Pretreatment and detection method for preventing generation of biological membrane of newly-built waterway system

Abstract

The invention discloses a pretreatment and detection method for preventing generation of a biological membrane of a newly-built waterway system, and belongs to the technical field of pipeline integrity detection. The method comprises the steps that after gas pressure maintaining is conducted on the waterway system, an actually-measured acoustic signal on the pipe wall of the waterway system is obtained; processing the actually measured acoustic signal based on a preset acoustic fingerprint database associated with a digital model of the waterway system to generate a leakage probability distribution representing the leakage probability of each area on the waterway system; and according to the leakage probability distribution, determining one or more high-probability leakage candidate areas to guide the targeted check of the one or more high-probability leakage candidate areas so as to determine the accurate position of the leakage point. According to the method, through the mode of combining acoustic coarse positioning and targeted checking, the pipeline structure does not need to be modified, the search space of the leakage points is greatly compressed, and the efficiency and accuracy of positioning the leakage points of the newly-built waterway system are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pipeline integrity detection, and in particular to a pretreatment and detection method for preventing biofilm generation in a newly built water system. BACKGROUND

[0002] In the new or renovation project of key areas such as medical institutions, the integrity and cleanliness of the water system are of great importance. Before being put into use, it is necessary to ensure that the pipeline is free of any leakage. Traditionally, the closed water test is a common method for detecting leakage, but this method requires filling the pipeline with water. If the system is idle for a long time after the test, the wet environment on the inner wall of the pipeline is prone to breed bacteria, fungi and other microorganisms, forming biofilm that is difficult to remove. Biofilm not only pollutes the water quality of subsequent water flow, but also may increase the fluid resistance and corrode the pipeline. In order to avoid this problem, using gas (such as compressed air or nitrogen) for pressure test becomes a more superior pretreatment method. However, although the gas pressure test can determine whether the system has leakage, it is difficult to locate the specific position of the leakage point. Once the pressure drop is detected, the engineering personnel often need to conduct a carpet-like investigation on the large and complex pipe network by smearing soap water or using an ultrasonic leak detector. This process is time-consuming and labor-intensive, especially in the case where the pipeline has been shielded by walls or suspended ceilings, the positioning work becomes extremely difficult and inefficient. Therefore, how to quickly and accurately locate the gas leakage point in the newly built water system without structural modification and water filling is a key technical problem to prevent biofilm generation from the source. SUMMARY

[0003] The present application aims to provide a pretreatment and detection method, system and computer readable storage medium for preventing biofilm generation in a newly built water system, which aims to solve the technical problem of being difficult to quickly and accurately locate the leakage point after the gas pressure test as proposed in the background.

[0004] To achieve the above-mentioned purpose, in a first aspect, the present application provides a pretreatment and detection method for preventing biofilm generation in a newly built water system, comprising: obtaining an acoustic fingerprint library associated with a digital model of a to-be-tested water system, wherein the acoustic fingerprint library contains acoustic fingerprint features of a plurality of preset nodes on the water system; after the water system is subjected to gas pressure, collecting a measured acoustic signal through an acoustic sensing device deployed on the water system; processing the measured acoustic signal to extract a measured acoustic feature vector; matching the measured acoustic feature vector with the acoustic fingerprint features in the acoustic fingerprint library to generate a leakage probability distribution representing the leakage probability of each region on the water system; and determining one or more high-probability leakage candidate regions according to the leakage probability distribution to guide targeted verification of the one or more high-probability leakage candidate regions, thereby determining the accurate position of a leakage point.

[0005] Optionally, the step of obtaining the acoustic fingerprint library associated with a digital model of a water system under test comprises: obtaining a geometric design drawing of the water system, and parsing the geometric design drawing to construct the digital model, the digital model containing a topological structure, geometric parameters and physical parameters of the water system; and based on the digital model, respectively calculating the acoustic fingerprint features for the plurality of preset nodes by a preset acoustic propagation analysis method.

[0006] Optionally, the geometric parameters include pipe length, diameter, wall thickness, and the positions and types of elbows and tees; and the physical parameters include the acoustic velocity and acoustic attenuation coefficient of the pipe material.

[0007] Optionally, the step of processing the measured acoustic signal comprises: band-pass filtering the measured acoustic signal to filter out noise outside a preset frequency band; and performing time-frequency analysis on the filtered measured acoustic signal to extract the measured acoustic feature vector, the measured acoustic feature vector containing at least one of peak frequency, energy spectrum distribution and signal envelope shape.

[0008] Optionally, the step of matching the measured acoustic feature vector with the acoustic fingerprint features in the acoustic fingerprint library comprises: calculating a similarity score between the measured acoustic feature vector and each of the acoustic fingerprint features in the acoustic fingerprint library; and based on the similarity score, converting the similarity score to the leakage probability distribution by a normalization function.

[0009] Optionally, the normalization function is a Softmax function.

[0010] Optionally, the step of guiding the targeted verification of the one or more high-probability leakage candidate regions comprises: guiding an operator to use a thermal imaging device to scan the one or more high-probability leakage candidate regions; and determining the precise location of the leakage point based on a low-temperature abnormal point detected by the thermal imaging device due to gas throttling effect.

