A multi-sensor fusion-based building pipeline leakage and aging intelligent identification system

Abstract

The application discloses a kind of building pipeline leakage and aging intelligent identification system based on multi-sensing fusion, it is related to building engineering intelligent monitoring and health diagnosis technical field, including multi-sensing array unit, edge computing unit, cloud data processing unit and visual early warning unit.The application multi-sensing array integrates five complementary sensors;Edge computing realizes local data preprocessing and preliminary decision-making, response time≤1s;Cloud end uses improved D-S evidence theory fusion algorithm, in combination with machine learning model realizes leakage accurate identification and aging degree quantitative evaluation.The application leakage detection accuracy≥96%, false alarm rate≤4%, can detect the flow of no less than 0.1mL / min Microleakage, 3~6 months earlier than traditional technology to find hidden trouble, suitable for all kinds of building pipeline full life cycle intelligent monitoring and predictive maintenance.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring and health diagnosis technology for building engineering, and more specifically to the field of intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion. Background Technology

[0002] Building pipeline systems are a crucial component of building infrastructure, responsible for the core function of transporting energy media such as water, gas, and heat. With the rapid advancement of urbanization and the continuous increase in the stock of existing buildings, a large number of building pipelines have entered their aging phase. According to statistics from the Ministry of Housing and Urban-Rural Development, over 40% of existing urban buildings in my country were constructed before 2000. The pipeline systems in these buildings generally suffer from low design standards, inconsistent construction quality, and inadequate maintenance, leading to frequent accidents such as leaks and ruptures.

[0003] Pipeline leaks not only waste significant amounts of water resources but also cause a series of problems such as wall mold, ground subsidence, and structural corrosion, seriously affecting the lifespan of buildings and residential safety. For gas pipelines, leaks can also lead to major safety accidents such as explosions and poisoning, causing casualties and property damage. Traditional pipeline maintenance methods mainly rely on regular manual inspections, which have many limitations: they can only detect obvious leaks and cannot detect hidden pipelines or early micro-leaks; the inspection cycle is long, and the rate of missed detection is high; they depend on the experience of the inspectors, are highly subjective, and make it difficult to form standardized assessment results.

[0004] In recent years, with the development of sensor technology and Internet of Things (IoT) technology, automatic monitoring technology has been gradually applied to the field of building pipeline monitoring. However, existing automatic monitoring systems still face the following technical bottlenecks: 1) Insufficient early micro-leakage detection capability: Existing sensors generally have low sensitivity, typically only able to detect leaks with a flow rate greater than 10 mL / min, by which time the leak has already caused some damage. For pipelines concealed within walls and floors, leaks often take months or even years to be discovered, resulting in incalculable losses.

[0005] 2) Low detection accuracy and high false alarm rate of single sensors: Existing systems mostly use a single type of sensor for detection. Temperature and humidity sensors are easily affected by fluctuations in ambient temperature and humidity, damp walls, and other factors, with a false alarm rate as high as 45% or more; acoustic sensors cannot distinguish between leakage sounds and environmental noises such as water flow sounds, footsteps, and equipment vibration sounds, and can hardly work properly in complex building environments; flow sensors can only detect large leaks and require the destruction of pipelines for installation, which is difficult to carry out.

[0006] 3) Lack of quantitative assessment capability for aging: Existing systems can only passively detect leaks that have already occurred, and cannot monitor and quantitatively assess the aging conditions of pipelines in real time, such as corrosion, scaling, wall thinning, and material fatigue. Pipeline aging is a gradual process, and when leaks occur, the pipeline is often already severely damaged and can only be replaced, resulting in high maintenance costs.

[0007] 4) Inadequate data processing architecture: Traditional centralized cloud processing architectures require all sensor data to be transmitted to the cloud for processing, resulting in problems such as high data transmission latency, high bandwidth consumption, and strong network dependence. This architecture cannot meet the requirements of applications that require real-time early warning.

[0008] 5) Insufficient multi-source data fusion capability: Data from different types of sensors have different characteristics and advantages. For example, temperature and humidity sensors are sensitive to liquid leakage, acoustic sensors are sensitive to gas leakage, and electrochemical sensors are sensitive to corrosion processes. However, existing systems lack effective multi-source data fusion mechanisms, failing to fully utilize the complementarity of multi-source data to improve detection accuracy and reliability.

[0009] To address the shortcomings of existing technologies, this invention proposes an intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion. By integrating multiple complementary sensors, adopting a distributed architecture combining edge computing and cloud computing, and utilizing improved multi-sensor fusion algorithms and machine learning models, it achieves early and accurate detection of building pipeline leakage and quantitative assessment of aging, effectively solving many problems existing in the prior art. Summary of the Invention

[0010] The purpose of this invention is to address the technical problems in existing technologies, such as the inability to detect early micro-leakage, high false alarm rate of single sensors, inability to quantitatively assess aging, slow response speed, and insufficient data fusion capabilities. This invention provides an intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion, enabling intelligent monitoring and management of building pipelines throughout their entire lifecycle.

