leak-type defect is identified. It then estimates the spatial coordinates of the leak point by combining a nonlinear propagation model with a weighted least squares optimization method.

[0113] Furthermore, the process of estimating the spatial coordinates of the leak point in the leak location module includes:

[0114] After identifying the presence of a leakage defect, acoustic signals received by multiple sensors were collected.

[0115] The time difference of arrival between any two sensors is calculated using the cross-correlation function method. The cross-correlation function is defined as follows:

[0116]

[0117] Among them, s i (t), s j (t) represent the signals collected by the i-th and j-th sensors, respectively, and τ represents the time delay variable;

[0118] Take the τ corresponding to the maximum value of the cross-correlation function ij As the time difference of signal arrival between the two sensors;

[0119] Construct the time delay matrix τ = [τ] between all sensors ij ];

[0120] Combined with known sensor coordinates (x) i ,y i Given the set propagation speed v, establish a set of nonlinear positioning equations:

[0121] ||(x l ,y l )-(x i ,y i )||=v·t i i = 1, 2, ..., N;

[0122] Among them, (x l ,y l ) represents the coordinates of the leak point to be solved, t i This indicates the arrival time of the signal at the i-th sensor;

[0123] We introduce a weighted least squares optimization method to solve for the optimal leak point coordinates. The objective function is:

[0124]

[0125] Where, ω i The confidence weights for the sensors are derived from historical error statistics.

[0126] The final output is the spatial coordinates of the leak point (x). l ,y l );

[0127] It should be noted that the leak location module uses the cross-correlation function method to calculate the signal arrival time difference between multiple sensors, and combines the nonlinear propagation model and the weighted least squares optimization method to estimate the coordinates of the leak point. The method fully considers the complexity of the signal propagation path in the underground pipe network and the inconsistency of sensor measurement errors. By introducing weight factors to optimize the solution process, the accuracy and robustness of leak location are improved, making it particularly suitable for the rapid location of leak points in complex urban underground environments.

[0128] The digital twin module, connected to the data acquisition module, is used to build a three-dimensional virtual model based on the pipeline's physical structure and material properties, receive real-time monitoring data and synchronously update the model's status, forming a digital twin that is consistent with the actual pipeline.

[0129] Furthermore, the specific implementation of synchronously updating the model state in the digital twin module includes:

[0130] A three-dimensional topology model is established based on the actual pipe's geometric dimensions, material properties, and connection methods.

[0131] Receive real-time monitoring data from the data acquisition module, including temperature, pressure, flow rate, and defect location information;

[0132] The finite element method was used to simulate the stress distribution and damage propagation trend inside the pipeline.

[0133] Within each update cycle, the following state update formula is used:

[0134] S t =F(S) t-1 D t );

[0135] Among them, S t S represents the state of the digital twin at time t. t-1 D represents the state at the previous moment. t Let F(.) represent the actual input data at time t, and let F(.) represent the state transition function, which is jointly modeled by the physical model and data-driven approach.

[0136] The updated digital twin status is used as data input for the health prediction module;

[0137] It should be noted that the digital twin module establishes a three-dimensional virtual model consistent with the actual pipeline and uses the finite element simulation method to simulate the internal stress distribution and damage propagation trend, thereby achieving a high-fidelity mapping of the physical pipeline state. Combined with real-time monitoring data for periodic updates, the digital twin always reflects the real pipeline operating status. It can not only be used as data input for the health prediction module, but also provides a visual management platform for operation and maintenance personnel, improving the overall intelligent perception and decision-making capabilities of the system.

[0138] The health prediction module, connected to the digital twin module, is used to train a prediction model based on historical flaw detection data and real-time status data, and outputs the pipeline health index and remaining service life prediction values.

[0139] Furthermore, the process of training the prediction model in the health prediction module includes:

[0140] Collect historical flaw detection records and operational data to form a training sample set;

[0141] Extract the input feature vector X for each sample, including the current health index, cumulative running time, historical defect type, and number of leaks;

[0142] Define the target output Y as the health index and remaining lifespan prediction for the next k steps;

[0143] The sequence data is modeled using a Long Short-Term Memory (LSTM) network, where the hidden state update formula is:

[0144] h t =LSTM(h t-1 ,X t );

[0145] Y t =W o ·h t +b o ;

[0146] Among them, ht W represents the hidden state at time t. o b o These are the weight matrix and bias term of the output layer, respectively;

[0147] After the model is trained, it periodically receives updated status data from the digital twin module and outputs a health index H. i Compared with the predicted remaining useful life;

[0148] It should be noted that the health prediction module constructs a training sample set by collecting historical flaw detection records and operational data, and uses an LSTM model to predict the pipeline health index and remaining service life. The model can capture the evolution pattern of pipeline deterioration, discover potential risks in advance, realize the transformation from passive maintenance to proactive maintenance, significantly improve the safety and operation and maintenance efficiency of building pipeline systems, and has good engineering application prospects.

[0149] The decision control module connects the defect identification module, the leak location module, and the health prediction module. It is used to generate maintenance suggestions and maintenance instructions by comprehensively considering the defect type, leak location, and health status prediction results.