[0011] In a second aspect, the application provides a pretreatment and detection system for preventing biofilm generation in a newly-built water system, comprising: a modeling module configured to obtain an acoustic fingerprint library associated with a digital model of a water system to be detected, wherein the acoustic fingerprint library comprises acoustic fingerprint features of a plurality of preset nodes on the water system; a collection module configured to collect a measured acoustic signal by an acoustic sensing device deployed on the water system after gas pressure maintenance is performed on the water system; a processing module configured to process the measured acoustic signal to extract a measured acoustic feature vector; an analysis module configured to match the measured acoustic feature vector with the acoustic fingerprint features in the acoustic fingerprint library to generate a leakage probability distribution representing leakage probabilities of regions on the water system; and a guidance module configured to determine one or more high-probability leakage candidate regions according to the leakage probability distribution, to guide targeted verification of the one or more high-probability leakage candidate regions, and to determine an accurate position of a leakage point.

[0012] Optionally, the acoustic sensing device of the collection module comprises a piezoelectric contact sensor and a low-noise preamplifier.

[0013] Optionally, the guidance module is further configured to: highlight the high-probability leakage candidate regions on the digital model of the water system on a three-dimensional display interface; and guide an operator to perform the targeted verification based on the highlighting using a thermal imaging device.

[0014] In a third aspect, the application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method of the first aspect.

[0015] The application has the following beneficial effects: by pre-constructing an acoustic fingerprint library of a pipe network and matching a leakage acoustic signal collected on site with the acoustic fingerprint library, a problem of global pressure drop without directionality can be converted into a positioning problem with spatial probability directionality. The method converts a traditional carpet method and linear search process into a process of one calculation plus several targeted verifications, greatly compresses the search space, shortens the leakage positioning time, and reduces the labor cost. Meanwhile, the entire detection process is completed in a gas medium, completely avoids pipe water accumulation, fundamentally eliminates the risk of biofilm generation caused by testing, and guarantees the highest cleanliness standard of a newly-built water system in the initial stage. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, brief introductions will be given to the drawings needed in the embodiment descriptions.

[0017] Figure 1A flowchart of a pretreatment and detection method for preventing biofilm generation in a newly-built water system according to an embodiment of the present application is shown in FIG. 1. Figure 2 A functional module and data flow diagram of a pretreatment and detection system for preventing biofilm generation in a newly-built water system according to an embodiment of the present application is shown in FIG. 2. Figure 3 A digital model of a water system to be detected according to an embodiment of the present application is shown in FIG. 3, including a topological structure composed of nodes and edges and associated geometric and physical parameters. Figure 4 A detailed flowchart of generating an acoustic fingerprint feature for a single preset node according to an embodiment of the present application is shown in FIG. 4. DETAILED DESCRIPTION

[0018] To make the objectives, technical solutions, and advantages of the present application clearer, the technical solutions of the present application will be described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0019] In addition, the terms "first" and "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first" and "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0020] The embodiments of the present application provide a pretreatment and detection method for preventing biofilm generation in a newly-built water system. In a specific implementation, the method converts a globally directionless pressure drop signal into a spatially directional and probabilistic leak location prediction by using acoustic fingerprint matching technology, and then uses high-precision thermal imaging technology to target verify the prediction. The method solves the technical problem in the prior art that it is difficult to quickly locate the leak point after gas pressure testing without structural modification, so that efficient positioning of the leak point can be achieved without any physical modification of the pipeline. Through the cooperative work of acoustic coarse positioning guiding targeted verification, the search space of the leak point can be reduced from the entire pipe network to one or more discrete candidate areas, and all integrity tests can be completed using dry gas as the medium before the newly-built water system is put into use, so as to fundamentally avoid the biofilm breeding problem caused by water on the inner wall of the pipeline, and to ensure the water safety in high-cleanliness-required scenarios such as medical treatment and pharmaceutical production.

[0021] Reference is made toFigure 1 FIG. 1 shows a flowchart of a method for pre-treatment and detection of biofilm formation in a newly built water system according to an embodiment of the present application.

[0022] S100: obtaining an acoustic fingerprint library associated with a digital model of a water system to be tested, wherein the acoustic fingerprint library contains acoustic fingerprint features of a plurality of preset nodes on the water system.

[0023] In one specific implementation, this step is a preparation stage before the start of the field detection work. The purpose is to create a digital twin of the physical pipe network system to be tested with acoustic propagation characteristics, and to pre-calculate the acoustic fingerprint library of all possible leakage conditions.

[0024] Exemplarily, S100 includes: S110: obtaining a geometric design drawing of the water system, and parsing the geometric design drawing to construct the digital model. Refer to Figure 3 which is a schematic diagram of a simplified T-shaped pipe network digital model.

[0025] In one specific embodiment of the present application, the digital model is a three-dimensional graphical data structure containing physical properties, which is a digital twin of the physical pipe network to be tested. The core of the model is composed of nodes (Nodes) representing the connection points of the pipe network and edges (Edges) representing the pipe sections, which describe the geometric shape and topological relationship of the pipe network.

[0026] Each edge object, i.e. pipe section, is associated with a set of geometric and physical parameters. The geometric parameters are data describing the physical shape of the pipe, exemplarily including but not limited to: the three-dimensional spatial length of each pipe section, for example, the length of a straight pipe can be 3250 mm; the nominal diameter and outer diameter of the pipe, for example, the outer diameter of a DN50 PVC-U pipe is 63 mm; the wall thickness of the pipe, for example, the wall thickness of the DN50 pipe is 3.0 mm; and the position and type of key structural components in the pipe network, for example, there is a 90-degree elbow at a coordinate point (12.3, 4.5, 3.0) meters, or there is an equal-diameter tee at another coordinate point.

