A pipeline water leakage detection method and system based on multi-modal composite sensing and a storage medium

Abstract

The application discloses a pipeline water leakage detection method and system based on multi-modal composite sensing and a storage medium, and belongs to the technical field of water leakage detection. The sensor array is used to collect the incremental changes of temperature, pressure and conductance signals arranged in an array around the pipeline, and form a standardized multi-dimensional feature data set. The long short-term memory network (LSTM) is used to extract the time sequence signal features of the multi-dimensional feature data, and the convolutional neural network (CNN) is used to extract the spatial distribution features. The time sequence signal features and the spatial distribution features are fused through a feature fusion layer, and finally, based on a multi-level classifier or a multi-layer fully connected neural network, the water leakage state, the water leakage position and the water leakage severity are output. The high-dimensional feature after fusion of the application comprehensively considers the multi-physical field information of water leakage, and significantly improves the recognition ability for small leakage and complex environmental changes. The application has high robustness, and can effectively inhibit the interference of abnormal signals and environmental noise on the judgment result.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of water leakage detection, and particularly relates to a pipeline water leakage detection method and system based on multi-modal composite sensing and a storage medium. BACKGROUND

[0002] At present, for the detection of pipeline water leakage in a water supply system, the following detection technologies are used:

[0003] (1) A water leakage monitoring method based on water-absorbing and heat-releasing materials and temperature sensors, which sets a detection layer containing water-absorbing and heat-releasing materials outside the pipeline, and uses temperature sensors to monitor the local temperature change caused by water leakage to realize water leakage detection. This kind of scheme usually analyzes the temperature change in the time or space dimension to infer the water leakage position and state.

[0004] For example, the existing Chinese patent CN101472543B discloses a moisture monitoring system for monitoring moisture in one or more absorbent articles, comprising an input for receiving one or more sensor signals indicative of the presence of moisture in the absorbent article; a processor for processing the one or more sensor signals and for performing an analysis on the signals in order to characterize a moisture event occurring in the absorbent article; and a user interface for communicating with a user of the system.

[0005] (2) A water leakage detection method based on distributed optical fiber temperature sensing, which fixes the temperature measuring optical fiber on the outer layer of the pipeline, and uses the signal change when the built-in water-absorbing and heat-releasing material changes in temperature to realize real-time monitoring of pipeline water leakage. For example, the existing Chinese patent CN223105855U discloses a buried water supply pipeline leakage detection device based on distributed optical fiber sensing, which comprises a temperature measuring optical fiber, a laying device and a reaction block. The laying device is a double-layer sleeve structure, the inner layer is a water permeable layer, and the outer layer is a waterproof layer. The inner diameter of the water permeable layer is equal to the outer diameter of the water supply pipeline. A spiral optical fiber groove with equal intervals is formed between the water permeable layer and the waterproof layer, and the temperature measuring optical fiber is fixed in the spiral optical fiber groove. The reaction block is made of a material that reacts with water and releases heat, and is fixed in the water permeable layer. One reaction block is arranged between every two adjacent optical fiber grooves. The optical fiber is fixed on the surface of the pipeline to be detected by the laying device, and the other end is connected to a DTS temperature measuring host. When the water supply pipeline leaks, the reaction block in the water permeable layer reacts and releases heat, and the temperature measuring optical fiber transmits the detected signal to the DTS temperature measuring host, thereby realizing real-time monitoring of the leakage of the water supply pipeline.

[0006] (3) Temperature sensing leakage monitoring platform based on signal amplification layer, reaction layer is set outside the detection pipeline to enhance the temperature change signal, so as to improve the sensitivity and accuracy of leakage detection. For example, the existing Chinese patent CN107167263B discloses a leakage detection experimental platform for tap water pipes based on optical fiber Raman temperature sensor, which comprises a water circulation system, a signal amplification system and an optical fiber Raman temperature sensor. The water circulation system simulates the leakage of urban tap water pipes in the laboratory environment, and the temperature signal is further amplified by the conversion layer set outside the detection pipeline, so as to facilitate the detection of the optical fiber Raman temperature sensor.

[0007] (4) Water pressure sensor method: pressure sensors are arranged at the main pipeline or key nodes to monitor the pressure change in the pipeline in real time. Pressure anomalies caused by leakage are captured by the data acquisition and analysis module for leakage alarm. This scheme is suitable for overall system anomaly monitoring.

[0008] (5) Humidity sensor method: humidity sensor array is arranged outside the pipeline or potential leakage point to detect local environmental humidity rise in real time to determine leakage. The signal acquisition and alarm unit is responsible for processing data and triggering alarm.

[0009] However, the above methods have the following defects:

[0010] 1) Single sensing mode limits the accuracy of detection, and the sensitivity to small leakage is low. Absorption heat release temperature monitoring, pressure monitoring or humidity monitoring are single physical quantities. If only single characteristics such as temperature change, pressure fluctuation or humidity rise are used for leakage detection, the detection accuracy is insufficient.