[0011] To achieve the above objectives, the present invention specifically adopts the following technical solution: This invention provides a building pipeline leakage and aging intelligent identification system based on multi-sensor fusion, including a multi-sensor array unit, an edge computing unit, a cloud data processing unit, and a visualization early warning unit; The multi-sensor array unit includes multiple sensor nodes distributed along the building pipelines. Each sensor node integrates a micro-leakage detection module, an acoustic vibration detection module, a temperature and humidity detection module, an electrochemical corrosion detection module, a high-definition image acquisition module, and a wireless communication module. The edge computing unit includes a data preprocessing module, a feature extraction module, a local decision-making module, a data compression and transmission module, and a local storage module; The cloud-based data processing unit includes a multi-sensor fusion module, a leakage identification module, an aging assessment module, a prediction and early warning module, and a database module. The visualization and early warning unit includes a 3D visualization module, a hierarchical early warning module, and a report generation module.

[0012] In one embodiment, the microleakage detection module employs a flexible humidity sensor array based on a multi-walled carbon nanotube / polydimethylsiloxane (MWCNT / PDMS) composite material. The flexible humidity sensor array consists of 8×8 independent sensing units, each with a size of 2mm×2mm, a substrate thickness of 0.1mm, a sensitivity ≥0.5pF / %RH, a response time ≤2s, and a recovery time ≤5s. It is capable of detecting microleakage with a flow rate of not less than 0.1mL / min. In one implementation, the multi-sensor fusion module employs an improved DS evidence theory fusion algorithm, with the following specific steps: Building a recognition framework ,in Indicates "no leakage". This indicates "minor leakage". This indicates "moderate leakage". This indicates "serious leakage"; The detection results of each sensor are normalized to obtain the basic probability allocation function. ,in These correspond to the micro-leakage detection module, acoustic vibration detection module, temperature and humidity detection module, electrochemical corrosion detection module, and high-definition image acquisition module, respectively. Introducing the credibility factor of evidence The basic probability assignment function is modified as follows: ; in, The sensor is dynamically adjusted based on its historical detection accuracy and current ambient temperature and humidity conditions, with a value range of [0,1]. The modified basic probability assignment function is fused using the Dempster combination rule: ; in, The conflict coefficient,

[0013] Based on the fused basic probability assignment function, the final decision is made using the maximum membership principle.

[0014] In one implementation, the pipeline remaining service life assessment model established by the aging assessment module is as follows: ; RUL stands for Remaining Service Life of the Pipeline. For comprehensive correction factors, This represents the initial wall thickness of the pipeline. The minimum wall thickness allowed for the pipeline. The average corrosion rate; The average corrosion rate Corrosion rate measured in real time The calculation yielded: ; The comprehensive correction coefficient Considering the effects of pipeline material, operating pressure, medium temperature and pH value, the value range is 0.8~1.2.

[0015] In one implementation, the local decision-making module performs preliminary anomaly detection based on a quantized and compressed MobileNetV2 neural network model. The model size is ≤5MB, the single-frame feature data processing time is ≤100ms, and the overall response time of the system from data acquisition to local early warning is ≤1s.

[0016] In one implementation, the 3D visualization module is developed based on the Unity3D engine, supports importing building information models in the IFC standard format, identifies pipeline operating status on the model with different colors, and supports pipeline cross-section, monitoring point data query, and historical curve display.

[0017] In one implementation, the graded early warning module divides the early warning into four levels: blue warning indicates slight aging with no risk of leakage, yellow warning indicates moderate aging with potential leakage risk, orange warning indicates severe aging with a high risk of leakage, and red warning indicates leakage has occurred or is about to break. Different levels of early warning are sent through various means such as SMS, email, APP push and sound and light alarms.

[0018] In one embodiment, the sensor node adopts an integrated waterproof encapsulation design with an IP67 protection rating; it uses the LoRaWAN communication protocol, operates at a frequency of 433MHz, and has a communication distance of ≥1km; it has a built-in 3.6V lithium-thionyl chloride battery with a capacity of 19A. h, in the working mode of collecting data once every 15 minutes, the battery life is ≥5 years.

[0019] In one embodiment, the acoustic vibration detection module employs a composite detection structure of a MEMS accelerometer and a piezoelectric ceramic sensor. The MEMS accelerometer has a frequency response range of 0.1Hz to 10kHz and a resolution of 1μg, and is used to collect low-frequency vibration signals of the pipeline. The piezoelectric ceramic sensor has a frequency response range of 1kHz to 100kHz and a sensitivity of 50mV / g, and is used to collect high-frequency leakage sound signals. In one embodiment, the electrochemical corrosion detection module employs a three-electrode system, including a working electrode, a reference electrode, and an auxiliary electrode. The working electrode is made of the same material as the pipeline under test and is in close contact with the outer wall of the pipeline through conductive adhesive. The reference electrode is a saturated calomel electrode with a potential stability of ≤±1mV. The auxiliary electrode is a platinum electrode. The corrosion current density is measured by the linear polarization resistance method, with a detection accuracy of ≤0.1μA / cm².

