Building curtain wall detection method

Abstract

The invention relates to the technical field of building curtain wall detection, in particular to a building curtain wall detection method, which comprises the following steps that a mobile detection platform is deployed, and a sensor is carried on the mobile detection platform; acquiring an acoustic signal and / or a tracer gas concentration signal generated by air permeation through a sensor; processing the acoustic signal and / or the tracer gas concentration signal to identify an air permeation point and determine the position of the air permeation point; and carrying out quantitative evaluation on the air permeation points, and generating a maintenance report. Through the technical scheme, the problem of failure of a detection method based on temperature difference under the compensation of an automatic heating, ventilation and air conditioning system is solved, and an accurate and efficient solution is provided for energy-saving maintenance of a building curtain wall.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of building curtain wall detection, and particularly relates to a building curtain wall detection method. BACKGROUND

[0002] As an important part of modern buildings, the energy-saving performance of building curtain walls has a direct impact on the overall energy consumption of buildings. The sealing strip of the curtain wall will produce a small gap due to aging in long-term use, leading to air permeation, and further affecting the air tightness and energy-saving effect of the building. The traditional detection method cannot effectively identify and quantify these scattered and distributed small leakage points, and the automatic heating, ventilation and air conditioning system inside the building will compensate for the local temperature changes caused by air permeation in real time, making the detection method based on temperature difference ineffective. SUMMARY

[0003] The present application aims to overcome the above-mentioned deficiencies, and provides a building curtain wall detection method.

[0004] The present application adopts the following technical solutions: A building curtain wall detection method, comprising the following steps: Deploying a mobile detection platform, the mobile detection platform being equipped with sensors; Acquiring acoustic signals and / or tracer gas concentration signals generated by air permeation through the sensors; Processing the acoustic signals and / or tracer gas concentration signals to identify air permeation points and determine the positions of the air permeation points; Quantitatively evaluating the air permeation points and generating a repair report.

[0005] Through the technical solutions, the present application can effectively identify and locate the small air permeation points in the building curtain wall, and quantitatively evaluate them to generate a repair report, thereby overcoming the limitations of traditional detection methods in the face of distributed micro-leakage, solving the problem of the ineffectiveness of detection methods based on temperature differences under the compensation of automatic heating, ventilation and air conditioning systems, and providing a precise and efficient solution for the energy-saving maintenance of building curtain walls.

[0006] Further, the method further comprises the following steps: Deploying a frequency-sweeping pneumatic device, the frequency-sweeping pneumatic device generating differential pressure pulses with frequency modulation characteristics; Acquiring acoustic signals and / or tracer gas concentration signals generated by air permeation through the sensors under the action of the differential pressure pulses; Performing time-frequency spectrum analysis on the acoustic signals and / or tracer gas concentration signals to extract response spectrum features related to the frequency of the differential pressure pulses; According to a pre-set response spectrum feature library, identifying the types of the extracted response spectrum features and attributing them to the permeation sources; Determining the location of the air infiltration source.

[0007] Further, the step of quantitatively evaluating the air infiltration point and generating a maintenance report comprises: According to the type of air infiltration source and the response spectrum intensity, the air leakage amount of the air infiltration point under the standard pressure difference condition is estimated; Obtaining indoor and outdoor pressure difference data, adaptive ventilation module operating state, local wind speed, ambient temperature and humidity; Based on the indoor and outdoor pressure difference data, adaptive ventilation module operating state, local wind speed, ambient temperature and humidity, the estimated air leakage amount is dynamically corrected, and the long-term energy consumption contribution of the air infiltration point under the current and predicted environmental conditions is calculated; According to the type of air infiltration source, material aging rate data and historical maintenance records, the leakage amount change trend of the air infiltration point in the future period of time is predicted; Comprehensive dynamic correction of air leakage, long-term energy consumption contribution prediction, air infiltration source type, air infiltration source location and predicted leakage amount change trend, the maintenance priority of the air infiltration point is calculated, and a report containing adaptive maintenance strategy is generated.

[0008] Further, the step of performing time-frequency spectrum analysis on the acoustic signal and / or tracer gas concentration signal to extract the response spectrum features related to the frequency of the pressure difference pulse comprises: Multi-channel synchronous acquisition of acoustic signals and / or tracer gas concentration signals is performed; Using adaptive filtering technology, according to the excitation frequency characteristics of the frequency-sweeping pneumatic device, the response components synchronized with the excitation frequency of the pressure difference pulse are separated from the acoustic signals and / or tracer gas concentration signals, and the background noise and non-infiltration airflow disturbance signals are suppressed; Multi-resolution time-frequency decomposition is performed on the separated response components to obtain time-frequency spectrum under different frequency resolutions; In the time-frequency spectrum under different frequency resolutions, the resonance peaks, energy envelopes and frequency shift patterns generated by different air infiltration sources or non-infiltration airflow disturbances in the inner and outer layers of the double-layer curtain and the air cavity under the specific frequency-sweeping frequency are identified and extracted as response spectrum features; The response spectrum features are processed for feature enhancement.

[0009] Further, according to the pre-set response spectrum feature library, the step of identifying the type of the extracted response spectrum features and attributing the air infiltration source comprises: Preliminary similarity calculation is performed on the extracted response spectrum features to obtain preliminary identification results; When the preliminary identification result shows that the similarity between the extracted response spectrum feature and each of the plurality of feature templates in the preset response spectrum feature library is higher than the preset threshold, fine structures, transient change rates and dynamic trajectories of the corresponding response spectrum features are extracted to obtain a fine feature comparison result; In combination with the position of the penetration source and the operating state of the adaptive ventilation module, it is analyzed whether the response spectrum feature is affected by airflow disturbance in the air cavity; According to the fine feature comparison result and the analysis result of the position of the penetration source and the operating state of the adaptive ventilation module, the preliminary identification result is corrected to determine the type of the penetration source and the attribution.

[0010] Further, the step of performing feature enhancement processing on the response spectrum feature includes: Performing nonlinear enhancement processing on the response spectrum feature based on an adaptive threshold; Performing enhancement processing on the response spectrum feature after nonlinear enhancement processing based on multi-scale detail fusion; Performing sparse representation and reconstruction on the response spectrum feature after multi-scale detail fusion enhancement.