[0011] On the one hand, single sensing signal is easily disturbed by environmental temperature fluctuation, water pressure variation and humidity natural change, resulting in frequent false alarms and missed alarms; on the other hand, these technologies are difficult to capture subtle signal changes caused by small leakage, especially in hidden and early leakage, it is difficult to achieve timely and accurate detection. For example, pressure sensor is difficult to capture weak pressure change caused by small flow leakage, and pressure signal lags, so the leakage position cannot be accurately positioned. Although the pressure sensor method is suitable for overall pressure anomaly alarm, the leakage positioning ability is weak. Secondly, the humidity sensor is densely arranged and the maintenance cost is high: the humidity sensor needs to be directly exposed to the humid environment, the coverage area is limited, and the sensor is easily damaged by long-term immersion, resulting in complex system maintenance and high cost;

[0012] 2) Lack of multi-modal feature fusion and intelligent analysis: the existing technology does not combine temperature, pressure and electrical conductivity for joint intelligent analysis, resulting in insufficient leakage event identification and positioning accuracy; system linkage and response capability is insufficient: most schemes lack intelligent alarm linkage mechanism, and cannot realize automatic identification and rapid response of leakage event, affecting the safety protection of actual application. SUMMARY

[0013] The present application aims to provide a pipeline leakage detection method and system based on multi-modal composite perception and a storage medium, and aims to solve the above problems.

[0014] The present application is mainly realized by the following technical solutions:

[0015] A pipeline leakage detection method based on multi-modal composite perception, comprising the following steps:

[0016] Step S1: Based on the sensor array, the temperature, pressure and conductivity signal increment changes of the pipeline circumferential side are collected to obtain the temperature increment ΔT, pressure increment ΔP and conductivity increment Δσ arranged in an array; then, data preprocessing is performed to form a standardized multi-dimensional feature data set;

[0017] Step S2: Based on the standardized multi-dimensional feature data set, a fusion analysis model is used to identify the leakage feature mode;

[0018] Step S21: Extracting time series signal features; based on the time series data of temperature increment ΔT, pressure increment ΔP and conductivity increment Δσ, a long short-term memory network LSTM is used to extract the time correlation features of multi-dimensional feature data, to capture the dynamic rules and mutation features of temperature, pressure and conductivity changes;

[0019] Step S22: Extracting spatial distribution features; based on the spatial distribution data of temperature increment ΔT, pressure increment ΔP and conductivity increment Δσ, a convolutional neural network CNN is used to extract two-dimensional or three-dimensional features;

[0020] Step S23: Multi-modal feature fusion; the time series signal features in step S21 and the spatial distribution features in step S22 are multi-modal fused through a feature fusion layer;

[0021] Step S24: Finally, based on a multi-level classifier or a multi-layer fully connected neural network, the leakage state, leakage position and leakage severity are output.

[0022] In order to better realize the present application, further, in step S21, first, the time series data of temperature increment ΔT, pressure increment ΔP and conductivity increment Δσ is normalized and time-synchronized; then, the processed time series data is input into a long short-term memory network LSTM, which captures the mutation features generated by the leakage event through a memory cell, while retaining the trend information of long time scale, so as to take into account short-term dramatic changes and long-term changes; finally, a time feature vector containing the dynamic mode and time dependence of each channel at different times is output.

[0023] In order to better realize the present application, further, in the step S22, firstly, the one-dimensional spatial distribution data of the collected temperature increment ΔT, pressure increment ΔP and conductivity increment Δσ are aligned and superimposed, so that one spatial position x contains the detection values of the temperature increment ΔT, pressure increment ΔP and conductivity increment Δσ at the same time, and a data matrix is formed; then, the spatial distribution characteristics of water leakage are obtained based on the convolutional neural network CNN processing. Specifically, the spatial position x corresponds to the serial number position of the point sensor in each sensor array.

[0024] In order to better realize the present application, further, the feature fusion layer is an attention mechanism fusion module, and the weights of each feature channel are dynamically adjusted based on the attention mechanism.

[0025] In order to better realize the present application, further, in the step S24, the spatio-temporal features ffusion fused by the feature fusion layer are input into the MLP multilayer perceptron to obtain the water leakage state, water leakage position and water leakage severity; the MLP multilayer perceptron includes three parallel classification branches:

[0026] The water leakage state branch outputs the probability distribution of no water leakage and water leakage;

[0027] The water leakage position branch outputs the water leakage position coordinates of the sample;

[0028] The water leakage severity branch outputs the probability distribution of micro-leakage, medium and severe.

[0029] In order to better realize the present application, further, it further includes a step S3 of outputting the two-dimensional / three-dimensional coordinates of the water leakage point based on the water leakage position and sensor position information.

[0030] The present application is mainly realized by the following technical solutions:

[0031] A pipeline water leakage detection system based on multi-modal composite sensing, realized based on the above-mentioned pipeline water leakage detection method based on multi-modal composite sensing, comprising a material layer module, a sensor array module and a control processing module, the control processing module comprising a data acquisition and preprocessing module and an intelligent analysis and identification module, the sensor array module being arranged inside the material layer module, and the sensor array module being connected with the data acquisition and preprocessing module; the data acquisition and preprocessing module is used for real-time acquisition of temperature, pressure and conductivity signals in the sensor array module, and pre-processing is performed to form a standardized multi-dimensional feature data set; the intelligent analysis and identification module is used for identifying water leakage feature patterns based on the multi-dimensional feature data set by using a fusion analysis model;

[0032] The material layer module comprises, from bottom to top, a bottom layer, a moisture absorption and heat release expansion layer and an outer protective layer; the sensor array module comprises, from bottom to top, a temperature sensor array, a pressure sensor array and an electric conductivity sensor array; the moisture absorption and heat release expansion layer comprises an adsorbing salt and an expansion material, the adsorbing salt generates an exothermic reaction when meeting water and transmits heat to the temperature sensor array; the expansion material expands when absorbing water to generate mechanical stress and directly acts on the pressure sensor array; the electric conductivity sensor array changes resistance when meeting water and generates an electric conductivity characteristic signal.