[0027] The physical parameters are material inherent properties that determine the behavior of acoustic waves propagating in the pipe. Exemplarily, for pipe materials, it is necessary to define the longitudinal wave speed, shear wave speed and acoustic attenuation coefficient. For example, for polyvinyl chloride (PVC) material, the longitudinal wave speed can be set to 2395 m / s, and the acoustic attenuation coefficient is about 0.2 decibels per meter at the typical frequency band (such as 50 kHz) of the leakage signal.

[0028] Each node object records its precise three-dimensional spatial coordinates and topological type, such as elbow, tee, cross or pipe end.

[0029] The acquired geometric design drawings can be in a variety of standard formats, such as AutoCAD DWG format files, or IFC format files commonly used in building information modeling (BIM). The parsing process can be completed by a special software module that can automatically identify the pipeline elements (such as straight lines, arcs, polylines) in the drawings and extract their key geometric parameters and topological relationships.

[0030] Exemplarily, in order to construct the above structured digital model from unstructured geometric design drawings (such as AutoCAD DWG format files), the system executes a set of preset parsing algorithms. The process first filters the input drawings by layers, isolating relevant pipeline elements according to preset layer naming rules (for example, filtering out the layer named "P-PIPE-WATER-SUPPLY"). Subsequently, the module traverses all geometric primitives (such as straight lines, arcs, polylines) on the filtered layer to identify elements representing pipe centerlines and extract their geometric parameters, while searching for text annotations in the local neighborhood to associate the nominal diameter of the pipe section. The most critical is the reconstruction of topological relationships. The module organizes the endpoint coordinates of all centerline elements by constructing a spatial index (such as a k-d tree), and identifies all spatially coincident endpoint sets with a preset coincidence tolerance (such as 1 mm). Each such coincident set is defined as a "node", and each centerline element is defined as a "edge" connecting two nodes. Through this process, static drawing information is successfully converted into a computable graph data structure composed of nodes and edges.

[0031] In a specific embodiment, the acquisition of physical parameters can be: for commonly used materials (such as PVC, stainless steel, etc.), their acoustic parameters can be directly obtained by consulting public engineering manuals or material science databases. When the pipe material is unknown or a non-standard composite material, calibration can be done by standard non-destructive testing methods on the pipe samples obtained from the field. For example, a commercial ultrasonic thickness gauge or acoustic velocity measuring instrument can be used to determine the longitudinal wave velocity and attenuation coefficient of the material according to its standard operating procedures, In another alternative embodiment, it should be appreciated that the process of constructing a digital model from geometric design drawings can also be done manually or semi-automatically by a skilled person in the art, instead of the aforementioned fully automated parsing algorithm. For example, a dedicated data entry interface can be provided, where a skilled person can refer to the two-dimensional or three-dimensional geometric design drawings, manually measure or read the geometric parameters such as length, diameter of each pipe segment from the attributes of the drawings, and input them into the interface. For the topological relationship, the skilled person can define the unique identifiers of the start node and end node for each pipe segment, and input the three-dimensional coordinates and node type (e.g. elbow, tee) for each node identifier in a separate node list. In this way, a graph data structure consisting of nodes and edges that is logically equivalent to the automated parsing result can also be constructed, so as not to affect the subsequent calculation of the acoustic fingerprint library.

[0032] S120: based on the digital model, respectively calculating and generating the acoustic fingerprint features for the plurality of preset nodes by a preset acoustic propagation analysis method. Refer to Figure 4 which shows the process of generating acoustic fingerprints for individual preset nodes and finally constructing a complete fingerprint library.

[0033] After the digital model is constructed, the system discretizes the entire pipe network model into a series of preset nodes at a certain spatial resolution (e.g. every 0.2 meters), simulating a standardized small gas leak event occurring at each node.

[0034] In a preferred embodiment, the choice of spatial resolution follows the Nyquist spatial sampling theorem to ensure sufficient sampling of the spatial variation of the sound field. Specifically, the length of the resolution should not be worse than one quarter of the corresponding wavelength of the highest frequency sound wave (e.g. 140 kHz) in the pipe medium (e.g. air) to accurately capture the phase change of the sound wave during propagation. Then, the system simulates a standardized small gas leak event occurring at each node.

[0035] In a preferred embodiment, the spatial resolution is chosen to follow the Nyquist sampling theorem to ensure sufficient sampling of the spatial variation of the sound field. Specifically, the length of the resolution should not be worse than one quarter of the corresponding wavelength of the highest frequency sound wave to be detected (e.g. 140 kHz) in the medium of the pipe (e.g. air) to accurately capture the phase variation of the sound wave during propagation. Then, the system simulates a standardized small gas leak event occurring at each node. The acoustic propagation analysis method can employ Acoustic Finite Element Analysis (Acoustic FEA), Boundary Element Method (BEM) or the more efficient Transfer Matrix Method for one-dimensional pipe networks. In practical applications, the choice of method can be adapted according to the complexity of the pipe network and the analysis frequency band: for low frequency band analysis with pipe diameters much smaller than the sound wave wavelength or for straight pipe networks with relatively simple geometry, the most computationally efficient Transfer Matrix Method can be preferred; for scenarios with complex three-dimensional geometry (such as valve cavities, reducers) and the need for accurate analysis of high-frequency sound scattering effects, Acoustic Finite Element Analysis or Boundary Element Method capable of full three-dimensional solution should be chosen to ensure the physical fidelity of the model.

