An intelligent water data acquisition control method, system and terminal device

Abstract

The application provides a kind of intelligent water consumption data acquisition control method, system and terminal device, comprising: the application discloses a kind of intelligent water consumption data acquisition control method, system and terminal device, suitable for urban water supply, industrial buried pipeline micro leakage scene;Through the arrangement distributed optical fiber acoustic sensor, temperature sensor, nanometer film vibration sensor array and high-precision pressure sensor collect multi-physical field data;Through multi-scale wavelet transform, dynamic time warping realizes multi-modal data space-time alignment and feature fusion, eliminates temperature interference using long short-term memory-causal network, combines with the space-time characteristics of Transformer network processing;Based on the particle swarm optimization algorithm and physical information generated adversarial network of Navier-Stokes equation constraint, realize the high-precision inversion and positioning of leakage point coordinate, the scheme integrates multi-physical field data and intelligent algorithm, improves the positioning accuracy and reliability of leakage point, and is suitable for different material pipes.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of smart water, and in particular to an intelligent water use data acquisition and control method, system and terminal device. The present application adopts a buried pipeline micro leakage detection and positioning method, system and terminal device based on multi-physical field fusion perception and intelligent algorithm driving, and is suitable for urban water supply and industrial buried pipeline micro leakage scenarios. BACKGROUND

[0002] In the field of smart water, buried pipelines are the infrastructure for urban water supply and industrial fluid transportation, and their safe operation is directly related to resource efficiency and public safety.

[0003] However, micro leakage caused by pipeline aging, corrosion or external damage is difficult to detect in a timely manner, especially in non-metallic pipelines such as plastic, composite materials or deep buried scenarios, and leakage positioning faces significant technical challenges.

[0004] To this end, the present application provides an intelligent water use data acquisition and control method, system and terminal device. SUMMARY

[0005] In order to make up for the deficiencies of the prior art and solve at least one technical problem raised in the background art.

[0006] The technical scheme adopted by the present application to solve its technical problems is:

[0007] In a first aspect, the present application provides an intelligent water use data acquisition and control method, comprising:

[0008] S1: arranging sensors to collect multi-physical field data, the multi-physical field data including: leakage point sound wave first wave arrival time, pipeline temperature gradient along the line, vibration signal, pressure data;

[0009] S2: based on the collected multi-physical field data, performing time-frequency decomposition on the leakage point sound wave first wave arrival time sequence and the vibration signal through multi-scale wavelet transform, detecting mutation points to optimize feature reliability, and aligning the time axes of different sensors using dynamic time warping and cross-spectral entropy, and extracting effective feature segments;

[0010] S3: using the long short-term memory-causal network to analyze the correlation between the temperature gradient data and the vibration energy after aligning the time axes and extracting the effective features from part of the multi-physical field data, constructing a physical relationship model, calculating a temperature influence factor, compensating the vibration energy using the temperature influence factor, and eliminating noise introduced by environmental temperature fluctuations;

[0011] S4: combining the vibration energy after eliminating temperature interference with the spatiotemporal domain joint feature matrix data to realize leakage point positioning through an optimization algorithm constrained by a physical equation and a generative adversarial network.

[0012] In a second aspect, the present application provides an intelligent water use data acquisition and control system, comprising:

[0013] A multi-physical field data acquisition module: sensors are arranged to acquire multi-physical field data, including: leak point sound wave first wave arrival time, pipeline temperature gradient along the line, vibration signal, pressure data;

[0014] A multi-modal data space-time alignment module: based on the collected multi-physical field data, the time-frequency decomposition of the first wave arrival time sequence and the vibration signal containing the leak point sound wave is carried out through multi-scale wavelet transform, the mutation point is detected to optimize the feature reliability, and the time axis of different sensors is aligned by using dynamic time warping and cross-spectral entropy, and an effective feature segment is extracted;

[0015] A temperature interference elimination and feature fusion module: part of the multi-physical field data after the time axis alignment and the effective feature optimization and extraction is compensated by using the temperature influence factor to eliminate the noise introduced by the environmental temperature fluctuation.

[0016] A physical constraint intelligent positioning module: the vibration energy after the temperature interference elimination is combined with the space-time domain joint feature matrix data to realize the leak point positioning through the optimization algorithm of the physical equation constraint and the generative adversarial network.

[0017] In a third aspect, the present application provides an intelligent water use data acquisition and control terminal device, comprising:

[0018] A distributed optical fiber acoustic sensor, a distributed optical fiber temperature measurement sensor, a nano thin film vibration sensor array, a high-precision pressure sensor, and a computer processing unit.

[0019] The present application has the following advantages:

[0020] 1. Deploying a distributed optical fiber acoustic sensor, a temperature measurement sensor, a nano thin film vibration sensor array, and a high-precision pressure sensor, synchronously acquiring multi-dimensional data such as sound waves, temperature, vibration, and pressure, covering the key physical characteristics of leak detection, and solving the problem of detection difference between metal and non-metal pipes by enhancing signal coupling through a sonar patch, realizing adaptability to pipes of different materials, using multi-scale wavelet transform to perform time-frequency decomposition on non-stationary signals, detecting signal mutation points through layering, avoiding single frequency band noise interference, improving feature reliability, using dynamic time warping (DTW) and cross-spectral entropy (CSE) to solve the time sequence offset problem caused by the difference in sampling frequency of different sensors, and ensuring the accurate alignment of multi-modal data on the time axis and the frequency domain, providing synchronous and reliable data basis for subsequent analysis.