[0033] In order to better realize the present application, further, the adsorbing salt is calcium chloride, the expansion material is cross-linked polyacrylic acid sodium, and the mass ratio of calcium chloride and cross-linked polyacrylic acid sodium is 4:1.

[0034] In order to better realize the present application, further, an alarm and linkage module connected with the control processing module is further included, and the alarm and linkage module is used for triggering multi-channel alarm.

[0035] A computer readable storage medium, having stored thereon a computer program, the program being executed by a processor to implement the above-mentioned pipeline leakage detection method based on multi-modal composite sensing.

[0036] The beneficial effects of the present application are as follows:

[0037] (1) The present application forms a standardized multi-dimensional feature data set based on the incremental changes of temperature, pressure and electric conductivity signals, and adopts a long short-term memory network (LSTM) to extract the time dependence of the multi-dimensional feature data, capture the dynamic rules and mutation characteristics of temperature, pressure and electric conductivity changes; a convolutional neural network (CNN) is used to perform convolution operation on the spatial distribution data of the sensor array, extract the spatial patterns of leakage heat, pressure distribution and electric conductivity changes, and assist in locating the leakage point and area. Then, the time features and spatial distribution features are multi-modal fused through a feature fusion layer, and input into a classifier to complete the state recognition, position estimation and severity classification of the leakage event. The high-dimensional features after fusion of the present application comprehensively consider the multi-physical field information of the leakage, significantly improve the recognition ability of small leakage and complex environmental changes; at the same time, the present application has high robustness, and can effectively suppress the interference of abnormal signals and environmental noise on the judgment result.

[0038] (2) The multifunctional composite material layer designed by innovation realizes multi-physical field sensitive capture of water leakage events by integrating salt adsorption heat release reaction, pressure sensing of expanded material and electrically conductive fine iron wire electric conduction change, and significantly improves the sensitivity and accuracy of water leakage detection. Compared with the traditional single sensor method, the present application effectively overcomes the technical problems of difficult detection of small water leakage, accurate positioning of water leakage position and high false alarm rate caused by environmental interference. Secondly, based on the fusion analysis model, the present application uses machine learning and deep learning algorithm to jointly analyze the multi-modal time series data, realizes intelligent discrimination of water leakage state, position and severity, effectively distinguishes water leakage signal from environmental interference, and improves the robustness and self-adaptive ability of the system.

[0039] (3) The present application further constructs a real-time alarm mechanism, which can ensure quick response when water leakage occurs, and realizes accurate alarm. The present application supports linkage with pipe network control device, such as automatic closing of valve and notification of maintenance personnel, and improves the efficiency of water leakage disposal. The present application has the advantages of flexible layout, fast response speed, high detection accuracy and low maintenance cost, and meets the large-scale application requirements of multiple scenes such as family, business and industry. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The present application is a schematic diagram of the principle of the pipe water leakage detection method based on multi-modal composite sensing;

[0041] Figure 2 The present application is a simulated temperature change curve ΔT-t, pressure change curve ΔP-t and conductivity change curve Δσ-t in example 1;

[0042] Figure 3 The present application is a schematic diagram of the principle of the pipe water leakage detection system based on multi-modal composite sensing;

[0043] Figure 4 The present application is a schematic diagram of the principle of the linkage alarm;

[0044] Figure 5 The present application is a structural schematic diagram of the moisture absorption and heat release expansion layer and the sensor array module.

[0045] Wherein: 1-bottom layer, 2-moisture absorption and heat release expansion layer, 3-sensor array module, 4-outer protective layer. DETAILED DESCRIPTION

[0046] Example 1:

[0047] A pipe water leakage detection method based on multi-modal composite sensing, based on a fusion analysis model to identify water leakage feature patterns, such as Figure 1As shown, the fusion analysis model in the application includes a long short-term memory network LSTM and a convolutional neural network CNN arranged side by side, and a multi-level classifier. The application performs intelligent analysis based on multi-dimensional sensor data, adopts advanced machine learning and deep learning technologies, and realizes accurate identification, positioning and severity judgment of the water leakage event. Specifically, the following steps are included:

[0048] (1) Multi-dimensional data acquisition and preprocessing;

[0049] The application forms a sensor array composed of multiple types of sensing units integrated in the composite material layer, and collects multi-dimensional physical signals such as temperature, pressure and conductivity in real time. The application fuses temperature, pressure, conductivity and other multi-physical field signals, and through denoising, normalization and synchronization preprocessing of the data, forms a unified format of multi-dimensional spatio-temporal data set, realizes sensitive perception and accurate identification of the water leakage event. The specific collected features include:

[0050] 1) Temperature change feature;

[0051] Relying on the exothermic reaction of adsorbed salt and water leakage, the temperature sensor captures the instantaneous rise of temperature, and reflects the occurrence and intensity of water leakage by calculating the incremental change (ΔT) of temperature.

[0052] 2) Pressure change feature;

[0053] The local pressure change of the water-absorbing expansion material to the pipeline is measured by the pressure sensor, forming a pressure increment (ΔP), which provides a basis for judging the severity of water leakage.