[0011] Further, the step of performing nonlinear enhancement processing on the response spectrum feature based on an adaptive threshold includes: Performing adaptive gamma correction on the response spectrum feature based on local signal-to-noise ratio, and the gamma correction factor is dynamically adjusted according to the local signal-to-noise ratio of the response spectrum feature, providing stronger enhancement in the low signal-to-noise ratio region, and maintaining the original form of the response spectrum feature in the high signal-to-noise ratio region; During the adjustment of the gamma correction factor, the instantaneous energy change rate of the response spectrum feature is monitored; When the instantaneous energy change rate exceeds the preset change rate threshold, a noise suppression strategy based on frequency segmentation is enabled, additional attenuation is performed on the high-energy disturbance frequency band, and the gamma correction factor of the adjacent frequency band is smoothed; In combination with the speed and attitude information of the mobile detection platform, the spatial position of the response spectrum feature is calibrated to compensate for the estimation deviation of the local signal-to-noise ratio caused by the movement of the mobile detection platform.

[0012] Further, the step of performing enhancement processing on the response spectrum feature after nonlinear enhancement processing based on multi-scale detail fusion includes: Performing multi-scale detail fusion based on feature similarity clustering on the response spectrum feature after nonlinear enhancement processing; The clustering groups similar resonance peaks, energy envelopes and frequency shift patterns; Each group of resonance peaks, energy envelopes and frequency shift patterns is weighted and combined to enhance the distinguishability of the resonance peaks, energy envelopes and frequency shift patterns.

[0013] Further, the step of performing sparse representation and reconstruction on the response spectrum features enhanced by multi-scale detail fusion includes: For the air cavity inside the double-layer curtain wall, the response spectrum features enhanced by multi-scale detail fusion are performed multi-dictionary collaborative sparse representation; The multi-dictionary collaborative sparse representation includes a general dictionary and one or more scene-specific sub-dictionaries, the general dictionary is used to capture the common features of the penetration source, and the scene-specific sub-dictionaries are constructed for penetration sources of different depths and different structural gap types in the air cavity to capture the fine features of the penetration source; The joint sparse decomposition of the response spectrum features enhanced by multi-scale detail fusion on the general dictionary and the scene-specific sub-dictionary is performed to obtain joint sparse coefficients; According to the similarity of the preliminary position information of the penetration source and the response spectrum features, the joint sparse coefficients are weighted and fused to dynamically adjust the contribution weight of the general dictionary and the scene-specific sub-dictionary, so as to improve the environmental adaptability of detection; The joint sparse coefficients after weighted fusion, the general dictionary and the scene-specific sub-dictionary are reconstructed to obtain enhanced response spectrum features.

[0014] Further, the step of obtaining the acoustic signal and / or the tracer gas concentration signal generated by air penetration by the sensor includes: The acoustic signal generated by air penetration is obtained by an acoustic sensor; The tracer gas concentration signal generated by air penetration is obtained by a gas sensor; When the acoustic signal generated by air penetration and the tracer gas concentration signal generated by air penetration need to be obtained, the acoustic signal generated by air penetration and the tracer gas concentration signal generated by air penetration are collected in a synchronous manner to ensure that the time stamps of the acoustic signal generated by air penetration and the tracer gas concentration signal generated by air penetration are consistent.

[0015] The present application quantitatively evaluates the air penetration point and generates a repair report, providing specific repair basis and priority for building operation and maintenance team, realizing accurate positioning and quantitative evaluation of the root cause of curtain wall energy saving performance decline, significantly improving detection efficiency and repair pertinence, reducing detection cost, and ultimately helping to improve the overall energy saving effect of the building.

[0016] In order to further understand the features and technical contents of the present application, please refer to the following detailed description and drawings of the present application. However, the provided drawings are only used for reference and illustration, and are not used to limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The present application is a method flow chart of a building curtain wall detection method. Detailed Implementation

[0018] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can understand the advantages and effects of the present invention from the content disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the spirit of the present invention. Furthermore, the accompanying drawings of the present invention are for simple illustrative purposes only and are not depictions of actual dimensions; this is stated in advance. The following embodiments will further describe the relevant technical content of the present invention in detail, but the disclosed content is not intended to limit the scope of protection of the present invention.

[0019] This embodiment provides a method for inspecting building curtain walls, combined with... Figure 1 As shown.

[0020] A method for inspecting building curtain walls, comprising the following steps: Deploy a mobile detection platform, which is equipped with sensors; Acoustic signals and / or tracer gas concentration signals generated by air infiltration are acquired through sensors; The acoustic signal and / or tracer gas concentration signal are processed to identify the air infiltration point and determine its location; Quantitatively assess air infiltration points and generate maintenance reports.

[0021] The mobile inspection platform mentioned can be understood as a carrier capable of moving along the surface of a building's curtain wall, integrating various inspection devices. This platform can be a drone, a tracked robot, or a manually operated inspection vehicle. Its core function is to provide a stable and controllable mobile carrier so that sensors can systematically cover the entire curtain wall area.

[0022] A sensor is a device used to sense and measure physical quantities and convert them into usable signals. In this application, the sensor is mainly used to acquire acoustic signals and / or tracer gas concentration signals generated by air infiltration. Acoustic signals refer to weak sounds generated by airflow or leakage, which may be outside the range of human hearing and require highly sensitive acoustic sensors (such as ultrasonic sensors or high-frequency microphone arrays) to capture. Tracer gas concentration signals are determined by releasing a small amount of harmless tracer gas on one side of the curtain wall and detecting its concentration change on the other side using a gas sensor (such as an infrared gas analyzer, semiconductor gas sensor, etc.), thereby determining whether gas infiltration has occurred.

[0023] Air infiltration points refer to the tiny gaps or holes in the curtain wall structure where air leakage occurs. Identifying these infiltration points and determining their locations are key steps in the method. Quantitative evaluation refers to the quantitative analysis of the identified air infiltration points, such as estimating their leakage amount, impact on building energy consumption, etc. Finally, the generation of the repair report aims to provide specific repair recommendations and priority ranking for building operation and maintenance.

[0024] The building curtain wall detection method of the present application is implemented by the following steps: First, deploy the mobile detection platform. The mobile detection platform can take various forms, for example, it can be a ground robot equipped with a wheeled or tracked drive mechanism, which can move along the bottom of the curtain wall or accessible areas. In another embodiment, the mobile detection platform can be a multi-rotor drone that can fly outside the curtain wall through remote control, thereby covering high-level areas. In addition, the mobile detection platform can also be a detection vehicle operated by a human, which moves along the curtain wall surface through a track or suspension system. Regardless of the form, the mobile detection platform is equipped with sensors that are fixed on the platform and calibrated and configured appropriately according to the detection requirements. For example, acoustic sensors can be installed at multiple positions on the platform to form an array to improve the accuracy of sound source positioning; gas sensors can be installed at the front end of the platform to collect gas samples in real time during movement.