[0021] 2. The physical relationship model between temperature gradient and vibration energy attenuation is constructed by long short-term memory-causal network (LSTM-CN), dynamic learning of temperature influence factor, quantitative compensation of the interference of temperature fluctuation on vibration signal, ensuring that the detection result is not affected by environmental temperature change; combined with Transformer network and attention mechanism, the spatial and temporal domain joint feature extraction is carried out on the multi-modal data: time flow Transformer processes the time sequence of sound wave arrival time, and identifies the multi-path reflection interference; the spatial flow Transformer fuses the GIS pipe network topology, weights the high leakage risk nodes, guides the algorithm to preferentially search the high-risk area, and improves the detection efficiency; the particle swarm optimization (PSO) algorithm based on Navier-Stokes equation constraint introduces the pipe material parameters and fluid mechanics equation to dynamically adjust the particle search step, and ensures that the search path conforms to the physical law; the physical information generative adversarial network (PI-GAN) filters the candidate solution that violates the wave equation in real time, and excludes the physically infeasible solution; the collaborative positioning algorithm realizes high-precision inversion of the leakage point coordinates, and the positioning error is less than or equal to 0.5 meters, and the physical consistency is verified by COMSOL simulation, ensuring the reliability of the result. BRIEF DESCRIPTION OF DRAWINGS

[0022] The application will be further described below in combination with the drawings.

[0023] Figure 1 is a step flow chart of embodiment 1 of the application;

[0024] Figure 2 is a system module diagram of embodiment 2 of the application;

[0025] Figure 3 is a device connection diagram of embodiment 3 of the application. DETAILED DESCRIPTION

[0026] In order to make the technical means, creative features, purposes and effects realized by the application easy to understand, the application will be further described below in combination with specific embodiments.

[0027] Embodiment 1

[0028] As shown in Figure 1 , the intelligent water use data acquisition and control method described in the embodiment of the application comprises:

[0029] S1: arranging a sensor to collect multi-physical field data, the multi-physical field data comprising: leakage point sound wave first wave arrival time, pipe temperature gradient along the line, vibration signal, and pressure data;

[0030] The distributed optical fiber acoustic sensor DAS, the distributed optical fiber temperature measurement sensor DTS, the nanometer thin film vibration sensor array, and the high-precision pressure sensor are arranged to collect multi-physical field data;

[0031] In some embodiments, the distributed optical fiber acoustic sensor DAS is deployed along the pipeline at a preset distance to extract the leak point acoustic first wave arrival time, and the distributed optical fiber temperature sensor DTS is integrated to monitor the temperature gradient along the pipeline. For the low acoustic impedance characteristics of plastic pipes, the transmission efficiency of the leak sound to the sensor is improved by the sonar patch, and the leak sound signal coupling is enhanced. It should be noted that for metal pipes, the distributed optical fiber acoustic sensor DAS, the distributed optical fiber temperature sensor DTS, and the nano thin film vibration sensor array can be directly deployed;

[0032] In some embodiments, a nano thin film vibration sensor array is deployed on the outer wall of the pipeline to monitor the vibration signal generated by the leak point, and a high-precision pressure sensor is installed along the pipeline at a preset distance to collect pressure data in real time, and a computer processing unit is used to calculate the real-time pressure derivative Unit Pa / s;

[0033] The leak point acoustic first wave arrival time, the temperature gradient along the pipeline, the vibration signal, the pressure data, and the multi-physical field data are combined.

[0034] S2: Based on the collected multi-physical field data, the time-frequency decomposition of the leak point acoustic first wave arrival time sequence and the vibration signal is performed by multi-scale wavelet transform, the mutation point is detected to optimize the feature reliability, and the time axis of different sensors is aligned by dynamic time warping and cross-spectral entropy to extract effective feature segments.

[0035] The acoustic wave and vibration signal generated by the leak point of the buried pipeline are non-stationary signals, and direct analysis cannot capture detailed features.

[0036] In some embodiments, the distributed optical fiber acoustic sensor DAS and the nano thin film vibration sensor array data are wavelet packet decomposed to generate 3 layers of time-frequency subbands, including: low frequency layer, medium frequency layer, high frequency layer; Daubechies wavelet basis function is used, decomposition series J=3, and multi-resolution analysis of non-stationary signals is realized.

[0037] In each subband, the signal mutation point is detected, the wavelet coefficient modulus maximum value is calculated, and the signal mutation point is located. The mutation point is direct evidence of the existence of the leak point, and the mutation point includes: pressure step, vibration peak. The mutation points are independently detected in the low frequency layer, the medium frequency layer, and the high frequency layer to avoid single frequency band noise interference.

[0038] Through cross-subband mutation point correlation analysis, false features in a single frequency band are excluded.

[0039] Through layered detection, single frequency band noise interference can be avoided to optimize feature reliability.