[0054] 3) Conductivity change feature;

[0055] The conductive fine iron wire embedded in the material layer shows a high resistance state when dry, and the resistance drops sharply when water is encountered. The system realizes rapid detection of water leakage by monitoring the conductivity change (Δσ) in real time.

[0056] The features of the above three dimensions are obtained by calculating the increment, i.e. the total change relative to the baseline value before water leakage. The multi-channel sensor data collected by the system in a certain period of time (such as the past 1 minute) before water leakage can be calculated by sliding average or median filtering to obtain the "normal baseline" under the current environment. For example: temperature sensor: the average value of the past 60 seconds is 25.0℃, so take 25℃ as the baseline, and if the water-absorbing exothermic temperature is 30.0℃, then ΔT is 5. Therefore, after real-time data collection, the algorithm module will perform a simple operation.

[0057] The calculation of pressure increment ΔP and conductivity increment Δσ is the same. In this way, the change relative to the current environmental state is obtained, rather than a rigid absolute value. Moreover, it can adapt to seasonal, day and night, air conditioning opening and other environmental changes. Under normal conditions, the data change during water leakage is as follows:

[0058] The temperature increment ΔT can generally reach 8-20℃ or even higher, and the change is particularly obvious;

[0059] The pressure increment ΔP is obviously changed due to the water absorption and expansion of the pipeline, and is generally 5-50 kPa;

[0060] The conductivity increment Δσ can jump from very low to high conductivity state, showing a step change.

[0061] All original signals are filtered, denoised, drift corrected, normalized and time synchronized by the acquisition and preprocessing module to form a standardized multi-dimensional feature data set.

[0062] (2) Then, the data is input into the fusion analysis model;

[0063] For time series signals, the temperature change curve ΔT-t, the pressure change curve ΔP-t and the conductivity change curve Δσ-t, a long short-term memory network LSTM is used to extract time-related features; specifically, the time-dependent characteristics of sensor data are extracted by using a time convolution network TCN and a long short-term memory network LSTM, and the dynamic rules and mutation characteristics of temperature, pressure and conductivity changes are captured.

[0064] For spatial distribution data, the temperature spatial change curve T-x, the pressure spatial change curve P-x and the conductivity spatial change curve σ-x curve, a convolutional neural network CNN is used for two-dimensional or three-dimensional feature extraction; specifically, the spatial distribution data of the sensor array is convolved by using a convolutional neural network CNN, and the spatial patterns of water leakage heat, pressure distribution and conductivity change are extracted to assist in locating the water leakage point and area.

[0065] Finally, the time features and spatial distribution features are fused by a feature fusion layer, and input into a classifier to complete the state recognition (whether water leakage), position estimation (water leakage point coordinates or area) and severity classification (micro leakage / medium / severe) of the water leakage event. Specifically, the feature fusion layer jointly learns the time series features and spatial distribution features, dynamically adjusts the weights of each feature channel through an attention mechanism, and improves the sensitivity and recognition accuracy of the model to key signals. The high-dimensional features after fusion integrate the multi-physical field information of water leakage, significantly improving the recognition ability of small leakage and complex environmental changes; at the same time, this layer has a certain robustness, which can effectively suppress the interference of abnormal signals and environmental noise on the judgment result.

[0066] Based on deep feature representation, the system uses a multi-layer fully connected neural network or a lightweight classifier (e.g., XGBoost, Random Forest) to perform binary classification (leaking / normal) of leakage status and multi-level severity classification (minor leak, moderate leak, severe leak). The classifier is trained on a large number of labeled leakage datasets, supporting continuous iterative optimization of the model to improve generalization ability. It can also be combined with a threshold strategy to achieve a balance between sensitivity and false alarm rate, ensuring stable system operation.

[0067] Specifically, such as Figure 1 As shown, the fusion analysis model includes the following steps:

[0068] Step A1: Time series signal feature extraction;

[0069] For three types of time-series signals—temperature change curve ΔT-t, pressure change curve ΔP-t, and conductivity change curve Δσ-t—normalization and time synchronization processing are first performed (i.e., the time is adjusted to be consistent so that the three detection values ​​are at the same time). Then, the signals are input into a Long Short-Term Memory (LSTM) network to extract time-related features.

[0070] LSTM captures the abrupt changes caused by leakage events through its memory cells (e.g., instantaneous peak of ΔT, step increase of ΔP, sudden jump of Δσ); at the same time, it retains trend information over long time scales (e.g., slow signal decline, sustained high voltage range), thus taking into account both short-term drastic changes and long-term variations; the final output time feature vector contains the dynamic patterns and time dependencies of each channel at different times.

[0071] Step A2: Spatial distribution feature extraction;

[0072] First, there are three curves representing Tx, Px, and σ-x that vary with location, representing one-dimensional spatial distribution data. Assume there are n spatial locations (x1, x2, ..., x...). n Let Tx be the temperature increment ΔT, Px be the pressure increment ΔP, and σ be the conductivity increment Δσ. Then Tx, Px, and σ-x are all vectors of length n. Superimposing these three vectors results in:

[0073] Temperature increment per layer: ΔT1 ΔT2 ΔT3 ... ΔTn

[0074] Second-layer pressure increment: ΔP1 ΔP2 ΔP3 ... ΔPn

[0075] The conductivity increments of the three layers are: Δσ1 Δσ2 Δσ3 ... Δσn.