[0020] The beneficial effects of this invention are as follows: 1. Strong early micro-leakage detection capability: The flexible humidity sensor array based on multi-walled carbon nanotube / polydimethylsiloxane (MWCNT / PDMS) composite material can detect micro-leakage with a flow rate of not less than 0.1 mL / min, which can detect potential leakage 3 to 6 months earlier than traditional sensors, effectively avoiding the expansion of damage caused by leakage.

[0021] 2. High detection accuracy and low false alarm rate: It integrates five complementary sensors and uses an improved DS evidence theory fusion algorithm to fuse multi-source data, making full use of the advantages of different sensors, improving the leakage detection accuracy to over 96% and reducing the false alarm rate to below 4%.

[0022] 3. Accurate quantitative assessment of aging degree: The three-electrode system electrochemical corrosion detection module monitors the corrosion rate of pipelines in real time and establishes a pipeline remaining service life assessment model based on corrosion rate integral. It can accurately quantify and assess the aging degree and remaining service life of pipelines, realizing the transformation from "post-maintenance" to "predictive maintenance".

[0023] 4. Fast response speed: Adopting a distributed architecture that combines edge computing and cloud computing, the edge computing unit can perform preliminary analysis and decision-making locally. The overall response time of the system from data collection to local early warning is ≤1 second, realizing real-time early warning and effectively avoiding the occurrence of accidents.

[0024] 5. Low system power consumption and long service life: With low power consumption design and sleep wake-up mechanism, the battery life of sensor nodes can reach more than 5 years, which greatly reduces the system maintenance cost.

[0025] 6. High degree of visualization and simple operation: Based on BIM technology, the three-dimensional visualization of pipeline operation status is realized, which allows managers to intuitively understand the distribution and operation status of pipelines, improving management efficiency by more than 50%.

[0026] 7. Easy installation and no need to damage pipelines: All sensor nodes are installed by adhesive bonding, which does not require damage to the pipeline structure, making construction simple and having little impact on the normal use of the building.

[0027] 8. Wide range of applications: Applicable to various building pipelines such as water supply and drainage, gas, HVAC, etc., and can meet the monitoring needs of different buildings. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0029] Figure 1 This is a structural diagram of an intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion. Figure 2 This is a schematic diagram of the structure of a multi-sensor array unit; Figure 3 This is a schematic diagram of the edge computing unit structure; Figure 4 This is a schematic diagram of the structure of a cloud-based data processing unit; Figure 5 This is a schematic diagram of the structure of the visual early warning unit; Figure 6 This is a flowchart of the DS evidence theory fusion algorithm; Reference numerals: 1. Multi-sensor array unit; 2. Edge computing unit; 3. Cloud data processing unit; 4. Visual early warning unit; 101. Micro-leakage detection module; 102. Acoustic vibration detection module; 103. Temperature and humidity detection module; 104. Electrochemical corrosion detection module; 105. High-definition image acquisition module; 106. Wireless communication module; 201. Data Preprocessing Module; 202. Feature Extraction Module; 203. Local Decision Module; 204. Data Compression and Transmission Module; 205. Local Storage Module; 301. Multi-sensor fusion module; 302. Leakage identification module; 303. Aging assessment module; 304. Prediction and early warning module; 305. Database module; 401. 3D visualization module; 402. Hierarchical early warning module; 403. Report generation module. Detailed Implementation

[0030] To make the technical problems, technical solutions, and technical effects of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0031] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0032] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0033] In the description of the embodiments of the present invention, it should be noted that the terms "inner", "outer", "upper", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of the invention is usually placed when in use. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting the present invention.

[0034] This invention provides an intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion, including a multi-sensor array unit 1, an edge computing unit 2, a cloud data processing unit 3, and a visualization early warning unit 4.

[0035] The multi-sensor array unit 1 includes multiple sensor nodes distributed along the building pipelines. Each sensor node integrates a micro-leakage detection module 101, an acoustic vibration detection module 102, a temperature and humidity detection module 103, an electrochemical corrosion detection module 104, a high-definition image acquisition module 105, and a wireless communication module 106.

[0036] The edge computing unit 2 includes a data preprocessing module 201, a feature extraction module 202, a local decision-making module 203, a data compression and transmission module 204, and a local storage module 205.

[0037] The cloud data processing unit 3 includes a multi-sensor fusion module 301, a leakage identification module 302, an aging assessment module 303, a prediction and early warning module 304, and a database module 305.

[0038] The visualization early warning unit 4 includes a three-dimensional visualization module 401, a hierarchical early warning module 402, and a report generation module 403.

[0039] The detailed introduction of each module is as follows: The multi-sensor array unit 1 adopts a distributed node arrangement, and the arrangement density of the sensor nodes is dynamically determined according to the risk level of the pipeline. High-risk areas (basement, pipe shaft, bathroom, kitchen, pipe joints, valves, diameter changes): sensor node spacing ≤ 5m; Medium-risk areas (corridors, public areas, stairwells): sensor node spacing ≤ 10m; Low-risk areas (offices, meeting rooms, storage rooms): sensor node spacing ≤ 15m; Each sensor node features an integrated waterproof encapsulation design, with the shell made of ABS engineering plastic, achieving an IP67 protection rating. This allows for long-term stable operation in humid, dusty, and temperature-fluctuating building environments. The node dimensions are 100mm × 80mm × 30mm, and the weight is ≤200g. Installation is convenient and requires no damage to the pipeline structure.