[0040] The sampling frequencies of the distributed acoustic sensor (DAS) and the nano thin film vibration sensor array can be different, resulting in a mismatch in the timestamps of the same physical event in the data of the two sensors;

[0041] In some embodiments, dynamic time warping (DTW) is used to stretch or compress the feature point sequences of the distributed acoustic sensor (DAS) and the nano thin film vibration sensor array on the time axis, the matching distance is minimized, the Sakoe-Chiba bandwidth constraint is used to limit the DTW search range, avoid global search, reduce the computational difficulty, improve real-time performance, the window size r is set to 0.1N, where N is the sequence length; and the turning points on the optimal path of DTW are selected as time synchronization marker points;

[0042] The cross power spectral density (CPSD) of the data segments 200 ms before and after the synchronization marker points is calculated, and the calculation formula is as follows: Where A(f) and B(f) are the results of windowed fast Fourier transform, K is the number of segments, and B k * (f) represents the complex conjugate of B(f), A is the distributed acoustic sensor (DAS) signal data, and B is the nano thin film vibration sensor array signal data;

[0043] The cross spectral entropy (CSE) is defined to reflect the similarity of the two signals in the frequency domain, and the specific formula is as follows: CSExy=-Σ f P xy (f)logP xy (f), where, The calculated cross spectral entropy (CSE) value is analyzed, and if the cross spectral entropy (CSE) value is greater than 0.85, it is determined that the effective synchronization feature segment is extracted;

[0044] After wavelet packet decomposition and dynamic time warping (DTW) alignment, the distributed acoustic sensor (DAS) and the nano thin film vibration sensor array data are synchronized on the time axis and the frequency components are separated, ensuring that the subsequent cross spectral entropy (CSE) analysis can accurately calculate the similarity of the two signals in the frequency domain, and avoiding false positives due to time sequence or frequency mismatch;

[0045] S3: The part of the multi-physical field data after time axis alignment and effective feature extraction is analyzed by using the long short-term memory-causal network to analyze the correlation between the temperature gradient data and the vibration energy, a physical relationship model is constructed, a temperature influence factor is calculated, the vibration energy is compensated using the temperature influence factor, and the noise introduced by the environmental temperature fluctuation is eliminated;

[0046] Firstly, the temperature interference is eliminated by long short-term memory-causal network LSTM-CN, and the physical relationship model of temperature gradient and vibration energy attenuation is constructed; then the spatio-temporal features are processed by the Transformer network, and finally high-quality input is provided for the physically constrained positioning model to improve the leak positioning accuracy;

[0047] The data sequence collected by the distributed optical fiber temperature sensor DTS is processed by sliding window, the temperature gradient is calculated, and the temperature gradient sequence is calculated by multiple temperature gradients;

[0048] The vibration signal data collected by the distributed optical fiber acoustic sensor DAS is wavelet denoised, and the vibration energy feature sequence is extracted;

[0049] The temperature gradient sequence and the vibration energy feature sequence are input into the input layer of the long short-term memory network LSTM, and the bidirectional LSTM unit in the hidden layer is used to capture the time sequence relationship between temperature and vibration. The output layer outputs the temperature influence factor a, which represents the influence weight of temperature change on vibration signal;

[0050] In the causal network CN, the causal graph G=(V, E) of temperature gradient and vibration energy is constructed, where V={T g ,H} H vibration energy, representing the energy features related to vibration, T g is the temperature gradient, reflecting the change of temperature, and E represents the causal relationship strength; the causal structure is learned by PC algorithm, and the causal strength is calculated by conditional mutual information CMI: CMI(T g ;E v |T g-1 ) = H(T g ;E v |T g-1 )-H(T g |T g-1 )-H(E v |T g-1 ), where H(·) is the information entropy, T g is the temperature gradient, E v is the H vibration energy, and T g-1 is the temperature gradient at the previous moment. When CMI(T g ;E v |T g-1 )>0.3, it indicates that the change of temperature gradient has substantial influence on vibration energy;

[0051] The vibration energy after temperature compensation is calculated, which is specifically calculated by the formula: H'=H·(1-α t ·|T g |) to eliminate the interference of temperature fluctuation on vibration signal, where α tis the temperature gradient, which represents the change of temperature, and is taken as an absolute value to focus on the amplitude of temperature change rather than the direction; g Temperature gradient, representing the change of temperature, is taken as an absolute value to focus on the amplitude of temperature change rather than the direction;

[0052] Feature enhancement is performed on the compensated vibration signal: where σ T is the temperature gradient standard deviation, used for normalizing the temperature gradient |T g | to reasonably control the amplitude of feature enhancement and highlight the vibration features of temperature anomaly areas, and H' is the vibration energy after temperature compensation, T g is the temperature gradient;

[0053] Traditional positioning algorithms do not make full use of the topological relationship of pipe networks, wasting computing resources in high leakage risk areas;

[0054] In some embodiments, the time flow Transformer is used to process the TDOA sound wave arrival time sequence, and the multi-path reflection interference is identified through the self-attention mechanism; the spatial flow Transformer is used to encode the GIS pipe network topology into a graph structure, and the GAT network is used to weight the high leakage risk nodes to assist the positioning algorithm in preferentially searching the high-risk areas;

[0055] The calibrated multi-modal data is filtered through the attention mechanism, and the spatio-temporal domain joint feature matrix is output as the input of the downstream positioning model;