[0076] The above superposition forms a 3-row, n-column "matrix", where 3 is the number of channels and n is the number of spatial positions. At this time, when the CNN convolution kernel scans, it is equivalent to looking at three physical quantities at the same position at a time, and also looking at how they change at adjacent positions to obtain the spatial distribution feature pattern of the water leakage. For example, the water leakage position feature or the water leakage range feature.

[0077] Step A3: Multimodal feature fusion

[0078] The time feature vector output by the LSTM and the spatial distribution feature vector output by the CNN are spliced and weighted in the feature fusion layer (for example, a fully connected layer or an attention mechanism fusion module). The attention mechanism fusion module dynamically allocates weights to different feature channels based on the attention mechanism. For example, in the initial water leakage stage, the time feature weight is higher, and in the diffusion stage, the spatial distribution feature weight is higher.

[0079] For example: Assuming that the time feature vector ft output by the LSTM is [0.2, 0.5, 0.1] and the spatial distribution feature vector fs output by the CNN is [0.3, 0.7, 0.4], after fusion in the feature fusion layer, the spatiotemporal feature ffusion is [0.2, 0.5, 0.1, 0.3, 0.7, 0.4]. The fused vector contains both time evolution information and spatial distribution information and can be directly input to the classifier for water leakage recognition, position estimation, and severity judgment.

[0080] A4: Classification task

[0081] The spatiotemporal feature ffusion obtained after fusion is input into an MLP multilayer perceptron, and three tensors are output, representing three classification tasks. Specifically, the MLP multilayer perceptron includes:

[0082] The water leakage state branch (status_pred) outputs two values, representing the probability distribution of "no water leakage / water leakage", and the classification result is the event with the highest probability;

[0083] The water leakage position branch (position) does not pass through an activation function (such as softmax) to convert to probability, but outputs two sample coordinates, so it also outputs two values, representing the water leakage position coordinates [x1, x2] of each sample.

[0084] The water leakage severity branch (severity_pred) outputs three values, representing the probability distribution of "micro leakage / medium / severe", and the classification result is the degree level with the highest probability.

[0085] For example Figure 1As shown, the loss functions of the water leakage state branch and the water leakage severity branch are cross-entropy loss functions, and the loss function of the water leakage position branch is a mean square error loss function, and the loss function of the MLP multi-layer perceptron is the sum of the loss functions of the three branches.

[0086] The MLP multi-layer perceptron is trained based on simulation data, the model is optimized by jointly calculating the loss of each task and performing back propagation, and the opening and closing of the softmax activation function are controlled, and it is proved that the MLP multi-layer perceptron model can simultaneously predict three tasks. The MLP multi-layer perceptron is tested based on simulation data, as shown in Table 1, in the water leakage state column, 1 represents water leakage, and 0 represents no water leakage; it can be analyzed that, as shown in Table 2, from the performance report of the MLP multi-layer perceptron model, it can be seen that the water leakage state detection accuracy is 92%, the water leakage severity detection accuracy is 83%, and the predicted information is available. This simulation proves that the MLP multi-layer perceptron model can indeed complete the classification of "water leakage state, water leakage position and water leakage severity" through a large amount of different feature data. Subsequently, a large amount of data can be used for iterative learning to make the model more accurate.

[0087] Table 1: Comparison of model prediction information after initial simulation training

[0088]

[0089] Table 2: Performance report of the MLP multi-layer perceptron model

[0090]

[0091] (3) Water leakage position estimation;

[0092] Combined with the output water leakage position coordinates and sensor position information, a regression model or positioning network can be used to output the two-dimensional / three-dimensional coordinates of the water leakage point. Through model inversion technology, the spatial distribution of the sensor signal is mapped back to the actual pipeline position, realizing accurate positioning. The regression model, positioning network and model inversion technology at this place are all prior art, and will not be described again.

[0093] The present application can automatically update model parameters based on continuously collected operation data, adapt to different pipe network environments and sensor characteristic changes. Through transfer learning and semi-supervised learning techniques, a small amount of labeled data can be effectively used to expand the application range of the model; intelligent maintenance and performance self-optimization are realized, the system life is prolonged and the maintenance cost is reduced.

[0094] As shown in Table 3, in order to verify the effect of multi-dimensional feature fusion, we simulated data in multiple scenarios, and the simulated scenarios are:

[0095] T1 micro-leakage (1-3 drops of water per minute, position close to sensor);

[0096] T2 Moderate water leakage (fine water flow, flow rate 0.1-0.3 L / min, location close to sensor);

[0097] T3 Severe water leakage (continuous water flow, >1 L / min);

[0098] T4 Remote micro-leakage (leakage point far from sensor, weak signal);

[0099] T5 Normal state (no conductance, pressure change, to be excluded);

[0100] T6 Temperature fluctuation false signal (no water leakage, but significant external temperature change).

[0101] As shown in Figure 2 , the present application simulates the temperature change curve ΔT-t, the pressure change curve ΔP-t, and the conductivity change curve Δσ-t, respectively. As shown in Table 4, the analysis results based on the simulation data of the present application are accurate and have high confidence.