[0025] Second, acquire acoustic signals and / or tracer gas concentration signals generated by air infiltration through sensors. When the mobile detection platform moves along the curtain wall, the sensors carried by it will work continuously. For example, acoustic sensors will listen to the acoustic environment of the curtain wall area in real time, capturing ultrasonic or infrasonic signals generated by tiny air leaks. These signals usually have specific frequency and energy characteristics and are distinguishable from background noise. At the same time, if tracer gas detection is used, a small amount of tracer gas can be released on one side of the curtain wall (e.g. indoors), while the gas sensor on the mobile detection platform detects the concentration on the other side (e.g. outdoors). When there are air infiltration points, the tracer gas will leak from one side to the other through these infiltration points, causing the gas sensor to detect an increase in tracer gas concentration. The data collected by the sensors will be transmitted to the processing unit in real time for subsequent analysis.

[0026] Next, the acoustic signals and / or tracer gas concentration signals are processed to identify and determine the location of air infiltration points. After obtaining the raw acoustic signals and / or tracer gas concentration signals, a series of data processing is needed. For example, for acoustic signals, digital signal processing techniques such as Fourier transform, wavelet analysis, etc. can be used to extract the characteristic signals related to air infiltration from the complex background noise. By analyzing the intensity, frequency and propagation direction of these characteristic signals, it can be preliminarily judged whether there is air infiltration. Combined with the position information of the mobile detection platform (such as obtained by GPS, inertial measurement unit IMU or visual odometry), the source of the acoustic signal can be triangulated or sound source imaged, so as to accurately determine the location of the air infiltration point. For tracer gas concentration signals, a concentration threshold can be set, and when the detected tracer gas concentration exceeds the threshold, it is considered that there is air infiltration. Combined with the position information of the gas sensor when detecting high concentration, the location of the infiltration point can be directly determined.

[0027] Finally, the air infiltration points are quantitatively evaluated, and a maintenance report is generated. After identifying and locating the air infiltration points, they need to be quantitatively evaluated. For example, by analyzing the intensity of the acoustic signal or the peak value of the tracer gas concentration, combined with the preset leakage model, the air leakage of each infiltration point can be estimated. These leakage amounts can be converted into the impact on building energy consumption, such as calculating the additional heating or cooling load caused by these infiltration points. The results of quantitative evaluation will be the basis for generating a maintenance report. The maintenance report will list in detail the location of all identified air infiltration points, the estimated leakage amount, the impact on energy consumption, and the recommended maintenance solutions and priorities. For example, infiltration points with large leakage amounts or located in critical areas may be marked as high-priority maintenance. The maintenance report can be output in the form of an electronic document or a printed copy, providing decision support for building operation and maintenance teams.

[0028] The above building curtain wall detection method further comprises the following steps: Deploying a swept-frequency pneumatic device that generates differential pressure pulses with frequency modulation characteristics; Under the action of the differential pressure pulses, the acoustic signals and / or tracer gas concentration signals generated by air infiltration are obtained by sensors; Performing time-frequency spectrum analysis on the acoustic signals and / or tracer gas concentration signals to extract response spectrum features related to the frequency of the differential pressure pulses; According to a pre-set response spectrum feature library, the extracted response spectrum features are identified by type and attributed to the source of the infiltration; Determining the location of the source of the infiltration.

[0029] Specifically, deploying the sweep-frequency pneumatic device refers to setting a device capable of actively generating differential pressure changes near the building curtain wall area to be detected. The sweep-frequency pneumatic device is designed to generate differential pressure pulses with frequency modulation characteristics, which means that the frequency of the generated differential pressure pulses is not fixed but varies continuously or jumps within a certain frequency range according to a preset pattern (e.g., linear scanning, logarithmic scanning, etc.). The purpose is to simulate the wind pressure effect at different frequencies by active excitation, thereby stimulating the unique response of different types of penetration sources.

[0030] Wherein, under the action of the differential pressure pulse, the acoustic signal and / or tracer gas concentration signal generated by air penetration are obtained through the sensor, which can be understood as follows: when the sweep-frequency pneumatic device generates a differential pressure pulse acting on the building curtain wall, if there is an air penetration point, air passing through these penetration points will produce specific acoustic signals and / or changes in tracer gas concentration. At this time, the sensors (such as acoustic sensors, gas sensors) carried on the mobile detection platform are used to synchronously collect these response signals caused by active excitation. The purpose is to capture the dynamic response of the penetration source under specific frequency excitation, rather than relying solely on passive signals in the environment.

[0031] In practical applications, time-frequency spectrum analysis is performed on the acoustic signal and / or tracer gas concentration signal to extract response spectrum features related to the frequency of the differential pressure pulse, which specifically refers to joint analysis of the time domain and frequency domain of the collected acoustic signal and / or tracer gas concentration signal. For example, short-time Fourier transform, wavelet transform, etc. can be used to generate a time-frequency spectrum diagram of the signal. On this basis, since the differential pressure pulse has a known frequency modulation characteristic, the signal features corresponding to the excitation frequency variation pattern can be identified and extracted from the time-frequency spectrum diagram, which are referred to as response spectrum features. Response spectrum features can include resonance peaks, energy envelopes, frequency shifts, etc., which can reflect the unique physical response of the penetration source under different excitation frequencies. The purpose is to distinguish the response of the penetration source from background noise and non-penetrating airflow disturbances, and obtain fine features of the penetration source.

[0032] Further, according to the preset response spectrum feature library, the extracted response spectrum features are identified and attributed to the penetration source, which means that the response spectrum features extracted through time-frequency spectrum analysis are compared with a pre-established database. The response spectrum feature library stores typical response spectrum feature patterns of various known penetration source types (e.g., sealant aging, gasket failure, profile deformation, loose connections, etc.) under different sweep-frequency excitations. Through pattern matching or machine learning algorithms, it can be identified which known penetration source type the currently detected response spectrum feature is closest to, thereby realizing type identification and attribution of the penetration source. The purpose is not only to find the penetration point, but also to determine its cause, providing a basis for subsequent maintenance.

[0033] Thus, determining the location of the air infiltration point refers to accurately calculating the specific spatial coordinates of the air infiltration point on the building curtain wall based on the identification of the type and attribution of the air infiltration point, in combination with the layout of the sensor array, the signal propagation characteristics, and the position information of the mobile detection platform. This can be achieved through signal strength attenuation, time difference of arrival, or beamforming techniques. The purpose is to provide accurate geographic location information for maintenance work.