[0056] The calibrated multi-modal data sound wave features, temperature gradient sequence, vibration energy feature sequence, and real-time pressure derivative are weighted and filtered through the attention mechanism,

[0057] When the temperature gradient sequence shows an anomaly, the weight of the vibration energy feature is reduced;

[0058] When the pressure derivative shows a step, the weight of the sound wave feature is increased, and the pressure step directly reflects the transient pressure wave generated by the leakage, which is strongly correlated with the sound wave feature;

[0059] According to the real-time data quality, the importance of the features is dynamically adjusted to avoid the dominance of a single modal data anomaly in the positioning result;

[0060] The filtered features are integrated into a spatio-temporal domain joint feature matrix, with rows representing time sequences and columns representing spatial nodes, and each element containing multiple physical field features;

[0061] S4: The vibration energy feature after eliminating temperature interference, sound wave feature, and spatio-temporal domain joint feature matrix are cooperated through the optimization algorithm constrained by physical equation and the generative adversarial network to realize leak point positioning;

[0062] Based on the Navier-Stokes equation constraint PSO algorithm, in the particle swarm optimization process, the pipeline fluid dynamic equation is introduced, the pressure wave propagation speed correction coefficient caused by leakage is calculated according to the real-time pressure derivative and material parameter, and the particle search step is dynamically adjusted; The gradient direction is automatically generated by the accompanying method, so that the particle moving path meets the mass conservation equation, and the invalid iteration is reduced;

[0063] Specifically, define each particle to represent the potential leak point coordinates (x, y, z), and the dimension corresponds to the pipe network space coordinate system;

[0064] Physical parameter input, including: real-time pressure derivative Pipe material parameters and fluid properties; Fluid properties include: water density, water viscosity; Pipe material parameters include: material type, elastic modulus F, unit GPa, wall thickness h, unit mm, inner diameter D, unit mm;

[0065] Based on the Navier-Stokes equation of the propagation speed modeling, the pressure wave speed c caused by leakage satisfies the simplified wave equation: Wherein, K is the fluid bulk modulus, D is the pipe inner diameter, h is the pipe wall thickness, h is the pipe wall thickness, which refers to the thickness of the pipe wall, F is the elastic modulus of the pipe material, which measures the ability of the pipe material to resist elastic deformation;

[0066] It should be noted that the elastic modulus F is strongly related to the pipe material, and the value of the metal pipe is usually 100-300 times that of the plastic pipe, which directly affects the calculation result of the pressure wave speed c;

[0067] According to the real-time pressure derivative And the material parameter, the correction coefficient λ is calculated: Wherein, η is the correction factor, which is used to dynamically adjust the wave speed estimate value;

[0068] Particle search step update, particle velocity v i And position x i Update formula: Wherein, ω is the inertia weight, The velocity of particle i at t+1 time, c1, c2 is the learning factor, r1, r2 is a random number between 0 and 1, which increases the randomness of the search process and avoids the algorithm falling into local optimum, λ is the correction coefficient, pbest i The individual optimal position of particle i is the best position found in the search process of the particle itself, and gbest is the global optimal position found in the search process of all particles, The position of particle i at t+1 time is obtained by updating the current position And velocity Iteration;

[0069] The calculation result of the particle step length is consistent with the actual leakage propagation rule by introducing the physical wave speed constraint;

[0070] The target function f(x) is constructed by automatically deriving by the adjoint method, and the multi-sensor TDOA positioning error is calculated by the adjoint method. The gradient Δf(x) guides the particle to move in the direction of the fastest error descent, and reduces blind search: Wherein, α is the learning rate, The gradient of the target function f(x) at guides the particle to move in the direction of the fastest error descent;

[0071] The mass conservation equation constraint forces the particle movement path to meet: The particle velocity is projected to the divergence-free subspace by the projection method to ensure that the search path meets the physical law and avoid generating invalid solutions that violate fluid mechanics;

[0072] Data input and physical criterion definition, input layer input pipe network parameter vector: Including the original pressure data collected by the high-precision pressure sensor in real time, the temperature gradient T g is the temperature gradient along the pipeline monitored by the distributed optical fiber temperature sensor DTS, and E is the pipe material parameter;

[0073] Generator and discriminator architecture, input pipe network parameters θ plus random noise z; output mixed waveform s gen =G(θ,z), including simulated leak point signal and noise, wherein the pipe network parameters include: pipe internal pressure data collected by high-precision pressure sensor in real time, pipe temperature gradient data along the pipeline monitored by distributed optical fiber temperature sensor DTS, and pipe material related characteristic parameters;

[0074] Physical constraint embedding, adding wave equation penalty term in generator loss function: Wherein, E θ,z The expectation operation on θ and z, is the partial derivative of the mixed waveform s gen with respect to time t, reflecting the rate of change of the waveform with time, and c is the pressure wave speed generated by the leak, is the Laplacian of the mixed waveform s gen , describing the second-order change of the waveform in space, input s gen , output true or false probability D(s);

[0075] Adversarial training and interference filtering, training target: Wherein, λ phys is the physical constraint weight, balancing data authenticity and physical rationality, min GThe generator G is optimized to generate waveforms that can fool the discriminator as much as possible, max D The discriminator D is optimized to distinguish real from fake waveforms more accurately, E θ,z Again, the expectation operation is performed on θ and z;