[0040] The microleakage detection module 101 employs a flexible humidity sensor array based on a multi-walled carbon nanotube / polydimethylsiloxane (MWCNT / PDMS) composite material. This composite material possesses high specific surface area, high conductivity, and good flexibility. When water molecules are adsorbed onto the surface of the composite material, it causes a change in the material's dielectric constant, resulting in a change in the sensor's capacitance. The sensor array consists of 8×8 independent sensing units, each measuring 2mm×2mm, with a substrate thickness of 0.1mm. The sensor sensitivity is ≥0.5pF / %RH, with a response time ≤2s and a recovery time ≤5s, capable of detecting microleakage with a flow rate of not less than 0.1mL / min. The sensor array is adhered to the outer wall of the pipeline using medical-grade acrylic adhesive, adaptable to pipelines of different diameters (DN15~DN300) and materials (carbon steel, stainless steel, copper, PVC, PE, PPR).

[0041] The acoustic vibration detection module 102 employs a composite detection structure combining a MEMS accelerometer and a piezoelectric ceramic sensor. The MEMS accelerometer selected is the ADXL355, a low-noise, low-power triaxial accelerometer with a frequency response range of 0.1Hz to 10kHz and a resolution of 1μg. It is used to acquire low-frequency vibration signals from pipelines to analyze their operational status and structural integrity. When fluid flows inside the pipeline, it generates vibration signals at specific frequencies. By analyzing the characteristics of these vibration signals, it is possible to determine whether there are problems such as blockage or scaling in the pipeline. The piezoelectric ceramic sensor selected is PZT-5H, a piezoelectric material with a high piezoelectric coefficient. It has a frequency response range of 1kHz to 100kHz and a sensitivity of 50mV / g, and is used to acquire high-frequency leakage sound signals. When a pipeline leaks, the fluid ejected from the leak point generates high-frequency acoustic emission signals. These signals typically have frequencies between 1kHz and 100kHz, much higher than the frequency of ambient noise. The two sensors are processed by different signal conditioning circuits, effectively covering the entire frequency range of pipeline vibration and leakage sound.

[0042] The temperature and humidity detection module 103 uses the SHT30 digital temperature and humidity sensor, a high-precision, low-power digital temperature and humidity sensor with a temperature measurement range of -40℃ to 125℃ and an accuracy of ±0.3℃; and a humidity measurement range of 0%RH to 100%RH with an accuracy of ±2%RH. The temperature and humidity detection module 103 is used to collect temperature and humidity data of the environment surrounding the pipeline, providing environmental compensation for micro-leakage detection and eliminating the influence of environmental temperature and humidity changes on the detection results. Simultaneously, the temperature and humidity data can also be used to assess the pipeline's operating environment, providing a reference for aging assessment.

[0043] The electrochemical corrosion detection module 104 employs a three-electrode system, including a working electrode, a reference electrode, and an auxiliary electrode. The working electrode is made of the same material as the pipeline under test and is in close contact with the pipeline's outer wall via conductive adhesive, ensuring good electrical contact. The reference electrode is a saturated calomel electrode (SCE), with a potential stability ≤ ±1mV, providing a stable potential reference. The auxiliary electrode is a platinum electrode with an area of ​​1cm², used for current conduction. The electrochemical corrosion detection module 104 measures the corrosion current density of the pipeline using the linear polarization resistance method (LPR), with a detection accuracy ≤ 0.1μA / cm² and a configurable measurement cycle (1h~24h). The linear polarization resistance method is a commonly used electrochemical corrosion measurement method. Its principle is to apply a small polarization potential (typically ±10mV) between the working electrode and the reference electrode, measure the polarization current, and then calculate the corrosion current density according to the Stern-Geary formula. The corrosion current density is proportional to the corrosion rate; by monitoring the corrosion current density in real time, the degree of corrosion and aging status of the pipeline can be assessed.

[0044] The high-definition image acquisition module 105 employs a 2-megapixel global shutter CMOS camera and is equipped with an 850nm infrared LED, enabling it to acquire clear images of pipeline surfaces in completely dark environments. The image acquisition module 105 is normally in sleep mode, only waking up when other sensors detect anomalies, effectively reducing system power consumption. The acquired image data is used to identify defects such as cracks, corrosion, and deformation on the pipeline surface, providing visual evidence for leakage and aging assessments.

[0045] The wireless communication module 106 adopts the LoRaWAN communication protocol, operates at a frequency of 433MHz, has a transmit power of 14dBm, and a communication distance of up to 1km, capable of penetrating 3-5 layers of concrete walls. LoRaWAN is a low-power wide-area network communication technology with advantages such as long communication distance, low power consumption, and strong anti-interference capabilities, making it very suitable for IoT applications in built-in environments. Each sensor node has a built-in 3.6V lithium-thionyl chloride battery with a capacity of 19A. Under normal operating conditions (data is collected every 15 minutes), the battery life can reach more than 5 years.