[0034] In some preferred embodiments, the glass curtain wall of a high-rise building is assumed to be detected. The curtain wall may have small cracks caused by sealant aging and structural gaps caused by loose connectors. Traditional passive detection methods may only be able to detect the presence of air infiltration, but it is difficult to distinguish between the two types of air infiltration. In the detection method of the present application, the mobile detection platform carries a sweep-frequency pneumatic device and a sensor. The sweep-frequency pneumatic device is deployed on the outside of the curtain wall to generate a differential pressure pulse with a linear sweep from 20 Hz to 500 Hz. When the differential pressure pulse acts on the curtain wall, the acoustic sensor synchronously collects the acoustic signals on the inside of the curtain wall. By performing time-frequency spectrum analysis on the collected acoustic signals, such as using continuous wavelet transform, it can be observed that in a certain frequency range, sealant cracks will produce a narrow-band resonance peak, while loose connectors may produce a wider-band energy envelope and frequency shift. These unique response spectrum features are extracted and compared with a pre-set feature library. The feature library pre-stores the typical response patterns of sealant aging and connector loosening under the same sweep-frequency excitation. Through comparison, the system can accurately identify which air infiltration points are caused by sealant aging and which are caused by connector loosening, and accurately mark the specific locations of these air infiltration points in combination with the positioning data of the sensor array. For example, the system may report that there is an air infiltration point caused by sealant aging at the upper left corner of the fifth glass on the third floor of the curtain wall, and an air infiltration point caused by connector loosening at the bottom of the second glass on the seventh floor.

[0035] The present application further proposes steps for quantitatively evaluating the air infiltration point and generating a maintenance report, including: According to the type of the air infiltration point and the response spectrum intensity of the air infiltration point, estimate the air leakage amount of the air infiltration point under standard differential pressure conditions; Obtain indoor and outdoor pressure difference data, adaptive ventilation module operating status, local wind speed, environmental temperature and humidity; Based on the indoor and outdoor pressure difference data, the adaptive ventilation module operating status, the local wind speed, the environmental temperature and the humidity, dynamically correct the estimated air leakage amount, and calculate the long-term energy consumption contribution of the air infiltration point under the current and predicted environmental conditions; According to the type of the air infiltration point, the material aging rate data and the historical maintenance records, predict the leakage amount trend of the air infiltration point in a future period of time; The dynamic corrected air leakage, long-term energy consumption contribution prediction, penetration source type, penetration source location, and predicted leakage trend are integrated to calculate the repair priority of the air penetration point, and a report containing adaptive repair strategies is generated.

[0036] Specifically, the steps of quantitatively evaluating the air penetration point and generating a repair report include: First, the air leakage of the air penetration point under standard pressure difference conditions is estimated according to the penetration source type and the penetration source response spectrum intensity. The penetration source type can be a specific type such as a crack, a hole, or a seal failure identified by the above method, and the penetration source response spectrum intensity reflects the severity of the penetration. This step aims to provide a baseline, standardized leakage evaluation.

[0037] Second, indoor and outdoor pressure difference data, adaptive ventilation module operating status, local wind speed, and environmental temperature and humidity are obtained. These data are key environmental parameters that affect actual air leakage and building energy consumption. Indoor and outdoor pressure difference data reflect the driving force for air penetration, adaptive ventilation module operating status affects air flow organization inside the building, and local wind speed, environmental temperature, and humidity directly affect material properties and air density.

[0038] Further, the estimated air leakage is dynamically corrected based on the obtained indoor and outdoor pressure difference data, adaptive ventilation module operating status, local wind speed, environmental temperature, and humidity, and the long-term energy consumption contribution of the air penetration point under current and predicted environmental conditions is calculated. Dynamic correction aims to make the estimated leakage closer to the actual operating conditions, while the calculation of long-term energy consumption contribution quantifies the continuous impact of the penetration point on the overall energy consumption of the building.

[0039] In addition, the leakage trend of the air penetration point in the future period of time is predicted according to the penetration source type, material aging rate data, and historical maintenance records. Material aging rate data can be obtained from technical manuals of material suppliers or industry standards, and historical maintenance records provide actual maintenance experience of the building or similar buildings. These information helps to evaluate the deterioration speed of the penetration point over time.

[0040] Finally, the dynamic corrected air leakage, long-term energy consumption contribution prediction, penetration source type, penetration source location, and predicted leakage trend are integrated to calculate the repair priority of the air penetration point, and a report containing adaptive repair strategies is generated. The repair priority aims to guide maintenance personnel to prioritize the processing of the penetration point with the greatest impact or the fastest deterioration, and the adaptive repair strategy provides customized repair recommendations according to the specific circumstances of the penetration source (such as type, location, and expected life).

[0041] In some preferred embodiments, assuming that a penetration source is identified at a sealant strip between a window frame and glass of a building curtain wall by the above detection method, the penetration source type is identified as sealant aging, and its intensity characteristics are obtained by response spectrum analysis.

[0042] First, according to the sealant aging penetration source type and its response spectrum intensity, the air leakage amount of the penetration point under standard pressure difference conditions is estimated to be X cubic meters / hour.

[0043] Subsequently, the system continuously obtains indoor and outdoor pressure difference data of the building, real-time running state of the adaptive ventilation module, local wind speed, environmental temperature and humidity of the region and other environmental parameters. For example, in winter, when the outdoor temperature is low and the wind speed is large, the indoor and outdoor pressure difference may increase significantly.

[0044] Based on these real-time environmental data, the estimated leakage amount of X cubic meters / hour is dynamically corrected, for example, to X' cubic meters / hour. At the same time, combined with the energy price and running time of the building, the long-term energy consumption contribution of the penetration point in the current and predicted winter running conditions is calculated to be Y yuan per year.

[0045] Further, combined with the penetration source type of sealant aging, the known aging rate data of the sealant strip material (for example, the expected service life of this type of strip in a specific environment is 10 years, and it has been used for 8 years), and the historical maintenance records of the building or similar buildings (for example, similar aging problems have occurred in other locations of the same batch of strips), it is predicted that the leakage amount of the penetration point will increase by Z% per year in the next 1-3 years.

[0046] Finally, combining the dynamically corrected leakage amount X', the long-term energy consumption contribution Y, the penetration source type sealant aging, the penetration source location (for example, the third layer east window of the curtain wall), and the predicted leakage amount change trend, the maintenance priority of the penetration point is calculated to be high, and a maintenance report is generated. The report not only contains all the quantitative data and predictions mentioned above, but also proposes adaptive maintenance strategies, such as recommending to replace the sealant strip within the next 6 months, and recommending to use a new type of sealant material with longer service life and better weather resistance to avoid similar problems from occurring again in the future.