[0076] Interference filtering mechanism, after training, the discriminator of PI-GAN can identify interference signals that do not conform to physical laws, and improve the purity of the input positioning model signal;

[0077] For example, when D(s) < 0.6, it is determined to be an interference signal;

[0078] Multi-modal feature input and leak point coordinate inversion, input the spatio-temporal domain joint feature matrix data processed by LSTM-CN and Transformer, the dimension is: T x N x M, where T is the time step, N is the number of sensor nodes, that is, the number of sensors used to collect multi-modal data, and M is the number of physical field modalities; The spatio-temporal domain joint feature matrix data includes: TDOA sequence, vibration energy compensation value, GIS topological encoded node risk weight, sensor position coordinates;

[0079] PSO and PI-GAN collaborative positioning steps:

[0080] Initialization search: PSO particles are randomly distributed in the pipe network space, and the matching degree of the initial feature matrix and the particle position is calculated, and the error of TDOA time difference and theoretical propagation time;

[0081] PI-GAN pre-screening: for high-error particles, use PI-GAN to generate the theoretical leakage waveform at that position, and compare it with the actual feature matrix to exclude physically infeasible solutions;

[0082] Further iteration optimization: PSO updates the particle position according to the Navier-Stokes constraint, and PI-GAN filters candidate solutions that violate the wave equation in real time;

[0083] Convergence determination: when the positioning error of the optimal solution of the particle swarm is less than a threshold value, such as 0.5 meters, or the number of iterations reaches the upper limit, output the leak point coordinates (x * ,y * ,z * );

[0084] Simulate the leakage scenario at this coordinate using COMSOL, calculate the original pressure data, temperature gradient distribution along the pipeline, and compare the sensor position coordinates with the actually collected multi-modal feature matrix to verify the physical consistency of the positioning results: Where f sim is the simulation feature, and f real,m is the measured feature;

[0085] The sensor position coordinates are: distributed optical fiber acoustic wave sensor DAS, distributed optical fiber temperature measurement sensor DTS, nanometer film vibration sensor array, and high-precision pressure sensor in the specific coordinate values in the pipe network.

[0086] If the verification index is lower than the preset threshold, it is determined that the leak point positioning is effective.

[0087] The technical scheme of the embodiment of the application is: deploying distributed optical fiber acoustic wave DAS, temperature measurement sensor DTS, nanometer film vibration sensor array, and high-precision pressure sensor, collecting acoustic wave, temperature, vibration, and pressure data, enhancing signal coupling by adding sonar patches to plastic pipelines, and performing data processing and feature fusion: detecting signal mutation points through wavelet packet decomposition, aligning time sequence differences by using dynamic time warping DTW and cross-spectral entropy CSE; eliminating temperature interference by means of long short-term memory-causal network LSTM-CN, and generating a spatiotemporal domain joint feature matrix in combination with a Transformer network and an attention mechanism.

[0088] Based on the cooperation of the particle swarm optimization PSO algorithm constrained by the Navier-Stokes equation and the physical information generation adversarial network PI-GAN, the leak point coordinates are high-precision inverted by pressure wave velocity modeling, gradient guidance by the adjoint method, and disturbance filtering by the wave equation, the positioning error is less than or equal to 0.5 meters, and the physical consistency is verified by COMSOL simulation. This scheme combines multi-modal data and intelligent algorithms to improve detection accuracy and engineering adaptability.

[0089] Embodiment 2

[0090] As shown in Figure 2 Based on embodiment 1, the application provides an intelligent water use data acquisition and control system, which comprises the following modules:

[0091] A multi-physical field data acquisition module: sensors are arranged to acquire multi-physical field data, and the multi-physical field data includes: leak point acoustic wave first wave arrival time, pipeline temperature gradient along the line, vibration signal, and pressure data.

[0092] Distributed optical fiber acoustic wave sensors DAS are arranged at a preset distance along the pipeline to extract the leak point acoustic wave first wave arrival time, and distributed optical fiber temperature measurement sensors DTS are integrated to monitor the pipeline temperature gradient along the line. In view of the low acoustic impedance characteristic of the plastic pipeline, the transmission efficiency of the leak sound to the sensor is improved by the sonar patch, and the coupling of the leak sound signal is enhanced. It should be noted that for metal pipelines, distributed optical fiber acoustic wave sensors DAS, distributed optical fiber temperature measurement sensors DTS, and nanometer film vibration sensor arrays can be directly deployed.

[0093] A nano-thin film vibration sensor array is arranged on the outer wall of the pipeline to monitor the vibration signals generated by the leakage point, high-precision pressure sensors are installed at preset distances along the pipeline, real-time pressure data is collected and real-time pressure derivatives are calculated in units of Pa / s.