[0036] In a specific embodiment, the acoustic propagation analysis method can be implemented with the help of commercial multi-physics simulation software (e.g. COMSOL Multiphysics, ANSYS, etc.) or dedicated acoustic analysis toolboxes. The specific technical path to implement this analysis includes the following steps: first, import or construct the three-dimensional digital model in the simulation software; second, assign the material physical parameters defined in S110 to the pipe wall of the model, for example, assign the longitudinal wave speed of 2395 m / s and the corresponding density and Poisson's ratio to the PVC material. Next, select a suitable physical field interface, such as the "Pressure Acoustics, Frequency Domain" interface, to solve the Helmholtz equation to describe the steady-state propagation of sound waves in the pipe medium (gas). Then, precise boundary conditions and excitation sources need to be set. At a pre-set node, a "monopole point source" is set to simulate gas leakage, which is set to have a flat broadband spectrum characteristic in the frequency range of 10 kHz to 140 kHz, and its acoustic power can be standardized to 1 watt. It should be understood that the simplification of the leakage source as a monopole point source is based on the far-field approximation theory in physical acoustics. For the spherical divergent sound wave produced by the leakage of high-pressure gas through a small aperture, its sound field characteristics are highly consistent with an ideal point source after its propagation distance is greater than several times the wavelength. Therefore, this model is an efficient and reasonable computational simplification of complex leakage phenomena while ensuring the main physical propagation characteristics (such as spherical wave attenuation, phase relationship).

[0037] For the inner wall of the pipe, it can be set as an "acoustic hard boundary" condition to simulate the high impedance reflection of sound waves at the solid-gas interface. At the monitoring points corresponding to the deployment positions of the acoustic sensing devices in S200 below, a virtual sound pressure pickup probe is set or a domain point detection is performed. Finally, the solver is configured to perform frequency domain scanning calculation. The solver is set to scan and solve in the frequency range of 10 kHz to 140 kHz with a frequency step of, for example, 100 Hz. Each time the solver is solved, the simulation software calculates the complex sound pressure value at the monitoring point position generated by the monopole point sound source at a specific frequency. By combining the calculation results of all frequency points, the complete frequency response function (FRF) of the monitoring point to the leakage source node can be obtained. This FRF in the form of a complex vector, after subsequent feature extraction process same as S300 (e.g., calculating its amplitude spectrum and extracting energy distribution features), finally forms the standardized acoustic fingerprint feature corresponding to the preset node. By automatic script or manual repeated control of the simulation software, all preset nodes are traversed as sound source positions, and the above calculation process is repeated to build a complete acoustic fingerprint library.

[0038] The core of this analysis method is to solve the wave equation of sound waves propagating in a complex pipe network structure. The calculation of the sound waves generated by each simulated leakage source (i.e., each preset node) propagates through the pipe wall, reflects and transmits at elbows and tees, and attenuates along the way, and finally reaches the preset monitoring point (i.e., the subsequent deployment position of the acoustic sensing device). The complete time-domain waveform or frequency-domain response of the acoustic signal formed is obtained. Subsequently, the same feature extraction process as in S300 is performed on the simulated received signal, and the result obtained after processing, i.e., a multi-dimensional feature vector, is taken as the acoustic fingerprint feature of the source node, and is stored in the acoustic fingerprint library together with the ID of the node. Exemplarily, an acoustic fingerprint feature can be a 142-dimensional floating-point number vector. The composition of the vector is consistent with the processing result of the measured signal, and its structure can be defined as: 1) 128-dimensional energy spectrum distribution feature, obtained by dividing the frequency band of 10 kHz to 140 kHz into 128 sub-bands and calculating the average energy in each sub-band; 2) 10-dimensional peak frequency feature, obtained by searching for the top 5 frequency points and their corresponding energy values in the entire frequency band; 3) 4-dimensional signal envelope shape feature, obtained by calculating the mean, variance, kurtosis and peak of the signal envelope line. In this way, through a unified multi-dimensional fingerprint, the acoustic characteristics corresponding to the sound source position can be described, and the comparability with the measured feature vector in the same feature space is ensured.

[0039] S200: After the gas in the waterway system is pressurized, an acoustic sensing device deployed on the waterway system acquires a measured acoustic signal.

[0040] During construction, the entire water system to be tested needs to be pressurized with gas first. This is the standard process in the industry, usually using compressed air or high-purity nitrogen as the medium, through the main valve to the completely closed pipe network, until the pressure reaches the design requirement of the test pressure. According to the "Building Water Supply and Drainage and Heating Engineering Construction Quality Acceptance Specification GB50242-2002", for metal and composite pipes, the test pressure should be 1.5 times the working pressure, but not less than 0.6 MPa. Exemplarily, the pipe network pressure can be stabilized at 0.6 MPa.

[0041] After the pressure is stabilized, an acoustic sensing device deployed on the water system is physically coupled to the exposed part of the pipe. The deployment location should be consistent with the monitoring point location set in step S100. Optionally, the location can be selected at the main valve, access hole or pump house of the pipe system, which is easy to access and can represent the starting point of the system, so as to maximize the propagation path of the sound wave and enhance the discrimination of the fingerprints of different leakage points.