[0047] The present application further proposes the steps of performing time-frequency spectrum analysis on the acoustic signal and / or tracer gas concentration signal, and extracting response spectrum characteristics related to the frequency of the pressure difference pulse include: Multi-channel synchronous acquisition of acoustic signal and / or tracer gas concentration signal; The adaptive filtering technology is used to separate the response component synchronized with the excitation frequency of the differential pressure pulse from the acoustic signal and / or the tracer gas concentration signal according to the excitation frequency characteristics of the sweep-type pneumatic device, and to suppress the background noise and the signal generated by the non-infiltration air flow disturbance; The separated response component is subjected to multi-resolution time-frequency decomposition to obtain time-frequency spectrograms under different frequency resolutions. In the time-frequency spectrograms under different frequency resolutions, the resonance peaks, energy envelopes and frequency shift patterns generated by different infiltration sources or non-infiltration air flow disturbances in the inner and outer layers of the double-layer curtain and the air cavity are identified and extracted at a specific sweep frequency as response spectrum characteristics. The response spectrum characteristics are subjected to feature enhancement processing.

[0048] Specifically, the multi-channel synchronous acquisition refers to simultaneously acquiring acoustic signals and / or tracer gas concentration signals at different positions or different types by using multiple sensors at the same time point or extremely short time intervals. The purpose is to ensure the time stamp consistency of the acquired signals, thereby providing accurate spatio-temporal correlation information for subsequent signal processing and feature extraction, which helps to improve the signal-to-noise ratio and positioning accuracy.

[0049] The adaptive filtering technology can be understood as a signal processing method that can automatically adjust the filter parameters according to the statistical characteristics of the input signal. In this application, the technology dynamically separates the response component generated by the differential pressure pulse excitation from the original signal according to the known excitation frequency characteristics of the sweep-type pneumatic device. At the same time, the technology can effectively suppress the background noise in the environment and the interference signal generated by the non-infiltration air flow disturbance (for example, normal ventilation air flow inside the curtain structure or external wind pressure fluctuation), thereby ensuring that the extracted response component is more pure and more accurate to reflect the characteristics of air infiltration.

[0050] In practical applications, multi-resolution time-frequency decomposition specifically refers to analyzing the separated response component in different time scales and frequency scales to generate a series of time-frequency spectrograms with different time resolutions and frequency resolutions. For example, wavelet transform, short-time Fourier transform combined with different window functions or multi-scale spectrum analysis can be used. The purpose is to comprehensively capture the transient and steady-state characteristics of the infiltration signal in different frequency ranges, because different types of infiltration sources may exhibit unique characteristics at different frequency resolutions.

[0051] In the time-frequency spectrograms with different frequency resolutions, the resonance peaks, energy envelope, and frequency shift patterns are identified and extracted as response spectral features. Resonance peaks refer to the phenomenon of significant enhancement of signal energy at specific frequencies, which is usually related to the geometric structure or material properties of the penetration path; energy envelope describes the trend of signal energy change over time or frequency, reflecting the dynamic characteristics of the penetration process; frequency shift patterns may indicate slight changes in the penetration path or fluctuations in air flow velocity. These features can finely characterize the unique responses of different penetration sources (such as aging of seals, structural cracks) or non-penetrating air flow disturbances (such as ventilation openings) in the inner and outer layers of the double-layer curtain wall and the air cavity at specific swept frequency.

[0052] Further, the response spectral features are subjected to feature enhancement processing, which aims to further improve the discriminability and robustness of these features, making them more reliable in subsequent type identification and attribution processes. This may include noise reduction, normalization, feature selection, or feature transformation operations.

[0053] In some preferred embodiments, it is assumed that air penetration detection needs to be performed on the double-layer glass curtain wall of a high-rise building. First, a mobile detection platform is deployed, which carries multiple high-sensitivity acoustic sensors and tracer gas concentration sensors, and ensures that these sensors perform synchronous acquisition. At the same time, a swept-frequency pneumatic device is deployed outside or inside the curtain wall, which generates a differential pressure pulse with a linearly varying frequency from 20 Hz to 200 Hz.

[0054] When the differential pressure pulse acts on the curtain wall, the acoustic signals and / or tracer gas concentration signals generated by air penetration are captured by the sensors. These original signals may be mixed with urban traffic noise, building internal air conditioning system operation noise, and air flow disturbances caused by normal ventilation at the curtain wall structural joints. At this time, using adaptive filtering technology, the response components synchronized with the swept-frequency excitation frequency are accurately separated from the collected signals according to the known swept-frequency characteristics of the swept-frequency pneumatic device, and the above background noise and non-penetrating air flow disturbances are effectively suppressed.

[0055] Subsequently, the isolated response components are subjected to multi-resolution time-frequency decomposition, such as continuous wavelet transform, to generate a series of time-frequency spectrograms at different time-frequency resolutions. In these time-frequency spectrograms, different characteristic patterns can be observed: for example, a tiny seal crack can manifest as a sharp resonance peak at higher frequency resolution, while a larger structural gap can manifest as a mode with a broad energy envelope and slight frequency shift at lower frequency resolution. By identifying and extracting these unique resonance peaks, energy envelopes, and frequency shift patterns, they can be used as response spectral features to characterize different types of air infiltration sources. Finally, these extracted response spectral features are subjected to feature enhancement processing, such as nonlinear transformation or sparse representation, to further highlight their uniqueness and provide clearer and more discriminative inputs for subsequent infiltration source type identification and attribution. In this way, even in complex detection environments, different types of air infiltration sources can be accurately identified and distinguished.

[0056] The application further proposes a step of identifying the type of the extracted response spectral feature according to the pre-set response spectral feature library and attributing the air infiltration source, which includes: performing preliminary similarity calculation on the extracted response spectral feature to obtain a preliminary identification result; when the preliminary identification result shows that the similarity of the extracted response spectral feature to multiple feature templates in the pre-set response spectral feature library is higher than the pre-set threshold, extracting the fine structure, transient change rate, and dynamic trajectory of the corresponding response spectral feature and obtaining a fine feature comparison result; combining the location of the air infiltration source and the running state of the adaptive ventilation module to analyze whether the response spectral feature is affected by airflow disturbance in the air cavity; correcting the preliminary identification result according to the fine feature comparison result and the analysis result of the running state of the adaptive ventilation module to determine the type of the air infiltration source and the attribution.

[0057] Among them, the preliminary similarity calculation on the extracted response spectral feature to obtain a preliminary identification result refers to comparing the response spectral feature extracted from the acoustic signal and / or tracer gas concentration signal with the feature templates of various types of air infiltration sources in the pre-set response spectral feature library. This comparison can be achieved by calculating the distance (such as Euclidean distance, Mahalanobis distance) or similarity (such as cosine similarity, correlation coefficient) between feature vectors, and its purpose is to quickly filter out the potential air infiltration source type that is closest to the detected air infiltration feature.