[0102] Table 3 Simulated scenario information

[0103] Scenario ΔΤ (°C) Δσ (mS / cm) ΔΡ (kPa) Explanation T1 6~8 0→50 Mutation 5~10 Obvious exotherm, conductance jump, slight pressure increase T2 10~12 0→150 Mutation 15~30 Obvious exotherm, large conductance increase, significant pressure increase T3 12~18 0→300 Near short circuit 30~50 Very strong exotherm, conductance near short circuit, large pressure increase T4 4~6 0→20 Small change 2~5 Signal attenuated, still captured by algorithm T5 2~3 0 0 No conductance, pressure change, will be excluded T6 12+6 300+100 40+20 Superimposed signal, AI can distinguish

[0104] Table 4 Analysis information based on simulation data

[0105]

[0106] Example 2

[0107] A pipeline water leakage detection system based on multi-modal composite sensing, as shown in Figure 3 , includes a material layer module, a sensor array module 3, and a control processing module, the control processing module includes a data acquisition and preprocessing module and an intelligent analysis and identification module, the sensor array module 3 is arranged inside the material layer module, and the sensor array module 3 is connected with the data acquisition and preprocessing module. The data acquisition and preprocessing module is used for real-time acquisition of temperature, pressure and conductance signals in the sensor array module 3, and pre-processing to form a standardized multi-dimensional feature data set; the intelligent analysis and identification module is used for identifying water leakage feature patterns based on the multi-dimensional feature data set using a fusion analysis model.

[0108] As shown in Figure 5 , the material layer module includes a bottom layer 1, a moisture absorption and heat release expansion layer 2, and an outer protective layer 4 arranged in order from bottom to top; the material layer module is directly attached to the outer wall of the pipeline to be monitored, and is responsible for sensing the physical changes (including temperature change, pressure change and conductance change) caused by water leakage.

[0109] The sensor array module 3 comprises a temperature sensor array, a pressure sensor array and an electric conductivity sensor array arranged in sequence from bottom to top inside the material layer module; the sensor array module 3 can realize high-density collection of multi-point and multi-dimensional data, and improve the spatial resolution and sensitivity of water leakage detection.

[0110] The hygroscopic and exothermic expansion layer 2 comprises adsorbing salt and expansion material; the adsorbing salt generates exothermic reaction when meeting water and transfers heat to the temperature sensor array; the expansion material expands when meeting water to generate mechanical stress and directly acts on the pressure sensor array; the electric conductivity sensor array comprises electrically conductive thin iron wire (for example, nickel-plated copper-clad iron wire), which changes resistance and generates electric conductivity characteristic signal when meeting water. Preferably, the electrically conductive thin iron wire can be embedded inside the hygroscopic and exothermic expansion layer 2.

[0111] Specifically, as shown in the drawings, Figure 5 The present application arranges a multifunctional composite structure composed of the bottom layer 1+hygroscopic and exothermic expansion layer 2 mixed with adsorbing salt and expansion material+embedded temperature sensor array+embedded pressure sensor array+electric conductivity sensor array+outer protective layer 4 at the key monitoring position of the pipeline. When water leakage occurs, the adsorbing salt generates exothermic reaction when meeting water, and the temperature change is transferred to the temperature sensor through the composite material to realize rapid capture of the temperature signal; the expansion material expands when meeting water to generate mechanical stress and directly acts on the pressure sensor to obtain the pressure change characteristics; the electrically conductive thin iron wire significantly changes resistance (short circuit or conduction) when meeting water to generate electric conductivity characteristic signal. The integrated material layer simultaneously provides three independent sensing channels of temperature, pressure and electric conductivity, constitutes a multi-physical characteristic fusion detection mechanism, and enables the system to obtain high enough signal-to-noise ratio even under the condition of small leakage, thereby improving the accuracy and robustness of water leakage identification.

[0112] Preferably, the bottom layer 1 is non-woven fabric, and the outer protective layer 4 is polyurethane (PU) film. The adsorbing salt and the expansion material are calcium chloride (CaCl2) and cross-linked sodium polyacrylate (SAP) respectively, and are mixed together according to a mass ratio of about 4:1.

[0113] The role of CaCl2 is to release heat immediately after absorbing water, so that the temperature sensor can quickly capture the change; the role of SAP is to rapidly increase in volume after absorbing water, and directly provide a measurable “pushing force” to the pressure sensor. The mass ratio of 4:1 is based on the reaction speed and signal contribution ratio of the two materials: the exothermic reaction of calcium chloride is strong but the duration is not as long as the expansion material, and the heat is released immediately, so it is suitable for providing instantaneous temperature peak value; the water absorption process of the expansion material is relatively slow, but it can be maintained for a long time, and it may be swollen for a long time after swelling. This ratio can not only ensure the rapid generation of temperature signal at the initial stage of water leakage, but also maintain the persistence of pressure signal at the later stage of reaction, so as to realize the complementation and enhancement of multi-channel signals in the whole water leakage process.

[0114] Preferably, the data acquisition and preprocessing module is responsible for real-time receiving, digitizing processing and preliminary sorting of temperature, pressure and conductivity signals collected by the sensor array; at the same time, it performs data filtering, denoising, calibration and feature extraction to ensure that the data input into the intelligent analysis and recognition module have high precision and high reliability.

[0115] The intelligent analysis and recognition module: unifies modeling and correlation analysis of multi-dimensional signals such as temperature, pressure and conductivity, and identifies the leakage feature mode by using fusion analysis model. It can simultaneously analyze time series features (temperature / pressure change rate, duration, etc.) and spatial distribution features (sensor array response position, etc.), thereby realizing accurate determination of the leakage event. The machine learning model is trained through historical data accumulation and simulated data generation, supports periodic update or online learning to adapt to environmental changes. It supports user feedback or manual correction mechanism to continuously optimize the accuracy of leakage identification and positioning.