[0046] The edge computing unit 2 uses an ARM Cortex-A53 quad-core processor with a clock speed of 1.2GHz, equipped with 2GB of LPDDR4 memory and 16GB of eMMC flash memory, and runs an embedded Linux operating system. Edge computing unit 2 is wall-mounted, measuring 200mm × 150mm × 50mm, and has an IP54 protection rating, enabling it to operate stably in indoor environments for extended periods. Each edge computing unit can connect to up to 200 sensor nodes, covering a building area of ​​approximately 50,000 m².

[0047] The data preprocessing module 201 preprocesses the raw data collected by the multi-sensor array unit 1 to remove noise and interference and improve data quality. For vibration and acoustic signals, a wavelet thresholding denoising algorithm is used, with the db4 wavelet basis selected, a decomposition level of 5, and a soft thresholding function used for thresholding to effectively remove environmental noise interference. For temperature and humidity signals and electrochemical signals, a moving average filtering algorithm is used with a window size of 5 to eliminate random noise and spike interference. For image data, histogram equalization algorithm is used to improve image contrast, and 3×3 median filtering algorithm is used to remove salt and pepper noise. Feature extraction module 202 extracts multi-dimensional features from the preprocessed data, which can reflect the pipeline's operating status and health condition: Sixteen time-domain features (mean, variance, peak value, peak-to-peak value, kurtosis, margin factor, impulse factor, waveform factor, etc.), eight frequency-domain features (centroid frequency, mean square frequency, frequency variance, frequency kurtosis, etc.), and twelve time-frequency domain features (wavelet packet energy entropy) are extracted from vibration and acoustic signals. Features such as temperature change rate, humidity change rate, temperature-humidity product, and temperature-humidity difference are extracted from temperature and humidity signals.

[0048] Electrochemical signals are used to extract features such as corrosion potential, corrosion current density, polarization resistance, and corrosion rate.

[0049] Image data extraction includes grayscale histograms, grayscale co-occurrence matrix texture features (energy, contrast, correlation, entropy), edge features, etc.

[0050] The local decision module 203 performs preliminary anomaly detection on the extracted features based on the quantized and compressed MobileNetV2 neural network model. MobileNetV2 is a lightweight convolutional neural network model that uses depthwise separable convolution and inverse residual structures, and has the advantages of small model size, high speed, and high accuracy.

[0051] Model training process: The dataset consists of publicly available data and field-collected data, containing over 100,000 samples, including over 32,000 positive samples (leakage) and over 68,000 negative samples (normal), covering various situations with different materials, pipe diameters, leakage levels, and environmental conditions.

[0052] The training method adopts the transfer learning approach. The pre-trained model completes basic feature extraction training on the ImageNet dataset, and then fine-tunes it on the domain dataset.

[0053] The training parameters used the Adam optimizer, with an initial learning rate of 0.001, a batch size of 32, and 50 training epochs. An early stopping mechanism was used to prevent overfitting.

[0054] The quantization compression uses the INT8 quantization method to compress the trained model, reducing the model size from 14MB to 4.2MB and improving inference speed by more than 3 times.

[0055] After quantization and compression, the model can run in real time on the edge computing unit, with a single frame feature data processing time of ≤100ms. When the local decision module 203 detects an anomaly, it immediately triggers a local audible and visual warning and transmits the raw data and preliminary analysis results to the cloud data processing unit 3 for further verification. This "local preliminary decision + cloud-based fine analysis" architecture ensures both the real-time nature of the warning and the accuracy of the detection.

[0056] The data compression and transmission module 204 uses the LZ4 lossless compression algorithm to compress transmitted data, achieving a compression ratio of up to 3:1, effectively reducing data transmission volume and bandwidth usage. Simultaneously, it employs a breakpoint resumption mechanism; when the network is interrupted, data is temporarily stored in the local storage module 205 and automatically resumes transmission after the network is restored, ensuring no data loss. For normal data, a timed upload method is used, uploading every 15 minutes; for abnormal data, an immediate upload method is used, ensuring that the cloud receives abnormal information promptly.

[0057] The local storage module 205 uses industrial-grade eMMC flash memory, capable of storing at least 90 days of monitoring data and analysis results. It also supports data cyclic overwriting, automatically overwriting the oldest data when storage space is insufficient. The local storage module 205 also stores system configuration parameters, model parameters, and sensor calibration data, ensuring the system can operate independently during network outages.

[0058] The cloud data processing unit 3 uses an Alibaba Cloud ECS server cluster, which supports elastic scaling and can process data from thousands of edge computing units simultaneously. The server is configured with: 8 cores and 16GB CPU, 500GB SSD hard drive, 10Mbps bandwidth, and runs Ubuntu 20.04 operating system.

[0059] The multi-sensor fusion module 301 employs a decision-level fusion algorithm that combines an improved DS evidence theory with deep learning. DS evidence theory is an effective method for handling uncertain information, capable of fusing detection results from multiple sensors to improve decision accuracy. However, traditional DS evidence theory can produce counterintuitive results when dealing with highly conflicting evidence. To address this issue, this invention introduces an evidence credibility factor to modify the basic probability allocation function before fusion.