[0042] In a preferred embodiment, the acoustic sensing device can be a high-sensitivity piezoelectric contact sensor. The frequency response range of the sensor needs to be wide enough to cover the high-frequency and ultrasonic signals generated by the small gas leakage. Exemplarily, a sensor with a frequency response range of 5 kHz to 200 kHz can be selected. In order to ensure that the sensor can receive the weak vibration of the pipe wall, it needs to be fixed tightly with the outer wall of the pipe through a strong magnetic base (suitable for ferromagnetic pipes) or an adjustable tension strap (suitable for non-ferromagnetic pipes such as PVC or stainless steel pipes), to achieve good acoustic coupling. The electrical signal output by the sensor is very weak, so it needs to be amplified by a low-noise preamplifier, for example, amplified by 60 decibels, to improve the signal-to-noise ratio.

[0043] The acquisition process is completed through a high-speed data acquisition card (DAQ), which digitizes the amplified analog signal at a high enough sampling rate. According to the Nyquist theorem, to effectively capture a 200 kHz signal, the sampling rate should be at least 400 kS / s (kilosamples per second). Exemplarily, a sampling rate of 1 MS / s can be used to collect data with a resolution of 16 bits, ensuring signal fidelity. The acquisition process lasts for a preset time window, for example, 30 seconds to 120 seconds, to form a stable measured acoustic signal time series data.

[0044] S300: Processing the measured acoustic signal to extract a measured acoustic feature vector.

[0045] The purpose of this step is to extract a standardized measured acoustic feature vector from the raw, noisy measured signal, which can be effectively compared with the features in the acoustic fingerprint library. Exemplarily, the S300 comprises: S310: Band-pass filtering the measured acoustic signal. The raw collected signal inevitably contains various environmental noises. For example, the frequency components below 1 kHz usually come from the structural vibration of the building, the operation of nearby large equipment, or the movement of personnel; while the components of too high frequency (e.g. close to half of the sampling rate) may come from the thermal noise of electronic equipment. In order to eliminate these disturbances, a digital band-pass filter needs to be applied to the signal. Exemplarily, an 8th order Butterworth band-pass filter can be designed, with the passband range set to 10 kHz to 140 kHz (the frequency band where the acoustic signal of gas leakage energy is most concentrated).

[0046] S320: Time-frequency analysis of the filtered measured acoustic signal to extract the measured acoustic feature vector. The extracted feature vector must be consistent with the feature extraction method used when constructing the fingerprint library in S100, to ensure the effectiveness of the comparison. The purpose of the time-frequency analysis is to convert the one-dimensional time-domain signal into a two-dimensional or multi-dimensional representation that can reflect the frequency components and energy changes over time.

[0047] Exemplarily, the short-time Fourier transform (STFT) can be used to obtain the spectrogram of the signal. The filtered signal is divided into a series of overlapping time windows (e.g. window length of 1024 sampling points, overlap rate of 50%), and after applying the Hanning window function to each window, the fast Fourier transform (FFT) is performed. In this way, a series of power spectra can be obtained, which are combined to form a distribution map of the signal's energy in time and frequency.

[0048] Exemplarily, the time-frequency analysis can be implemented by using a short-time Fourier transform (STFT). The filtered signal is first segmented into a series of time windows that overlap with each other. For example, in one specific implementation, the length of the time window can be set to 1024 sampling points, and the overlap rate between the window functions can be set to 50% to ensure a balance between time and frequency resolution. For each signal segment within a time window, a window function (e.g., a Hanning window) is applied to suppress spectral leakage, and then a fast Fourier transform (FFT) is performed to calculate the complex spectrum of the time window. The power spectrum corresponding to the time window can be obtained by taking the modulus square of each frequency component of the complex spectrum. By performing sliding window processing on the entire signal sequence, a two-dimensional matrix composed of a plurality of power spectra arranged in time sequence can be obtained, which is the spectrogram of the signal, representing the two-dimensional distribution of signal power with respect to time and frequency. However, since the leakage signal usually persists, in order to obtain a time-frequency distribution that can stably represent the intrinsic characteristics of the signal, time averaging processing also needs to be performed on the spectrogram. Specifically, the system performs element-wise arithmetic averaging of the power spectra of all time windows (i.e., all column vectors on the time axis) in the spectrogram. For example, for a frequency point with a frequency of , the average power is calculated as follows: , where is the total number of time windows, is the power value at the th time window , and the frequency of . After time averaging processing, a one-dimensional vector representing the average power spectrum density over the entire signal acquisition period is obtained. This average power spectrum vector, i.e., the distribution of signal energy in frequency, is used as the subsequent feature extraction signal, thereby eliminating the interference caused by instantaneous noise and signal fluctuations.

[0049] Exemplarily, assume that a small sequence containing 16 sampling points is extracted from the filtered signal : . The parameters of the STFT are set as follows: the window length is sampling points, and the overlap is sampling points (i.e., 50% overlap).

[0050] According to the above parameters, the 16-point signal sequence is segmented into 3 overlapping time windows. After applying the Hanning window function, the signal segments of the 3 windows are respectively:

[0051]

[0052]

[0053] For each of the above windowed signal segments Perform an 8-point fast Fourier transform and compute its power spectrum Due to symmetry, only the first frequency bins are of interest. Suppose the computed 3 power spectrum vectors (in arbitrary power units) are as follows:

[0054]

[0055]

[0056] It can be understood that the values here are illustrative, and are intended to show that there is a persistent energy peak at the 3rd frequency bin .

[0057] Take the element-wise (frequency bin-wise) arithmetic mean of the above 3 power spectrum vectors to obtain the final averaged power spectrum vector :

[0058] The specific calculation is as follows:

[0059]

[0060]

[0061]

[0062]

[0063] The final averaged power spectrum vector is . This vector is the frequency-wise distribution of the stable average energy of the signal segment, and will be used in the subsequent feature extraction step, e.g., to extract energy spectrum distribution features and peak frequency features from it.