[0058] Further, when the preliminary identification result shows that the similarity of the extracted response spectrum feature to multiple feature templates in the preset response spectrum feature library is higher than the preset threshold, the fine structure, transient change rate and dynamic trajectory of the corresponding response spectrum feature are extracted, and a fine feature comparison result is obtained. The fine structure can be understood as the more subtle frequency components, accurate shape of resonance peaks or energy distribution patterns in the response spectrum feature; the transient change rate refers to the rapid change trend of the response spectrum feature in the time dimension, for example, the signal rising or decay rate under the pressure difference pulse excitation; the dynamic trajectory refers to the change path of the frequency, amplitude and other parameters of the response spectrum feature with time or the sweep frequency in the sweep frequency process. The extraction of these fine features aims to deeply mine the unique information contained in the response spectrum feature that can distinguish highly similar penetration source types, thereby providing more abundant basis for subsequent accurate identification.

[0059] In addition, in combination with the penetration source position and the adaptive ventilation module operating state, it is analyzed whether the response spectrum feature is affected by the airflow disturbance in the air cavity. The penetration source position information can assist in judging whether there is a specific structural gap or material connection point at the position, which may correspond to a specific penetration source type. The adaptive ventilation module operating state provides real-time information of the airflow environment inside the air cavity, for example, when the adaptive ventilation module is in the working state, the airflow disturbance generated thereby can introduce additional noise or pseudo-features in the acoustic signal, or even may alias with the real air penetration signal. By analyzing these context information, the response spectrum feature caused by real penetration can be effectively distinguished from the signal caused by non-penetration airflow disturbance, and misjudgment can be avoided.

[0060] Finally, according to the fine feature comparison result, the analysis result of the penetration source position and the adaptive ventilation module operating state, the preliminary identification result is corrected, and the penetration source type and attribution are determined. This means that the ambiguity of the preliminary identification is eliminated or reduced through in-depth analysis of fine features and comprehensive judgment of environmental context information, so that a more accurate and reliable penetration source type is obtained, and it is attributed to a specific physical position or structural defect.

[0061] In some preferred embodiments, assuming that when detecting a certain double-layer curtain wall, the acoustic signals acquired by the sensor after time-frequency spectrum analysis, the similarity of the extracted response spectrum features to the two templates of the sealant aging cracking and the gasket failure at the connection between the glass curtain wall and the frame is higher than the preset threshold in the preset feature library, resulting in an unclear preliminary identification result. At this time, the scheme of the present application will further start fine feature comparison. Specifically, the system will extract the fine structure of the response spectrum features, for example, the sealant aging cracking may exhibit a wider resonance peak at a certain high frequency band, while the gasket failure may produce a sharper resonance peak at a lower frequency band; at the same time, the transient change rate is analyzed, for example, the leakage caused by gasket failure may exhibit a faster signal decay rate under the action of pressure difference pulse. In addition, the system will also combine the preliminary location information of the penetration source (for example, whether the location is close to the glass edge or the frame connection point) and the running state of the adaptive ventilation module (for example, if the ventilation module is running, whether the airflow disturbance generated by it can be confused with the features of gasket failure). Through comprehensive analysis of these fine features and context information, for example, if the fine structure analysis shows a wider high-frequency resonance peak, and the location is near the sealant strip, while excluding the interference of the ventilation module, the preliminary identification result can be corrected, and the type of the penetration source is finally determined as sealant aging cracking. This multi-dimensional and fine identification process ensures the accurate attribution of the penetration source in complex scenarios.

[0062] The present application further proposes a step of performing feature enhancement processing on the response spectrum features, which includes: performing nonlinear enhancement processing on the response spectrum features based on an adaptive threshold; performing enhancement processing on the response spectrum features after nonlinear enhancement processing based on multi-scale detail fusion; performing sparse representation and reconstruction on the response spectrum features after multi-scale detail fusion enhancement.

[0063] Specifically, the nonlinear enhancement processing on the response spectrum features based on an adaptive threshold means that the parameters of the nonlinear transformation are dynamically adjusted according to the local characteristics of the response spectrum features to achieve enhancement of the features. For example, the gamma correction factor, threshold function or gain coefficient can be adaptively adjusted according to local signal-to-noise ratio, energy density or frequency change rate, etc., so as to apply different intensities of enhancement effect in different areas. The purpose is to effectively suppress the signals generated by background noise and non-penetrating airflow disturbance, while highlighting the key response components related to air penetration, and improving the signal-to-noise ratio of the features.

[0064] Wherein, the response spectrum feature after the nonlinear enhancement processing is subjected to the enhancement processing based on the multi-scale detail fusion, which can be understood as analyzing and processing the response spectrum feature at multiple different scales or resolutions, and integrating the information at different scales. For example, wavelet transform, multi-resolution analysis or pyramid decomposition technology can be used to obtain the detail information and profile information of the response spectrum feature at different frequency ranges and time windows. Subsequently, the multi-scale details are fused by weighted average, feature splicing or decision-level fusion to construct a more comprehensive and robust feature representation. The purpose is to capture the diversified features of the penetration source at different frequency and time scales, and avoid the loss of details or misjudgment caused by single scale analysis.

[0065] In practical applications, the response spectrum feature after the multi-scale detail fusion enhancement is subjected to sparse representation and reconstruction, which specifically refers to projecting the fused response spectrum feature onto an over-complete dictionary and finding its sparse linear combination. For example, a dictionary containing various known penetration source feature modes can be constructed, and then matching pursuit, basis pursuit or orthogonal matching pursuit algorithm is used to express the response spectrum feature as a linear combination of a small number of atoms in the dictionary. The reconstruction process uses these sparse coefficients and the dictionary to reconstruct the enhanced response spectrum feature. The purpose is to remove the redundant information and residual noise in the feature, highlight the structural features related to the essence of the penetration source, and obtain a more pure and higher discriminant response spectrum feature.

[0066] In some preferred embodiments, it is assumed that when detecting the double-layer curtain of a high-rise building, the sensor carried by the mobile detection platform obtains an acoustic signal containing air penetration information. Preliminary time-frequency spectrum analysis shows that there are some response components related to the excitation frequency of the swept-frequency pneumatic device, but these components are severely disturbed by environmental wind noise, building structure vibration and indoor air conditioning system operation noise, resulting in a very low signal-to-noise ratio of the response spectrum feature, making it difficult to accurately determine the type of penetration source.