[0094] The multi-modal data space-time alignment module: based on the collected multi-physical field data, the time-frequency decomposition of the first wave arrival time sequence and the vibration signal containing the leakage point sound wave is carried out through multi-scale wavelet transform, the mutation point is detected to optimize the feature reliability, and the time axis of different sensors is aligned by using dynamic time warping and cross-spectral entropy, and the effective feature segment is extracted;

[0095] The data of the distributed optical fiber acoustic sensor DAS and the nano-thin film vibration sensor array are wavelet packet decomposed, the signal mutation points are detected in each sub-band, the wavelet coefficient modulus maximum is calculated, the signal mutation points are located, and the feature reliability is optimized;

[0096] The dynamic time warping DTW is used to stretch or compress the time axis of the feature point sequence of the distributed optical fiber acoustic sensor DAS and the nano-thin film vibration sensor array, the Sakoe-Chiba bandwidth constraint is used to minimize the matching distance, the DTW search range is limited, the global search is avoided, the calculation difficulty is reduced, the real-time performance is improved, the cross power spectral density CPSD of the data segment of 200 ms before and after the synchronous marker point is calculated, the cross-spectral entropy CSE is defined to reflect the similarity of the two signals in the frequency domain, and if the cross-spectral entropy CSE value is greater than 0.85, it is determined as an effective synchronization feature segment, and the effective feature segment is extracted;

[0097] The temperature interference elimination and feature fusion module: part of the multi-physical field data after time axis alignment and effective feature extraction is used to analyze the correlation between temperature gradient data and vibration energy through long short-term memory-causal network, a physical relationship model is constructed, a temperature influence factor is calculated, the vibration energy is compensated by using the temperature influence factor, and the noise introduced by environmental temperature fluctuation is eliminated.

[0098] The temperature interference is eliminated by the long short-term memory-causal network LSTM-CN, a physical relationship model of temperature gradient and vibration energy attenuation is constructed, then the Transformer network is used to process the space-time features, and finally high-quality input is provided for the physical constraint positioning model, and the leakage point positioning accuracy is improved.

[0099] The data sequence collected by the distributed optical fiber temperature measurement sensor DTS is processed by a sliding window, the temperature gradient is calculated, and the temperature gradient sequence is obtained by calculating multiple temperature gradients.

[0100] The vibration signal data collected by the distributed optical fiber acoustic sensor DAS is wavelet denoised, and the vibration energy feature sequence is extracted.

[0101] The temperature gradient sequence and the vibration energy feature sequence are input into the input layer of the long short-term memory network (LSTM), the time sequence relationship between the temperature and the vibration is captured through the bidirectional LSTM unit in the hidden layer, and the temperature influence factor a is output from the output layer, which represents the influence weight of the temperature change on the vibration signal.

[0102] In the causal network (CN), a causal graph of the temperature gradient and the vibration energy is constructed, the causal structure is learned through the PC algorithm, the causal strength is calculated using the conditional mutual information (CMI), and when the causal strength is greater than a threshold value, it is indicated that the temperature gradient change has a substantial impact on the vibration energy.

[0103] The vibration energy after temperature compensation is calculated, and the specific formula is: H' = H · (1-α t ·|T g |) eliminates the interference of temperature fluctuations on the vibration signal, wherein α t is the temperature influence factor, which is dynamically learned through the LSTM network, H is the original vibration energy, T g is the temperature gradient, and the absolute value is taken to focus on the temperature change amplitude rather than the direction.

[0104] The compensated vibration signal is enhanced in features: wherein σ T is the standard deviation of the temperature gradient, which is used to normalize the temperature gradient |T g | to reasonably control the amplitude of feature enhancement and highlight the vibration features of the temperature abnormal area, and H' is the vibration energy after temperature compensation, and T g is the temperature gradient.

[0105] Traditional positioning algorithms do not make full use of the topological relationship of the pipe network, and waste computing resources in high leakage risk areas.

[0106] The time flow Transformer is used to process the TDOA sound wave arrival time sequence, and the self-attention mechanism is used to identify multi-path reflection interference; the spatial flow Transformer is used to encode the GIS pipe network topology into a graph structure, and the GAT network is used to weight the high leakage risk nodes to assist the positioning algorithm in preferentially searching the high-risk area.

[0107] The calibrated multi-modal data is filtered through the attention mechanism, and a spatio-temporal feature matrix is output as the input of the downstream positioning model.

[0108] The calibrated multi-modal data sound wave features, temperature gradient sequence, vibration energy feature sequence, and real-time pressure derivative are weighted and filtered through the attention mechanism,

[0109] When the temperature gradient sequence shows an anomaly, the weight of the vibration energy feature is reduced.

[0110] When the pressure derivative appears a step, the weight of the acoustic feature is increased, and the pressure step directly reflects the transient pressure wave generated by the leakage, which is strongly related to the acoustic feature;

[0111] According to the real-time data quality, the importance of the feature is dynamically adjusted to avoid the single modal data anomaly dominating the positioning result;

[0112] The filtered features are integrated into a spatiotemporal feature matrix, with rows representing time series and columns representing spatial nodes, and each element containing multiple physical field features.

[0113] Physical constraint intelligent positioning module: the vibration energy feature, acoustic feature and spatiotemporal feature matrix after eliminating temperature interference are combined with the optimization algorithm based on physical equation constraint and the generative adversarial network to realize leak point positioning.