[0150] Furthermore, the process by which the decision control module generates maintenance suggestions and instructions includes:

[0151] The defect type output by the integrated defect identification module, the leak coordinates output by the leak location module, and the health index and remaining service life prediction value output by the health prediction module are combined.

[0152] A three-tiered early warning mechanism has been established, including:

[0153] When the health index is below the threshold T1 and no leakage has occurred, a "continuous monitoring" instruction is generated.

[0154] When the health index falls below the threshold T2 or a minor leak is detected, a "Schedule Inspection" instruction is generated.

[0155] When the health index falls below the threshold T3 or a serious leak is detected, an "immediate repair" command is generated, and valves in the designated area are shut down in conjunction with the command.

[0156] Maintenance instructions are sent to the remote control terminal via the communication interface, and operation logs are saved for traceability.

[0157] It should be noted that the decision control module generates maintenance suggestions and instructions by comprehensively considering the defect type, leakage location, and health status prediction results. It sets up a three-level early warning mechanism, which corresponds to operation strategies such as continuous monitoring, scheduled inspections, and immediate repairs. The module realizes linkage control of key equipment through a remote communication interface, such as automatically closing valves in designated areas, which helps prevent the escalation of accidents, ensures the safe operation of building facilities, and improves the intelligence level and emergency response capability of the entire system.

[0158] This embodiment also provides a computer device suitable for a building pipeline intelligent flaw detection and leak location system, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the building pipeline intelligent flaw detection and leak location system proposed in the above embodiment.

[0159] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0160] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the intelligent flaw detection and leak location system for building pipelines as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0161] In summary, this invention, by inputting the weighted and fused feature vectors into a deep neural network model containing convolutional, pooling, and LSTM layers, outputs four types of defect labels—normal, cracked, corroded, and perforated—along with their confidence information. This achieves automated identification and classification of defect types in building pipelines, improving detection efficiency and reducing errors caused by manual intervention. Simultaneously, by outputting confidence information, it provides a quantitative basis for subsequent decision-making, enhancing the system's intelligent diagnostic capabilities and interpretability. After identifying a leak-type defect, it collects acoustic signals from multiple sensors, calculates the time difference of arrival, constructs a time delay matrix, and estimates the leak point coordinates using a nonlinear propagation model and weighted least squares method. This effectively solves the positioning deviation problem caused by noise interference and irregular propagation paths in traditional TDOA positioning methods, achieving high-precision and rapid positioning of the leak point, providing accurate coordinate support for on-site maintenance, and significantly shortening fault diagnosis time.

[0162] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A smart flaw detection and leak location system for building pipelines, characterized in that: include: Data acquisition module, multimodal fusion module, defect identification module, leak location module, digital twin module, health prediction module, and decision control module; The data acquisition module is used to deploy multiple types of sensors along the building pipeline to collect acoustic emission signals, infrared thermal imaging images, pressure fluctuation data and fluxgate detection signals during pipeline operation, and output raw sensor data. The multimodal fusion module is connected to the data acquisition module and is used to extract features from the original sensing data, construct multimodal feature vectors, dynamically adjust the weights of each modality based on environmental parameters, and output a weighted fusion feature vector. The defect identification module is connected to the multimodal fusion module and is used to input the weighted fusion feature vector into the deep neural network model to identify whether there are cracks, corrosion or perforation defects in the pipeline, and output defect type labels and confidence information. The leak location module is connected to the data acquisition module. After identifying a leak-type defect, it uses the cross-correlation function method to calculate the signal arrival time difference between multiple sensors, and combines the nonlinear propagation model and the weighted least squares optimization method to estimate the spatial coordinates of the leak point. The digital twin module is connected to the data acquisition module and is used to establish a three-dimensional virtual model based on the pipeline's physical structure and material properties, receive real-time monitoring data and synchronously update the model status to form a digital twin that is consistent with the actual pipeline. The health prediction module is connected to the digital twin module and is used to train a prediction model based on historical flaw detection data and real-time status data, and output the pipeline health index and the predicted value of remaining service life. The decision control module is connected to the defect identification module, the leak location module, and the health prediction module. It is used to generate maintenance suggestions and maintenance instructions by comprehensively considering the defect type, leak location, and health status prediction results.

2. The intelligent flaw detection and leakage location system for building pipelines as described in claim 1, characterized in that: The process of feature extraction from the original sensing data in the multimodal fusion module includes: Wavelet packet transform is used to perform frequency band energy decomposition on acoustic emission signals, and the energy entropy values ​​of each frequency band are extracted. The gray-level co-occurrence matrix algorithm is used to extract texture features from infrared thermal imaging images to obtain contrast, correlation and energy parameters. Fourier transform was used to perform frequency domain analysis on the pressure fluctuation data to extract the main frequency amplitude and bandwidth information; The sliding window root mean square algorithm is used to quantify the fluctuation intensity of the fluxgate detection signal to obtain the trend index of metal wall thickness variation. Based on the above features, a multi-dimensional feature vector is constructed as the basic input for subsequent weighted fusion.