[0067] At this time, the feature enhancement processing method of the present application is enabled. First, the original response spectrum feature is subjected to adaptive gamma correction based on local signal-to-noise ratio, which is nonlinear enhancement processing based on adaptive threshold. In the frequency band with low signal-to-noise ratio, the gamma correction factor is dynamically adjusted to a larger value to amplify the penetration signal; while in the frequency band with high signal-to-noise ratio, the gamma correction factor is close to 1 to maintain the original signal form. This step effectively suppresses the broadband noise, making the penetration-related resonance peaks clearer.

[0068] Then, the response spectrum features after the nonlinear enhancement processing are enhanced based on multi-scale detail fusion. For example, multi-scale wavelet decomposition is used to decompose the response spectrum into sub-bands with different frequency resolutions. At the fine scale, sharp resonant peaks caused by small cracks or aging of the seal strip can be captured; at the coarse scale, the wideband energy envelope caused by larger gaps or loose structural connections can be identified. Subsequently, by weighting and fusing the features at different scales, for example, giving higher weight to high-frequency details and lower weight to low-frequency profiles, a fused feature is generated that contains both fine details and overall trends.

[0069] Finally, the response spectrum features after multi-scale detail fusion enhancement are sparsely represented and reconstructed. For example, a dictionary containing response spectrum features of various typical penetration sources (such as sealant cracking, gasket aging, profile connection gap, etc.) is constructed in advance. The fused features are jointly sparsely decomposed on the general dictionary and the scene-specific sub-dictionary constructed for different depths and structural gap types inside the air cavity of the double-layer curtain wall. By weighting and fusing the joint sparse coefficients and reconstructing, a highly sparse and strongly discriminative feature vector can be obtained. This final enhanced feature vector can clearly indicate the type of penetration source, for example, clearly distinguishing between penetration caused by sealant cracking or penetration caused by gasket aging, thereby providing accurate and reliable data support for subsequent maintenance reports and strategy formulation.

[0070] The present application further proposes steps for nonlinear enhancement processing of response spectrum features based on adaptive threshold value, including: Adaptive gamma correction of the response spectrum features based on local signal-to-noise ratio, with the gamma correction factor dynamically adjusted according to the local signal-to-noise ratio of the response spectrum features, providing stronger enhancement in low signal-to-noise ratio areas and maintaining the original form of the response spectrum features in high signal-to-noise ratio areas; During the adjustment of the gamma correction factor, the instantaneous energy change rate of the response spectrum features is monitored; When the instantaneous energy change rate exceeds a preset change rate threshold, a noise suppression strategy based on frequency segmentation is enabled, with additional attenuation of high-energy disturbance frequency bands and smoothing of the gamma correction factor of adjacent frequency bands; Combined with the speed and attitude information of the mobile detection platform, the response spectrum features are spatially calibrated to compensate for the estimation bias of the local signal-to-noise ratio caused by the movement of the mobile detection platform.

[0071] Specifically, the local SNR-based adaptive gamma correction aims to dynamically adjust the gamma correction factor according to the SNR level of the response spectrum feature in different frequency bands or time windows. The gamma correction factor is a nonlinear transformation parameter used to adjust the contrast and brightness of the signal. When the local SNR is low, the gamma correction factor is set to a small value to provide stronger nonlinear enhancement, thereby improving the visibility of weak signals; when the local SNR is high, the gamma correction factor is set to a value close to 1 to maintain the original form of the response spectrum feature and avoid introducing distortion.

[0072] During the adjustment of the gamma correction factor, the instantaneous energy change rate of the response spectrum feature is monitored, which aims to timely discover and respond to high-energy transient events that may be caused by burst noise or non-invasive airflow disturbance. When the instantaneous energy change rate exceeds the preset change rate threshold, it indicates that there is a significant transient disturbance. At this time, the frequency segment-based noise suppression strategy will be enabled to perform additional attenuation on the detected high-energy disturbance frequency band to effectively suppress the interference of these non-invasive signals. At the same time, in order to ensure the smoothness and continuity of the enhancement processing, the gamma correction factors of adjacent frequency bands are smoothed to avoid unnatural transitions caused by local drastic adjustments.

[0073] In practical applications, the speed and attitude information of the mobile detection platform is combined to calibrate the spatial position of the response spectrum feature, which aims to compensate for the estimation deviation of the local SNR caused by the movement of the mobile detection platform. During the movement of the mobile detection platform, changes in its speed and attitude (such as pitch, roll, and yaw) can cause dynamic changes in the signal strength and noise level received by the sensor, thereby affecting the accurate estimation of the local SNR. By introducing the speed and attitude information of the mobile detection platform, the estimated local SNR can be corrected to ensure that the gamma correction factor accurately reflects the actual signal quality, thereby improving the robustness and accuracy of the nonlinear enhancement processing.

[0074] In some preferred embodiments, the mobile detection platform carrying the acoustic sensor is assumed to scan and detect the surface of the building curtain wall. During the detection process, the acoustic signals received by the sensor in a certain area contain not only weak air-permeable acoustic signals, but also background environmental noise and occasional transient wind noise.

[0075] First, the system performs time-frequency spectrum analysis on the received acoustic signals to obtain response spectrum features. Then, the response spectrum features are subjected to adaptive gamma correction based on local signal-to-noise ratio. For example, at the joint of the curtain wall, the acoustic signal generated by air penetration is relatively weak, and the local signal-to-noise ratio is low. At this time, the gamma correction factor is dynamically adjusted to a small value (e.g. 0.5) to perform stronger non-linear enhancement on the response spectrum features in this area, so that the weak penetration signal can be highlighted. In the glass area of the curtain wall, the signal is relatively clear, and the local signal-to-noise ratio is high. The gamma correction factor is adjusted to a value close to 1 (e.g. 0.95) to maintain the original form of the signal and avoid over-enhancement.

[0076] At the same time, the system continuously monitors the instantaneous energy change rate of the response spectrum features. When a sudden burst of strong wind blows through the curtain wall, causing the instantaneous energy change rate of the acoustic signal to suddenly exceed the preset threshold, the system will immediately enable the frequency segment-based noise suppression strategy. For example, if it is identified that the wind noise is mainly concentrated in a certain frequency band (such as 500Hz-1000Hz), the response spectrum features in this frequency band are additionally attenuated, and the gamma correction factors of adjacent frequency bands are smoothed to eliminate the interference of wind noise and prevent it from being misidentified as a penetration signal.

[0077] In addition, since the mobile detection platform moves slowly on the surface of the curtain wall, its speed and attitude (such as slight tilt or sway) information is obtained in real time. These information is used to calibrate the spatial position of the response spectrum features. For example, if the platform tilts slightly at a certain time, causing a slight change in the distance between the sensor and the curtain wall, the system will correct the estimation of the local signal-to-noise ratio according to the tilt angle and speed information, so that the gamma correction factor can be accurately applied to the current spatial position, compensating for the signal attenuation or enhancement deviation caused by platform motion, further improving the accuracy of the enhancement processing.