[0064] Based on this time-frequency distribution, a variety of features can be extracted. The measured acoustic feature vector contains at least one of the information in the peak frequency, energy spectrum distribution, and signal envelope shape. In one specific implementation, the vector is a combination of multiple modal features, e.g.: Energy spectrum distribution: The frequency band from 10 kHz to 140 kHz is divided into 128 sub-bands, and the average energy in each sub-band is calculated to form a 128-dimensional vector. This feature can most directly reflect the frequency-selective attenuation and resonance enhancement effects caused by the geometric structure (such as bends, branches) and material properties of the pipe network during the propagation of sound waves.

[0065] Peak frequency: Search for the N highest frequency points and their corresponding energy values in the entire frequency band, for example, search for the top 5 frequency points to form a 10-dimensional vector (5 frequencies + 5 energies). This feature is to capture the specific main frequency components generated by the turbulent vortex shedding when gas leaks, as well as the most prominent energy peaks caused by the local structure resonance of the pipeline.

[0066] Signal envelope shape: Perform Hilbert transform on the original signal to obtain its envelope line, and then calculate the statistical properties of the envelope line, such as mean, variance, kurtosis, and so on, to form a 4-dimensional vector, for example. This feature is to quantify the transient characteristics and random fluctuations of the signal, which is closely related to the non-stationary physical nature of turbulent noise generated by high-pressure gas leaks, and can provide complementary recognition information about the time structure of the signal to the frequency spectrum feature.

[0067] By concatenating the above features, a 128+10+4=142-dimensional measured acoustic feature vector can be obtained. This vector is a highly condensed mathematical description of the on-site leakage situation.

[0068] S400: Match the measured acoustic feature vector with the acoustic fingerprint features in the acoustic fingerprint library to generate a leakage probability distribution representing the leakage probability of each region in the waterway system.

[0069] This step is to find the most matching point in the acoustic fingerprint library with the measured feature vector. Exemplarily, the S400 includes: S410: Calculate a similarity score between the measured acoustic feature vector and each of the acoustic fingerprint features in the acoustic fingerprint library. The system traverses each entry (i.e., the acoustic fingerprint feature of each preset node) in the acoustic fingerprint library. For each fingerprint feature vector , calculate the similarity between it and the measured acoustic feature vector .

[0070] The calculation of similarity can use various measurement methods. In a preferred embodiment, the cosine similarity is used, which measures the directional consistency of two vectors by calculating the cosine value of the angle between them in a multi-dimensional space, with a value range of [-1, 1]. The calculation formula is: The cosine similarity has the advantage that it is not sensitive to the absolute size of the vector (i.e. the overall energy intensity of the signal), but rather focuses on the "shape" of the vector (i.e. the distribution pattern of the energy over different features). This makes the method more robust to different sizes of leak apertures. After this step, a list of similarity scores corresponding to the nodes of the pipe network is obtained .

[0071] S420: Based on the similarity scores, convert the similarity scores to the leak probability distribution by a normalization function. The similarity score is only a relative value, in order to make it more physically meaningful and interpretable, it needs to be converted to a probability. The normalization function, in a preferred embodiment, is a Softmax function. The Softmax function can convert an arbitrary real number vector to a probability distribution vector with a sum of 1, and can highlight the contribution of the maximum value in the vector. Its calculation formula is: .

[0072] where, is the posterior probability of the leak point being located at the th node. is an adjustable temperature coefficient used to control the concentration of the probability distribution. If is large, the probability will be highly concentrated in a few nodes with the highest similarity scores, making the positioning result more "sharp"; if is small, the probability distribution will be smoother. Exemplarily, can be set to 20.

[0073] After this step, the final output is a probability vector , which constitutes the leak probability distribution. For example, the calculation result can be , , , and the probabilities of all other nodes are close to 0, i.e. the leak point has an 85% probability of being located near the 125th node.

[0074] S500: According to the leak probability distribution, determine one or more high-probability leak candidate zones to guide targeted verification of the one or more high-probability leak candidate zones, thereby determining the precise location of a leak point.

[0075] The system determines one or more high-probability leak candidate zones according to the leak probability distribution generated by S400. This can be achieved by setting a probability threshold. Exemplarily, all nodes with a probability greater than 0.05 and the pipe sections within a certain range (e.g. 0.5 meters in front and behind) around them can be defined as high-probability leak candidate zones.

[0076] Then, the system highlights these candidate zones on the 3D digital twin model through a human-machine interface (e.g., a laptop screen). For example, the zone with the highest probability is highlighted in red, and the second-highest in orange. Meanwhile, the specific physical location information of these zones within the building is provided (e.g., "2.5 meters away from the north wall in the suspended ceiling of the east corridor on the third floor").

[0077] The step of guiding the targeted verification of the one or more high-probability leakage candidate zones includes guiding an operator to use a thermal imaging device for scanning. Instead of blindly searching the entire pipe network, the operator directly carries a high-resolution thermal imager to the few candidate zones indicated by the system.

[0078] The thermal imaging device needs to have a high enough thermal sensitivity (NETD, Noise Equivalent Temperature Difference), such as better than 50 millikelvin (mK), to detect the subtle temperature difference caused by a tiny leakage. The operator scans the candidate zone's pipeline, especially the weak links such as flanges, valves, welded seams, and joints.