3. The intelligent flaw detection and leakage location system for building pipelines as described in claim 2, characterized in that: The process of dynamically adjusting the weights of each modality includes: The current temperature, humidity, and fluid medium type are obtained using an environmental sensing unit; Input the environmental parameters into the preset weight allocation function, the expression of which is: Among them, w i Let f represent the weighting coefficient corresponding to the i-th type of sensor, E represent the current set of environmental parameters, and f i (E) represents the sensitivity function of the i-th type of sensor under the current environment, β i Empirical moderating factor; Based on the calculation results, various feature vectors are weighted and fused to generate a weighted fused feature vector.

4. The intelligent flaw detection and leakage location system for building pipelines as described in claim 3, characterized in that: The deep neural network model in the defect identification module adopts the following structure and process: The input layer receives a weighted fusion feature vector from the output of the multimodal fusion module; The hidden layers sequentially consist of convolutional layers, pooling layers, and long short-term memory network layers, used to extract local temporal features and model long-term dependencies; The output layer uses the Softmax function to map the final features to four classification label spaces, with the labels being: normal, crack, erosion, and perforation. The classification result is determined by the following formula: Where C represents the final defect type, z k This represents the output score for the k-th class; If the output label is crack, corrosion, or perforation, it is determined that there is a leakage defect, triggering the leakage location process.

5. The intelligent flaw detection and leakage location system for building pipelines as described in claim 4, characterized in that: The process of estimating the spatial coordinates of the leak point in the leak location module includes: After identifying the presence of a leakage defect, acoustic signals received by multiple sensors were collected. The time difference of arrival between any two sensors is calculated using the cross-correlation function method. The cross-correlation function is defined as follows: Among them, s i (t), s j (t) represent the signals collected by the i-th and j-th sensors, respectively, and τ represents the time delay variable; Take the τ corresponding to the maximum value of the cross-correlation function ij As the time difference of signal arrival between the two sensors; Construct the time delay matrix τ = [τ] between all sensors ij ]; Combined with known sensor coordinates (x) i ,y i Given the set propagation speed v, establish a set of nonlinear positioning equations: ||(x l ,y l )-(x i ,y i )||=v·t i ,i=1,2,...,N; Among them, (x l ,y l ) represents the coordinates of the leak point to be solved, t i This indicates the arrival time of the signal at the i-th sensor; We introduce a weighted least squares optimization method to solve for the optimal leak point coordinates. The objective function is: Where, ω i The confidence weights for the sensors are derived from historical error statistics. The final output is the spatial coordinates of the leak point (x). l ,y l ).

6. The intelligent flaw detection and leakage location system for building pipelines as described in claim 5, characterized in that: The specific implementation of synchronously updating the model state in the digital twin module includes: A three-dimensional topology model is established based on the actual pipe's geometric dimensions, material properties, and connection methods. Receive real-time monitoring data from the data acquisition module, including temperature, pressure, flow rate, and defect location information; The finite element method was used to simulate the stress distribution and damage propagation trend inside the pipeline. Within each update cycle, the following state update formula is used: S t =F(S t-1 ,D t ); Among them, S t S represents the state of the digital twin at time t. t-1 D represents the state at the previous moment. t Let F(.) represent the actual input data at time t, and let F(.) represent the state transition function, which is jointly modeled by the physical model and data-driven approach. The updated digital twin status is used as data input for the health prediction module.

7. The intelligent flaw detection and leakage location system for building pipelines as described in claim 6, characterized in that: The process of training the prediction model in the health prediction module includes: Collect historical flaw detection records and operational data to form a training sample set; Extract the input feature vector X for each sample, including the current health index, cumulative running time, historical defect type, and number of leaks; Define the target output Y as the health index and remaining lifespan prediction for the next k steps; The sequence data is modeled using a Long Short-Term Memory (LSTM) network, where the hidden state update formula is: h t =LSTM(h t-1 ,X t ); Y t =W o ·h t +b o ; Among them, h t W represents the hidden state at time t. o b o These are the weight matrix and bias term of the output layer, respectively; After the model is trained, it periodically receives updated status data from the digital twin module and outputs a health index H. i Compared with the predicted remaining useful life.

8. The intelligent flaw detection and leakage location system for building pipelines as described in claim 7, characterized in that: The process by which the decision control module generates maintenance suggestions and instructions includes: The defect type output by the integrated defect identification module, the leak coordinates output by the leak location module, and the health index and remaining service life prediction value output by the health prediction module are combined. A three-tiered early warning mechanism has been established, including: When the health index is below the threshold T1 and no leakage has occurred, a "continuous monitoring" instruction is generated. When the health index falls below the threshold T2 or a minor leak is detected, a "Schedule Inspection" instruction is generated. When the health index falls below the threshold T3 or a serious leak is detected, an "Immediate Repair" command is generated, and valves in the designated area are shut down. The maintenance instructions are sent to the remote control terminal through the communication interface, and the operation log is saved for traceability.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the intelligent flaw detection and leakage location system for building pipelines as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the intelligent flaw detection and leakage location system for building pipelines as described in any one of claims 1 to 8.

Images