[0114] Based on the PSO algorithm constrained by the Navier-Stokes equation, the particle swarm optimization process is introduced into the pipe fluid dynamic equation, and the pressure wave propagation speed correction coefficient caused by the leakage is calculated according to the real-time pressure derivative and material parameters to dynamically adjust the particle search step. The gradient direction is automatically generated by the adjoint method, so that the particle movement path conforms to the mass conservation equation and reduces invalid iterations.

[0115] Data input and physical criterion definition: the input layer inputs the pipe network parameter vector, the generator and discriminator architecture, the input pipe network parameter adds random noise, and the output mixed waveform adds a wave equation penalty term in the generator loss function. The input mixed waveform outputs the true and false probability.

[0116] Adversarial training and interference filtering: training target: where λ phys is the physical constraint weight, balancing data authenticity and physical reasonableness, min G Optimize the generator G to make the generated waveform as possible to deceive the discriminator, max D Optimize the discriminator D to make it more accurate to distinguish between real and false waveforms, E θ,z Again, the expectation operation on θ and z is performed.

[0117] PSO and PI-GAN collaborative positioning steps:

[0118] Initialization search: PSO particles are randomly distributed in the pipe network space, and the matching degree of the initial feature matrix and the particle position is calculated, and the error of TDOA time difference and theoretical propagation time.

[0119] PI-GAN pre-screening: for high-error particles, use PI-GAN to generate the theoretical leakage waveform at this position, compare it with the actual feature matrix, and exclude physically infeasible solutions.

[0120] Further iteration optimization: PSO updates particle position according to Navier-Stokes constraint, and PI-GAN filters candidate solutions violating wave equation in real time;

[0121] Convergence determination: when the positioning error of the particle swarm optimal solution is less than a threshold value, such as 0.5 meters, or the number of iterations reaches an upper limit, output the leak point coordinates (x * ,y * ,z * );

[0122] Simulate the leakage scenario at the coordinates using COMSOL, calculate the original pressure data, temperature gradient distribution along the pipeline, and compare the sensor position coordinates with the actually collected multi-modal feature matrix to verify the physical consistency of the positioning results: Where f sim is the simulation feature, and f real,m is the measured feature;

[0123] The sensor position coordinates are: the specific coordinate values of the distributed acoustic sensor (DAS), the distributed temperature sensor (DTS), the nanometer thin film vibration sensor array, and the high-precision pressure sensor in the pipeline network;

[0124] If the verification index is lower than the preset threshold value, it is determined that the leak point positioning is effective;

[0125] Embodiment 3

[0126] As shown in Figure 3 , based on Embodiment 1 and Embodiment 2, the present application provides an intelligent water use data acquisition and control terminal device, which comprises:

[0127] a distributed acoustic sensor, a distributed temperature sensor, a nanometer thin film vibration sensor array, a high-precision pressure sensor, and a computer processing unit;

[0128] The distributed acoustic sensor is used to collect the first wave arrival time of the acoustic wave of the leak point.

[0129] The distributed temperature sensor is used to monitor the temperature gradient along the pipeline.

[0130] The nanometer thin film vibration sensor array is used to monitor the vibration signal generated by the leak point.

[0131] The high-precision pressure sensor is used to collect pressure data.

[0132] The computer processing unit utilizes the computer program to process the data collected by the distributed optical fiber acoustic sensor, the distributed optical fiber temperature sensor, the nanometer film vibration sensor array and the high-precision pressure sensor, and the computer program can execute the operation process required by any step or module in the intelligent water use data acquisition and control method and system according to the embodiments 1 and 2.

[0133] The distributed optical fiber acoustic sensor, the distributed optical fiber temperature sensor, the nanometer film vibration sensor array and the high-precision pressure sensor are connected with the computer processing unit through a CAN bus.

[0134] The basic principles, main features and advantages of the present application are shown and described above. It should be understood by those skilled in the art that the present application is not limited by the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent water use data acquisition and control, characterized in that: Comprise: S1: arrange sensor to collect multi-physical field data, the multi-physical field data includes: leak point sound wave first wave arrival time, pipeline temperature gradient along the line, vibration signal, pressure data; S2: based on the multi-physical field data collected, the time-frequency decomposition of the first wave arrival time sequence and the vibration signal containing the leak point sound wave is carried out through multi-scale wavelet transform, the reliability of the mutation point optimization feature is detected, and the time axis of different sensors is aligned using dynamic time warping and cross-spectral entropy, and the effective feature segment is extracted; S3: after the time axis alignment and effective feature extraction, part of the multi-physical field data is analyzed by long short-term memory-causal network to analyze the correlation between temperature gradient data and vibration energy, a physical relationship model is constructed, a temperature influence factor is calculated, and the vibration energy is compensated using the temperature influence factor to eliminate the noise introduced by environmental temperature fluctuation; S4: the vibration energy after eliminating temperature interference is combined with the space-time domain joint feature matrix data to realize leak point positioning through the optimization algorithm and the generative adversarial network constrained by physical equation; The specific process of realizing leak point positioning is: Multi-modal feature input and leak point coordinate inversion, input the space-time domain joint feature matrix data processed by long short-term memory-causal network LSTM-CN and time flow Transform, and locate by PSO and PI-GAN, PSO updates particle position according to Navier-Stokes constraint, PI-GAN filters candidate solutions that violate wave equation in real time, when the positioning error of the optimal solution of the particle group is less than the threshold value, output the leak point coordinates.