[0079] The physical principle of the targeted verification is based on the gas throttling effect, i.e., the Joule-Thomson effect. When high-pressure gas (such as 0.6 MPa compressed air in the pipe) is released through a tiny leakage hole into a low-pressure environment (atmospheric pressure), the gas will undergo rapid adiabatic expansion, do work on the outside, and cause its internal energy to drop significantly, resulting in a sharp temperature drop. This cooling effect will form a stable and clear low-temperature anomaly point outside the leakage hole.

[0080] Exemplarily, in the case of a background environment temperature of 25°C, a tiny leakage point may appear as a deep blue or black spot of -2°C to -5°C in the thermal imager's view, forming a strong visual contrast with the surrounding pipeline. Once the operator observes such a clear low-temperature anomaly point on the thermal imaging map, it can be confirmed that this is the precise physical location of the leakage point. At this point, the entire detection and positioning process is closed-loop completed.

[0081] Exemplarily, assume that the water pipeline system to be tested is a simplified T-shaped pipe network, and its digital model has been constructed in the S100 stage, generating an acoustic fingerprint library containing 15 preset nodes (Node-1 to Node-15). The acoustic sensing device is deployed at the inlet of the main pipeline, i.e., Node-1. Meanwhile, assume that there is a tiny, inaudible gas leakage at a welded joint near Node-14 at the end of the branch pipeline.

[0082] The T-junction was pressurized to 0.6 MPa. Subsequently, the acquisition module deployed at Node-1 started working. The data acquisition card continuously acquired the pipe wall vibration signal for 60 seconds at a sampling rate of 1 MS / s and a resolution of 16 bits, generating a time series digital signal containing 60,000,000 sampling points .

[0083] For , an 8th order Butterworth digital filter with a passband of 10 kHz to 140 kHz was applied to obtain the filtered signal . For , a short-time Fourier transform (STFT) was performed with a window length of 1024 sampling points and an overlap rate of 50%. By time-averaging the power spectrum of all time windows, a stable average power spectrum vector was obtained. Based on this vector and other analyses, the system extracted the following 142-dimensional measured acoustic feature vector : Energy spectrum distribution features (128 dimensions): It was found that the energy was mainly concentrated in the frequency bands of 35-45 kHz and 80-90 kHz, corresponding to specific dimension values in the vector being higher, while other dimension values being lower.

[0084] Peak frequency features (10 dimensions): The top 5 frequency peaks were detected as: 41.2 kHz, 88.7 kHz, 40.5 kHz, 89.1 kHz, 110.3 kHz, along with their corresponding power values, constituting this 10-dimensional vector.

[0085] Envelope morphology features (4 dimensions): The calculated kurtosis value was 3.8, indicating that the signal had stronger impact characteristics than Gaussian noise, which was consistent with the physical phenomenon of gas turbulent leakage.

[0086] The analysis module performed cosine similarity calculations between and the 15 acoustic fingerprint feature vectors stored in the acoustic fingerprint library to . Since the real leakage point was located at Node-14, its generated sound wave propagation path was the longest, having undergone reflection and attenuation at the T-junction, forming a unique propagation characteristic. Therefore, the calculation results showed that was the most similar to . The similarity scores obtained are shown as follows: (matched with Node-14 model) = 0.96 (matched with Node-13 model) = 0.87 = 0.85 (matches Node-15 model) = 0.15 (matches monitor point itself model) Similarity scores for all other nodes are all less than 0.3.

[0087] Using a temperature coefficient of the Softmax function, the above list of similarity scores is converted into a probability distribution. Due to the amplification effect of the exponential function, small differences in similarity scores are significantly stretched apart:

[0088]

[0089]

[0090] Probabilities for all other nodes are all much less than 0.01.

[0091] Nodes with probabilities greater than 0.05 (i.e. Node-14 and Node-13) and their adjacent pipe segments are defined as high-probability leak candidates. On the three-dimensional user interface, the area at the end of the branch pipe where Node-14 is located is rendered in deep red, indicating the highest probability. The area where Node-13 is located is rendered in orange as a secondary investigation area. At the same time, the physical coordinates of Node-14 within the building and its associated pipeline number are displayed on the interface.

[0092] The operator holds a thermal imager with a thermal sensitivity of 40 millikelvin (mK) and directly goes to the equipment room where the end of the branch pipe is located. The operator scans the pipe segment marked in red and finds a clear low-temperature anomaly area with an area of about 1 square centimeter on a welding seam of the pipe segment in the thermal imaging image, with a center temperature of -4.2 degrees Celsius (℃) and a background temperature of 23.5℃ around the pipe. Through this low-temperature anomaly point, the physical location of the leak point is finally confirmed to be the welding seam near Node-14.

[0093] Referring to Figure 2 , a structure block diagram of a pretreatment and detection system for preventing biofilm generation in a newly built water system is shown. The system can implement all steps in the above method, including a modeling module 210, a collection module 220, a processing module 230, an analysis module 240, and a guidance module 250.

[0094] The modeling module 210 is configured to perform the function of S100, i.e. to obtain the acoustic fingerprint library associated with the digital model of the waterway system to be tested. This module can be a software package running on a backend server or a local computer, integrating a CAD file parser, a physical parameter database and an acoustic simulation engine.

[0095] The acquisition module 220 is configured to perform the function of S200, i.e. to collect the measured acoustic signals in the field. This module physically corresponds to the hardware devices deployed in the field, including the piezoelectric contact sensor, the low-noise preamplifier, the data acquisition card and the device for fixation and coupling.