[0060] Evidence credibility factor Specific calculation method: ; in, The historical detection accuracy of sensor i is obtained from the statistical results of the sensor's detection over the past 30 days. The environmental impact factor is calculated based on the deviation of the current ambient temperature and humidity from the standard temperature and humidity (20℃, 50%RH): ; Where T is the current ambient temperature (°C) and RH is the current ambient humidity (%RH). When environmental conditions deviate from standard conditions, Increase This reduces the reliability of the sensor.

[0061] The specific steps of the fusion algorithm are as follows: Building a recognition framework ,in Indicates "no leakage". This indicates "minor leakage". This indicates "moderate leakage". It indicates "serious leakage".

[0062] The detection results of each sensor are normalized to obtain the basic probability allocation function. ,in These represent the micro-leakage detection module, acoustic vibration detection module, temperature and humidity detection module, electrochemical corrosion detection module, and high-definition image acquisition module, respectively.

[0063] Introducing the credibility factor of evidence The basic probability assignment function is modified as follows: in, The sensor's value is dynamically adjusted based on its historical detection accuracy and current ambient temperature and humidity conditions, with a range of [0,1].

[0064] The modified basic probability assignment function is fused using the Dempster combination rule: ; in, The conflict coefficient, ; Based on the fused basic probability assignment function, the final decision is made using the maximum membership principle.

[0065] The leakage detection module 302 uses a trained ResNet50 convolutional neural network model to identify the occurrence, location, and severity of leaks based on the fused feature data. ResNet50 is a deep residual network with powerful feature extraction and classification capabilities. The model is trained on a dataset containing over 100,000 samples, and testing shows that the model's leakage detection accuracy is ≥96%, with a false alarm rate ≤4%.

[0066] The aging assessment module 303 establishes a pipeline remaining service (RUL) assessment model based on the corrosion rate integral. Pipeline aging is mainly caused by corrosion, which leads to a gradual thinning of the pipeline wall thickness. When the wall thickness decreases to the minimum allowable thickness, the pipeline will rupture. Therefore, the remaining service life of the pipeline can be calculated using the corrosion rate and wall thickness. ; in, The initial wall thickness of the pipeline (mm) This represents the minimum allowable wall thickness (mm) for the pipeline. The average corrosion rate is given in mm / year.

[0067] Average corrosion rate Corrosion rate measured in real time by electrochemical corrosion detection module 104 The calculation yielded: ; Meanwhile, the model introduces a comprehensive correction coefficient. The effects of pipeline material, operating pressure, medium temperature, pH value, and other factors on the corrosion rate were considered. ; The specific method for determining the comprehensive correction coefficient u is as follows: ; in, Material correction factors: 0.9 for carbon steel, 1.2 for stainless steel, 1.1 for copper, and 1.0 for PVC / PE / PPR; Pressure correction factor: 1.0 for operating pressure ≤ 0.6MPa, 0.95 for operating pressure < 0.6MPa ≤ 1.0MPa, and 0.9 for operating pressure > 1.0MPa; Temperature correction factor: 1.0 for medium temperature ≤ 25℃, 0.95 for medium temperature < 25℃ ≤ 40℃, and 0.9 for medium temperature > 40℃; pH correction factor: 1.0 for pH ≤ 8.5, and 0.9 for pH < 6.5 or pH > 8.5; Comprehensive correction coefficient The value range is 0.8 to 1.2, which was determined based on a large amount of corrosion test data and engineering practice experience. Field verification shows that the model's remaining service life assessment error is ≤10%.

[0068] The prediction and early warning module 304 is based on a Long Short-Term Memory (LSTM) time series prediction algorithm, using historical monitoring data to predict future corrosion rates and leakage risks in pipelines. LSTM is a special type of recurrent neural network capable of processing long-sequence data and exhibiting excellent predictive performance. The model can predict the pipeline's operating status for the next 1-5 years with an accuracy rate of ≥85%, providing a scientific basis for developing medium- and long-term maintenance plans.

[0069] Database module 305 adopts a hybrid database architecture of MySQL and Redis. MySQL is used to store historical monitoring data, model parameters, evaluation results, and user information, while Redis is used to cache real-time data and improve system response speed. The database supports fast data querying, statistical analysis, and backup and recovery, and can meet the needs of large-scale data storage and processing.

[0070] The 3D visualization module 401 is developed based on the Unity3D engine and supports importing Building Information Models (BIM) in the IFC standard format. It visually displays the distribution of pipelines, the location of monitoring points, and their operational status on the BIM model. Different colors represent different operational statuses: green for normal, blue for minor anomalies, yellow for moderate anomalies, orange for severe anomalies, and red for urgent anomalies. Managers can use the mouse to rotate, zoom, and pan to view the pipeline's operational status from different angles; they can perform cross-section operations on the pipeline to view its internal structure; they can click on any monitoring point to view detailed monitoring data and historical curves; and they can query basic pipeline information such as material, diameter, installation time, and design service life.