[0116] Specifically, the control processing module (including main control chip and integrated circuit board) is designed in a compact box structure, which is convenient to install in the access hole, equipment cabinet or shallow slot of wall, has excellent environmental adaptability and heat dissipation performance. For the decorated environment, it is recommended to reserve a small maintenance access in the hidden parts of the wall corner, skirting line, ceiling and pipe well, which is used for wiring, replacement of adsorbed salt material and sensor maintenance, to ensure the long-term reliable operation and convenient maintenance of the system.

[0117] Preferably, as shown in Figure 3 , it also includes an alarm and linkage module connected with the control processing module, which is used to trigger multi-channel alarm. The alarm and linkage module includes the existing smart home central control, local sound and light alarm and maintenance module. As shown in Figure 4 , if the leakage anomaly is detected, the control processing module can issue a water cutoff control instruction to close the electromagnetic valve, start the sound and light alarm, remotely push the alarm information to the user end, and send the maintenance information to the maintenance personnel, so that the maintenance personnel can know the leakage event in the shortest time. Specifically, the control processing module supports local and remote communication functions, and can automatically link with the pipe network control system, emergency valve, fire fighting system and building management system to realize emergency valve closing, starting of drainage equipment or triggering of other emergency measures, thereby effectively reducing the property loss and safety risk caused by leakage.

[0118] Preferably, a hidden access hole can be reserved in inconspicuous positions such as wall decorative plate, cabinet or skirting line, which is mainly used for facilitating on-site personnel to quickly repair the leakage point when the pipeline leaks.

[0119] The core principle of the present application is to realize sensitive capture of water leakage events through a specially designed material layer structure, and to significantly improve the accuracy and robustness of water leakage identification through machine learning / deep learning analysis of multi-dimensional sensing features; at the same time, with intelligent alarm and linkage mechanism, fast and accurate response control is realized.

[0120] The system adopts modular and flexible design, which is convenient for rapid deployment in various practical scenarios, and the application scope covers synchronous pre-burying of newly built pipe networks and later upgrading of already decorated buildings. The core monitoring component of the system is a multifunctional composite flexible strip material integrating moisture-absorbing and heat-releasing expansion layer 2, temperature sensor array, pressure sensor array and electric conductivity sensor array, with a thickness of about 0.5 cm. The flexible strip material has the characteristics of lightness, bendability, high adhesion, etc., and can be closely attached to the outer wall of hot and cold water pipes to realize synchronous sensing of multi-dimensional signals. The installation method is various, which can be pasted with adhesive, or supported by binding, buckling or nesting type fixing, ensuring no damage to the original pipe structure, adapting to various pipe diameters and bending angles, and ensuring good contact with the pipe surface, thereby improving the sensing sensitivity and response speed.

[0121] The present application preferentially arranges the flexible strip material at key nodes (such as pipe corners, interface and end) in high water leakage areas (such as bathrooms, kitchens, etc.) to form a three-dimensional distributed multi-point sensing network. Each segment of sensor is connected in series through a bus to access a unified master control module, realizing centralized data collection and unified intelligent management, and improving the stability and expandability of the whole system. Based on the non-destructive and flexible composite structure, combined with the multi-dimensional sensor array, the present application realizes high sensitivity and accurate water leakage monitoring and intelligent early warning. The distributed and modular design takes into account the installation convenience, environmental adaptability and aesthetics, which not only provides an ideal solution for the intelligent early warning system of newly built buildings, but also provides a feasible technical path for the intelligent transformation of existing buildings.

[0122] The present application can be applied in the following fields:

[0123] (1) Centralized residential buildings such as hotels and apartments: the present application system is installed around the central water supply pipe in the building and the sanitary facilities of each household, which can collect multi-dimensional sensing data such as temperature, pressure and electric conductivity in real time, and upload the data to the property management system. The present application realizes accurate positioning and classification of water leakage events through integrated intelligent recognition algorithm, and timely triggers multi-level linkage alarm. The system can be linked with the fire and emergency response system to quickly execute automatic valve closing and personnel notification, effectively preventing structure penetration and wall mold caused by chronic water leakage, reducing maintenance cost, and realizing large-area distributed water leakage monitoring and centralized management.

[0124] (2) Industrial plant and production workshop: suitable for the key nodes of water supply and cooling pipe network in the factory, equipment-intensive area and underground concealed pipeline leakage detection. The application can simultaneously sense temperature rise, pressure change and electric conduction short circuit, real-time identify leakage event and trigger alarm, support automatic linkage to close related valves. The system can be integrated into the industrial automation platform, provide remote monitoring and maintenance, improve equipment safety and production continuity, reduce equipment damage and safety hazards caused by leakage.

[0125] (3) Public buildings such as hospitals and schools: the system is laid out in areas with dense water use and large personnel flow such as dormitory area and bathroom, supporting accurate identification and rapid alarm of local leakage. The application can reduce false alarm rate through multi-sensor fusion algorithm, ensure stable operation of the system. The application is connected to the building management system, realizes regional leakage risk grading management, ensures timely discovery and handling of leakage hazards in long-term unattended condition, protects public safety and building structure safety.