[0096] The processing module 230 is configured to perform the function of S300, i.e. to process the raw signals and extract the feature vectors. This module is a pure software module, containing a library of digital signal processing algorithms, responsible for tasks such as filtering, time-frequency transformation and feature calculation.

[0097] The analysis module 240 is configured to perform the function of S400, i.e. to match the measured features with the fingerprint library and generate the probability distribution. This module is the carrier of the core algorithm, containing the implementation of similarity calculation functions (such as cosine similarity) and normalization functions (such as Softmax).

[0098] The guidance module 250 is configured to perform the function of S500, i.e. to convert the calculation results into instructions for the operator. This module is usually represented as a graphical user interface (GUI), capable of rendering three-dimensional models, superimposing probability heat maps and giving explicit inspection location instructions in text or graphical form.

[0099] In a specific embodiment, the processing module 230, the analysis module 240 and the guidance module 250 can be integrated on the same portable industrial computer or high-performance notebook computer, constituting the central processing unit of the system. The computer is connected to the data acquisition card of the acquisition module 220 through a USB or Ethernet interface, forming a complete and portable detection system.

[0100] Those skilled in the art will understand that the embodiments of the present application can be provided as methods, systems or computer program products. Therefore, the present application can take the form of entirely hardware embodiments, entirely software embodiments or embodiments combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0101] In addition, each of the function units in each of the embodiments of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware.

[0102] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any change or replacement within the technical scope disclosed in the present application should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A pretreatment and detection method for preventing biofilm formation in newly constructed water systems, characterized in that, include: Acquire an acoustic fingerprint database associated with a digital model of a waterway system under test, wherein the acoustic fingerprint database contains acoustic fingerprint features of multiple preset nodes on the waterway system; After pressurizing the water system with gas, a measured acoustic signal is collected by an acoustic sensing device deployed on the water system. The measured acoustic signal is processed to extract a measured acoustic feature vector; The measured acoustic feature vector is matched with the acoustic fingerprint features in the acoustic fingerprint database to generate a leakage probability distribution that characterizes the leakage probability of each area on the water system. And based on the leakage probability distribution, one or more high-probability leakage candidate areas are identified to guide targeted verification of the one or more high-probability leakage candidate areas, thereby determining the precise location of a leakage point.

2. The method according to claim 1, characterized in that, The step of obtaining an acoustic fingerprint database associated with a digital model of a waterway system under test includes: Obtain the geometric design drawings of the waterway system and analyze the geometric design drawings to construct the digital model. The digital model includes the topology, geometric parameters and physical parameters of the waterway system. Based on the digital model, acoustic fingerprint features are calculated and generated for the multiple preset nodes using a preset acoustic propagation analysis method.

3. The method according to claim 2, characterized in that, The geometric parameters include pipe length, diameter, wall thickness, and the location and type of elbows and tees; the physical parameters include the sound velocity and sound attenuation coefficient of the pipe material.

4. The method according to claim 1, characterized in that, The steps for processing the measured acoustic signal include: The measured acoustic signal is bandpass filtered to remove noise outside the preset frequency band; The filtered measured acoustic signal is subjected to time-frequency analysis to extract the measured acoustic feature vector, which includes at least one of the following: peak frequency, energy spectrum distribution, and signal envelope morphology.

5. The method according to claim 1, characterized in that, The step of matching the measured acoustic feature vector with the acoustic fingerprint features in the acoustic fingerprint database includes: Calculate a similarity score between the measured acoustic feature vector and each acoustic fingerprint feature in the acoustic fingerprint database; And based on the similarity score, the similarity score is converted into the leakage probability distribution through a normalization function.

6. The method according to claim 5, characterized in that, The normalization function is a Softmax function.

7. The method according to claim 1, characterized in that, The step of guiding targeted verification of the one or more high-probability leakage candidate regions includes: Instruct an operator to use a thermal imaging device to scan the one or more high-probability leak candidate areas; And based on a low-temperature anomaly detected by the thermal imaging device due to a gas throttling effect, the precise location of the leak point is determined.

8. A pretreatment and detection system for preventing biofilm formation in newly constructed water systems, characterized in that, include: A modeling module is used to acquire an acoustic fingerprint database associated with a digital model of a waterway system under test, wherein the acoustic fingerprint database contains acoustic fingerprint features of multiple preset nodes on the waterway system. A data acquisition module is used to acquire a measured acoustic signal through an acoustic sensor deployed on the water system after the water system has been pressurized with gas. A processing module is used to process the measured acoustic signal to extract a measured acoustic feature vector; An analysis module is used to match the measured acoustic feature vector with the acoustic fingerprint features in the acoustic fingerprint database to generate a leakage probability distribution characterizing the leakage probability of each area on the water system. And a guidance module, used to determine one or more high-probability leak candidate areas based on the leak probability distribution, so as to guide the targeted verification of the one or more high-probability leak candidate areas, thereby determining the precise location of a leak point.

9. The system according to claim 8, characterized in that, The acoustic sensing device of the acquisition module includes a piezoelectric contact sensor and a low-noise preamplifier.

10. The system according to claim 8, characterized in that, The guidance module is further used for: On a three-dimensional display interface, the high-probability leak candidate area is highlighted on the digital model of the water system; And guide an operator to use a thermal imaging device to perform the target verification based on the highlighted display.

Images