[0071] The graded early warning module 402 classifies early warnings into four levels based on the severity of leakage and aging: Blue Alert (Level IV): The pipeline shows slight signs of aging, but there is no risk of leakage. It is recommended to strengthen daily inspections.

[0072] Yellow Alert (Level III): The pipeline is moderately aged and there is a potential risk of leakage. It is recommended to develop a maintenance plan within 3 months.

[0073] Orange Alert (Level II): The pipeline is severely aged and there is a high risk of leakage. It is recommended to carry out repairs within one month.

[0074] Red Alert (Level I): The pipeline has leaked or is about to rupture. It is recommended to stop using it immediately and carry out emergency repairs.

[0075] Different levels of early warning information are sent to relevant management personnel through various means such as SMS, email, app push notifications, and audible and visual alarms to ensure timely delivery of warning information. Simultaneously, the system records the processing of all early warning information, creating a complete early warning handling log.

[0076] The report generation module 403 can automatically generate daily, weekly, monthly, and annual reports. Report content includes pipeline operation status statistics, leakage accident statistics, aging assessment results, and maintenance recommendations. Reports can be exported to multiple formats such as PDF, Excel, and Word for easy archiving and submission. Additionally, the system supports custom reports, allowing administrators to select report content and formats as needed.

[0077] Example 1 like Figures 1 to 6 As shown in the figure, this embodiment discloses an intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion, applied to the monitoring of water supply and drainage pipelines in a hospital inpatient building. It includes a multi-sensor array unit 1, an edge computing unit 2, a cloud data processing unit 3, and a visualization and early warning unit 4.

[0078] This embodiment provides the system of the present invention. The hospital inpatient building has a building area of ​​35,000 m², with 12 floors above ground and 2 floors underground, and a total water supply and drainage pipeline length of approximately 8,500 m. Due to its early construction date (built in 2005), the pipeline is severely aged, and leakage accidents have occurred many times in recent years, which has seriously affected the normal operation of the hospital.

[0079] A total of 920 sensor nodes were installed on the water supply and drainage pipelines of the hospital building, of which 580 sensor nodes were installed in high-risk areas (basement, pipe shaft, toilet, operating room) with a spacing of 3 to 5 meters. Medium-risk areas (corridors, nurses' stations): 240 sensor nodes were deployed with a spacing of 8-10m; Low-risk areas (offices, meeting rooms): 100 sensor nodes were deployed with a spacing of 12-15m; A total of 4 edge computing units were deployed, which are responsible for the data processing of sensor nodes on underground floors 1-2, above-ground floors 1-4, above-ground floors 5-8, and above-ground floors 9-12, respectively.

[0080] Since its commissioning in March 2024, the system has operated stably and reliably, detecting 18 leakage incidents, 13 of which were early-stage micro-leakages, detected 4-5 months earlier than traditional manual inspections. The system's leakage detection accuracy rate is 97.2%, with a false alarm rate of 2.8%, far superior to traditional monitoring systems.

[0081] The system also conducted a comprehensive assessment of the aging of the hospital building's water supply and drainage pipelines. The assessment results showed that: 12% of the pipelines are severely aged, with less than 3 years of remaining service life. It is recommended to replace them within 1 year. 28% of the pipelines are of moderate aging, with a remaining service life of 3 to 8 years. It is recommended to strengthen daily inspections. 60% of the pipelines show only slight signs of aging, with a remaining service life of over 8 years, and are ready for normal use.

[0082] Based on the system's evaluation results, the hospital developed a phased pipeline upgrade and renovation plan, which is expected to save more than 40% in maintenance costs and effectively avoid medical accidents and property losses caused by pipeline leaks.

[0083] Example 2 This embodiment demonstrates the application of the system of the present invention in the monitoring of gas pipelines in a commercial complex. The commercial complex has a building area of ​​120,000 m², with 6 floors above ground and 3 floors underground, and a total gas pipeline length of approximately 3,200 m. The safe operation of the gas pipeline is directly related to the lives of thousands of customers and employees within the mall; therefore, the reliability and accuracy of the monitoring system are extremely important.

[0084] A total of 450 sensor nodes were installed on the gas pipeline of the commercial complex, all of which were explosion-proof with an explosion-proof rating of ExdIIBT4Gb. The sensor nodes were mainly located at key locations such as pipe joints, valves, and pressure regulating stations, with a spacing of 2-3 meters.

Claims

1. A smart identification system for building pipeline leakage and aging based on multi-sensor fusion, characterized in that, It includes a multi-sensor array unit (1), an edge computing unit (2), a cloud data processing unit (3), and a visualization and early warning unit (4); The multi-sensor array unit (1) includes multiple sensor nodes distributed along the building pipelines. Each sensor node integrates a micro-leakage detection module (101), an acoustic vibration detection module (102), a temperature and humidity detection module (103), an electrochemical corrosion detection module (104), a high-definition image acquisition module (105), and a wireless communication module (106). The system has a built-in dual-condition adaptive switching logic. Under the condition of water supply and drainage pipelines, all modules work synchronously. Under the condition of gas pipelines, the micro-leakage detection module (101) is automatically shut down. The acoustic vibration detection module (102), temperature and humidity detection module (103), electrochemical corrosion detection module (104), and high-definition image acquisition module (105) work together to complete the leak monitoring. The edge computing unit (2) includes a data preprocessing module (201), a feature extraction module (202), a local decision-making module (203), a data compression and transmission module (204), and a local storage module (205); The cloud data processing unit (3) includes a multi-sensor fusion module (301), a leakage identification module (302), an aging assessment module (303), a prediction and early warning module (304), and a database module (305); The visualization early warning unit (4) includes a three-dimensional visualization module (401), a hierarchical early warning module (402), and a report generation module (403).