[0126] The above is only the preferred embodiment of the application, not any form of limitation on the application, any simple modification or equivalent change of the above embodiment according to the technical essence of the application falls within the protection scope of the application.

Claims

1. A pipeline leakage detection method based on multimodal composite sensing, characterized in that, Includes the following steps: Step S1: Based on the sensor array, collect the incremental changes of temperature, pressure and conductivity signals around the pipe to obtain the temperature increment ΔT, pressure increment ΔP and conductivity increment Δσ arranged in an array; Then, data preprocessing is performed to form a standardized multidimensional feature dataset; Step S2: Based on a standardized multidimensional feature dataset, a fusion analysis model is used to identify leakage feature patterns; Step S21: Extract time series signal features; Based on the time series data of temperature increment ΔT, pressure increment ΔP and conductivity increment Δσ, the Long Short-Term Memory (LSTM) network is used to extract the time correlation features of the multidimensional feature data, and capture the dynamic laws and abrupt changes of temperature, pressure and conductivity. Step S22: Extract spatial distribution features; Based on the spatial distribution data of temperature increment ΔT, pressure increment ΔP, and conductivity increment Δσ, a convolutional neural network (CNN) is used to extract two-dimensional or three-dimensional features; Step S23: Multimodal feature fusion; The temporal signal features from step S21 and the spatial distribution features from step S22 are fused in a multimodal manner through a feature fusion layer; Step S24: Finally, based on a multi-level classifier or a multi-layer fully connected neural network, output the leakage status, leakage location, and leakage severity.

2. The pipeline leakage detection method based on multimodal composite sensing according to claim 1, characterized in that, In step S21, firstly, the time-series data of temperature increment ΔT, pressure increment ΔP, and conductivity increment Δσ are normalized and time-synchronized; then, the processed time-series data are input into a Long Short-Term Memory (LSTM) network, which captures the abrupt changes caused by the leakage event through memory units while retaining trend information over a long time scale; finally, a time feature vector containing the dynamic patterns and time dependencies of each channel at different times is output.

3. The pipeline leakage detection method based on multimodal composite sensing according to claim 1, characterized in that, In step S22, firstly, the one-dimensional spatial distribution data of the collected temperature increment ΔT, pressure increment ΔP, and conductivity increment Δσ are aligned and superimposed so that a spatial location x simultaneously contains the detected values ​​of temperature increment ΔT, pressure increment ΔP, and conductivity increment Δσ, forming a data matrix; then, the spatial distribution characteristics of the leakage are obtained based on the processing of a convolutional neural network CNN.

4. The pipeline leakage detection method based on multimodal composite sensing according to claim 1, characterized in that, In step S23, the feature fusion layer is an attention mechanism fusion module, and the weights of each feature channel are dynamically adjusted based on the attention mechanism.

5. The pipeline leakage detection method based on multimodal composite sensing according to claim 4, characterized in that, In step S24, the spatiotemporal features fused by the feature fusion layer are input into the MLP (Multilayer Perceptron) to obtain the leakage status, leakage location, and leakage severity. The MLP includes three parallel classification branches: The leakage status branch outputs the probability distributions for no leakage and leakage. The leak location branch outputs the coordinates of the leak location of the sample. The leakage severity branch outputs the probability distribution of minor, moderate, and severe leaks.

6. The pipeline leakage detection method based on multimodal composite sensing according to claim 1, characterized in that, It also includes step S3: based on the location of the leak and the sensor location information, output the two-dimensional / three-dimensional coordinates of the leak point.

7. A pipeline leakage detection system based on multimodal composite sensing, implemented based on the pipeline leakage detection method based on multimodal composite sensing as described in any one of claims 1-6, characterized in that, It includes a material layer module, a sensor array module, and a control processing module. The control processing module includes a data acquisition and preprocessing module and an intelligent analysis and recognition module. The sensor array module is located inside the material layer module and is connected to the data acquisition and preprocessing module. The data acquisition and preprocessing module is used to acquire temperature, pressure, and conductivity signals from the sensor array module in real time and perform preprocessing to form a standardized multidimensional feature dataset. The intelligent analysis and recognition module is used to identify leakage feature patterns based on a multi-dimensional feature dataset and a fusion analysis model. The material layer module includes, from bottom to top, a bottom layer, a moisture-absorbing and heat-releasing expansion layer, and an outer protective layer; the sensor array module includes, from bottom to top, a temperature sensor array, a pressure sensor array, and a conductivity sensor array; the moisture-absorbing and heat-releasing expansion layer includes adsorbed salt and an expansion material, wherein the adsorbed salt reacts with water to produce an exothermic reaction and transfers heat to the temperature sensor array; the expansion material absorbs water and expands to generate mechanical stress, which directly acts on the pressure sensor array; the conductivity sensor array changes its resistance and generates a conductivity characteristic signal after encountering water.

8. A pipeline leakage detection system based on multimodal composite sensing according to claim 7, characterized in that, The adsorbed salt is calcium chloride, the expanding material is cross-linked sodium polyacrylate, and the mass ratio of calcium chloride to cross-linked sodium polyacrylate is 4:

1.

9. A pipeline leakage detection system based on multimodal composite sensing according to claim 7, characterized in that, It also includes an alarm and linkage module connected to the control processing module, which is used to trigger multi-channel alarms.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements a pipeline leakage detection method based on multimodal composite sensing as described in any one of claims 1-6.

Images