2. The intelligent water data acquisition control method according to claim 1, characterized in that: The specific process of optimizing feature reliability is: The distributed acoustic sensor DAS and nano thin film vibration sensor array data are wavelet packet decomposed to generate 3 layers of time-frequency subbands, the wavelet coefficient modulus maximum is calculated in the 3 layers of time-frequency subbands respectively, the mutation points are independently detected, the false features of single frequency band are excluded through cross-subband mutation point correlation analysis, and the feature reliability is optimized.

3. The intelligent water data acquisition control method according to claim 1, characterized in that: The specific process of extracting effective feature segment is: Dynamic time warping is used to stretch or compress the feature point sequence of the distributed acoustic sensor and the nano thin film vibration sensor array, Sakoe-Chiba bandwidth constraint is used to minimize the matching distance, the DTW search range is limited, global search is avoided, and the turning points on the DTW optimal path are selected as time synchronization marker points; The cross power spectral density CPSD of the 200ms data segment before and after the synchronization marker point is calculated, the cross-spectral entropy CSE value calculated is analyzed, if the cross-spectral entropy CSE value is greater than 0.85, it is determined that it is an effective synchronization feature segment, and the effective feature segment is extracted.

4. The intelligent water data acquisition control method according to claim 1, characterized in that: The specific process of calculating the temperature influence factor is: Through long short-term memory-causal network LSTM-CN, the temperature interference is eliminated, the data sequence collected by the distributed fiber temperature sensor DTS is processed by sliding window, the temperature gradient is calculated, and the temperature gradient sequence is obtained by calculating multiple temperature gradients; The vibration signal data collected by the distributed acoustic sensor DAS is wavelet denoised, and the vibration energy feature sequence is extracted; The temperature gradient sequence and the vibration energy feature sequence are input into an input layer of a long short-term memory network (LSTM), the time sequence relationship between the temperature and the vibration is captured through a bidirectional LSTM unit in a hidden layer, and a temperature influence factor is output by an output layer.

5. The method of claim 1, wherein: The specific method for eliminating the noise introduced by the environmental temperature fluctuation is: Utilize the formula for: Eliminate the interference of environmental temperature fluctuation on vibration signal, wherein, is a temperature influence factor, which is dynamically learned by an LSTM network, H is the original vibration energy, Temperature gradient, indicating the change of temperature, taking the absolute value to pay attention to the amplitude of temperature change, not the direction.

6. The intelligent water usage data collection method of claim 1, wherein: The specific optimization algorithm constrained by the physical equation is: In the PSO algorithm constrained by the Navier-Stokes equation, the pipe material parameters are introduced when the pressure wave velocity is calculated, the theoretical wave velocity is calculated through a formula, and the particles are constrained by using a correction coefficient, so that the particle updating conforms to the fluid mechanics law.

7. The intelligent water usage data collection method of claim 1, wherein: The spatiotemporal domain joint feature matrix data include: a TDOA sequence, a vibration energy compensation value, a node risk weight coded by GIS topology, and a sensor position coordinate.

8. A smart water usage data acquisition control system characterized by: The method comprises the following steps: a multi-physical field data acquisition module: arranging sensors to acquire multi-physical field data, wherein the multi-physical field data include a leak point sound wave first wave arrival time, a pipeline temperature gradient along a line, a vibration signal, and pressure data; a multi-modal data spatiotemporal alignment module: based on the acquired multi-physical field data, performing time-frequency decomposition on the leak point sound wave first wave arrival time sequence and the vibration signal through multi-scale wavelet transform, detecting a mutation point to optimize feature reliability, and aligning time axes of different sensors by using dynamic time warping and cross-spectral entropy to extract effective feature fragments; a temperature interference elimination and feature fusion module: based on the multi-physical field data after the time axis alignment and the effective feature extraction, analyzing the correlation between the temperature gradient data and the vibration energy through a long short-term memory-causal network, constructing a physical relationship model, calculating a temperature influence factor, compensating the vibration energy by using the temperature influence factor, and eliminating the noise introduced by the environmental temperature fluctuation; a physical constraint intelligent positioning module: based on the vibration energy features, the sound wave features, and the spatiotemporal domain joint feature matrix after the temperature interference elimination, realizing leak point positioning through an optimization algorithm constrained by a physical equation and a generative adversarial network. The specific process for realizing leak point positioning comprises the following steps: multi-modal feature input and leak point coordinate inversion: inputting the spatiotemporal domain joint feature matrix data processed by the LSTM-CN and the time flow Transformer, and positioning by using the PSO and the PI-GAN, wherein the PSO updates the particle position according to the Navier-Stokes constraint, the PI-GAN filters candidate solutions that violate the wave equation in real time, and when the positioning error of the optimal solution of the particle group is less than a threshold value, the leak point coordinates are output.

9. A smart water use data acquisition control terminal device, characterized by: The method comprises the following steps: a distributed optical fiber acoustic sensor, a distributed optical fiber temperature sensor, a nano thin film vibration sensor array, a high-precision pressure sensor, and a computer processing unit; The computer processing unit is used to execute a computer program of the intelligent water use data acquisition and control method according to any one of claims 1 to 7.

Images