2. The intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion as described in claim 1, characterized in that, The microleakage detection module (101) adopts a flexible humidity sensor array based on multi-walled carbon nanotube / polydimethylsiloxane (MWCNT / PDMS) composite material. The flexible humidity sensor array consists of 8×8 independent sensing units, each with a size of 2mm×2mm, a substrate thickness of 0.1mm, a sensitivity ≥0.5pF / %RH, a response time ≤2s, and a recovery time ≤5s. It can detect microleakage with a flow rate of not less than 0.1mL / min.

3. The intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion as described in claim 1, characterized in that, The multi-sensor fusion module (301) employs an improved DS evidence theory fusion algorithm, the specific steps of which are as follows: Building a recognition framework ,in Indicates "no leakage". This indicates "minor leakage". This indicates "moderate leakage". This indicates "serious leakage"; The detection results of each sensor are normalized to obtain the basic probability allocation function. ,in These correspond to the micro-leakage detection module, acoustic vibration detection module, temperature and humidity detection module, electrochemical corrosion detection module, and high-definition image acquisition module, respectively. Introducing the credibility factor of evidence The basic probability assignment function is modified as follows: ; in, The sensor is dynamically adjusted based on its historical detection accuracy and current ambient temperature and humidity conditions, with a value range of [0,1]. The modified basic probability assignment function is fused using the Dempster combination rule: ; in, The conflict coefficient, Based on the fused basic probability assignment function, the final decision is made using the maximum membership principle.

4. The intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion as described in claim 1, characterized in that, The pipeline remaining service life assessment model established by the aging assessment module (303) is as follows: ; RUL stands for Remaining Service Life of the Pipeline. For comprehensive correction factors, This represents the initial wall thickness of the pipeline. The minimum wall thickness allowed for the pipeline. The average corrosion rate; The average corrosion rate Corrosion rate measured in real time The calculation yielded: ; The comprehensive correction coefficient Considering the effects of pipeline material, operating pressure, medium temperature and pH value, the value range is 0.8~1.

2.

5. The intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion as described in claim 1, characterized in that, The local decision module (203) performs preliminary anomaly detection based on the quantized and compressed MobileNetV2 neural network model. The model size is ≤5MB, the single-frame feature data processing time is ≤100ms, and the overall response time of the system from data acquisition to local early warning is ≤1s.

6. The intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion as described in claim 1, characterized in that, The three-dimensional visualization module (401) is developed based on the Unity3D engine. It supports importing building information models in the IFC standard format, marking pipeline operating status with different colors on the model, and supports pipeline sectioning, monitoring point data query and historical curve display.

7. The intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion as described in claim 1, characterized in that, The graded early warning module (402) divides the early warning into four levels: blue warning indicates slight aging with no risk of leakage, yellow warning indicates moderate aging with potential leakage risk, orange warning indicates severe aging with a high risk of leakage, and red warning indicates leakage has occurred or is about to break. Different levels of early warnings are sent through various means, including SMS, email, app push notifications, and audible and visual alarms.

8. The intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion as described in claim 1, characterized in that, The sensor node adopts an integrated waterproof encapsulation design with an IP67 protection rating; it uses the LoRaWAN communication protocol, operates at a frequency of 433MHz, and has a communication distance of ≥1km; it has a built-in 3.6V lithium-thionyl chloride battery with a capacity of 19A. h, in the working mode of collecting data once every 15 minutes, the battery life is ≥5 years.

9. The intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion as described in claim 1, characterized in that, The acoustic vibration detection module (102) adopts a composite detection structure of MEMS accelerometer and piezoelectric ceramic sensor. The frequency response range of the MEMS accelerometer is 0.1Hz~10kHz, and the resolution is 1μg. It is used to collect low-frequency vibration signals of pipelines. The frequency response range of the piezoelectric ceramic sensor is 1kHz~100kHz, and the sensitivity is 50mV / g. It is used to collect high-frequency leakage sound signals.

10. The intelligent identification system for building pipeline leakage and aging based on multi-sensor fusion according to claim 1, characterized in that, The electrochemical corrosion detection module (104) adopts a three-electrode system, including a working electrode, a reference electrode, and an auxiliary electrode. The working electrode is made of the same material as the pipeline being tested and is in close contact with the outer wall of the pipeline through conductive adhesive. The reference electrode is a saturated calomel electrode with a potential stability of ≤±1mV. The auxiliary electrode is a platinum electrode. The corrosion current density is measured by the linear polarization resistance method with a detection accuracy of ≤0.1μA / cm².

Images