[0078] Through the above series of adaptive and calibration measures, the final response spectrum features have higher signal-to-noise ratio and stronger distinguishability, thereby providing a more reliable data basis for subsequent penetration source type identification and attribution.

[0079] The present application further proposes a step of performing multi-scale detail fusion based enhancement processing on the response spectrum features after nonlinear enhancement processing, which includes: performing multi-scale detail fusion based on feature similarity clustering on the response spectrum features after nonlinear enhancement processing; clustering similar resonance peaks, energy envelopes and frequency shift patterns; performing weighted combination on each group of resonance peaks, energy envelopes and frequency shift patterns to enhance the distinguishability of resonance peaks, energy envelopes and frequency shift patterns.

[0080] Specifically, the multi-scale detail fusion based on feature similarity clustering refers to, in the process of multi-scale detail fusion, first conducting in-depth analysis on response spectrum features, and identifying different types of features contained therein, such as resonance peaks, energy envelopes, and frequency shift patterns. Subsequently, according to the similarity between these features, such as similarity in frequency, amplitude, duration, or morphology, similar features are grouped or clustered. This clustering can be a distance-based clustering algorithm (such as K-means, DBSCAN) or a density-based clustering algorithm. The purpose is to classify features with common sources or similar physical meanings into the same category for subsequent targeted enhancement processing.

[0081] Among them, the clustering of similar resonance peaks, energy envelopes, and frequency shift patterns can be understood as follows: after preliminary analysis of the response spectrum features, the system identifies multiple discrete resonance peaks, energy envelopes, and frequency shift patterns. These patterns may be generated by different penetration sources or non-penetration airflow disturbances, or may be generated by the same penetration source at different scales or different time points. By calculating the similarity between these patterns, such as using Euclidean distance, cosine similarity, or dynamic time warping, similar patterns are grouped into the same group. For example, multiple resonance peaks generated by the same penetration source may have a slight frequency shift, but their overall morphology and energy envelope have high similarity, so they will be grouped into the same group.

[0082] In practical applications, weighted combination is performed on each group of resonance peaks, energy envelopes, and frequency shift patterns to enhance the distinguishability of resonance peaks, energy envelopes, and frequency shift patterns. Specifically, after feature clustering, for the features in each clustering group, weighted average, weighted summation, or more complex fusion algorithms are used for combination. The weighting factor can be dynamically adjusted according to the signal-to-noise ratio, energy intensity, stability, or matching degree with known penetration source feature templates of the features. For example, features with high signal-to-noise ratio and strong energy are given greater weight to highlight their importance; stable features that appear at multiple scales can also have their weights increased. Through this weighted combination, noise and redundant information within the group can be effectively suppressed, while the expression of core features within the group can be strengthened, thereby improving the distinguishability between different feature groups and making subsequent type identification and penetration source attribution more accurate.

[0083] In some preferred embodiments, it is assumed that after time-frequency spectrum analysis and nonlinear enhancement processing of a certain acoustic signal, response spectrum features containing multiple resonance peaks, energy envelopes, and frequency shift patterns are obtained. For example, there may be three resonance peaks A, B, and C, two energy envelopes D and E, and one frequency shift pattern F.

[0084] First, the system will calculate the similarities between these features. For example, resonant peaks A and B can be very close in frequency and have similar energy envelope shapes, but resonant peak C is significantly different in frequency. Energy envelopes D and E can have different durations and amplitudes. Frequency shift pattern F can be associated with a particular type of infiltration source.

[0085] Based on these similarity calculation results, the system will perform clustering. For example, resonant peaks A and B can be clustered into a first group because they can originate from the vibration response of the same micro crack. Resonant peak C can be clustered into a second group because it can originate from a larger structural gap. Energy envelopes D and E can be clustered into different groups, or if they are highly correlated with a certain resonant peak, they can be clustered together with that resonant peak. Frequency shift pattern F can form a group alone, or be clustered together with the specific resonant peak or energy envelope that causes it.

[0086] Subsequently, for each cluster group, such as resonant peaks A and B in the first group, the system will perform weighted combination according to their signal-to-noise ratio, energy intensity, etc. For example, if the signal-to-noise ratio of resonant peak A is higher than that of resonant peak B, more weight will be given to resonant peak A in the combination, so as to generate a more representative and discriminative fused feature. In this way, even if there are subtle differences between the original features or they are affected by noise, the differences between the feature groups generated by different infiltration sources will be effectively amplified after clustering and weighted combination, thereby significantly enhancing the discriminability of the response spectrum features, providing a clearer basis for subsequent infiltration source type identification.

[0087] The present application further proposes the steps of sparse representation and reconstruction of the response spectrum features enhanced by multi-scale detail fusion, including: For the air cavity inside the double-layer curtain wall, the response spectrum features enhanced by multi-scale detail fusion are subjected to multi-dictionary collaborative sparse representation; The multi-dictionary collaborative sparse representation includes a general dictionary and one or more scene-specific sub-dictionaries. The general dictionary is used to capture the common features of the infiltration source, and the scene-specific sub-dictionaries are constructed for infiltration sources of different depths and different structural gap types in the air cavity to capture the fine features of the infiltration source. The response spectrum features enhanced by multi-scale detail fusion are subjected to joint sparse decomposition on the general dictionary and the scene-specific sub-dictionaries to obtain joint sparse coefficients. According to the preliminary position information of the infiltration source and the similarity of the response spectrum features, the joint sparse coefficients are subjected to weighted fusion to dynamically adjust the contribution weights of the general dictionary and the scene-specific sub-dictionaries, so as to improve the environmental adaptability of the detection. The response spectrum features with enhanced discriminability are obtained by reconstructing the weighted fused joint sparse coefficients and the general dictionary and the scene-specific sub-dictionaries.

[0088] Specifically, the multi-dictionary collaborative sparse representation refers to jointly decomposing and representing the response spectrum features by using multiple dictionaries. The general dictionary can be understood as a set of basis vectors containing various common penetration source response features, and its purpose is to capture the common response patterns of different penetration sources. For example, the general dictionary can be obtained by training and learning a large number of typical response spectrum features of different types of curtain wall penetration sources in advance. The scene-specific sub-dictionary is customized for specific environments and penetration source types in the double-layer curtain wall air cavity, for example, independent sub-dictionaries can be constructed for air cavities of different depths and structural gaps of different materials (such as aging of sealant, loosening of profile connection, etc.). These sub-dictionaries aim to capture the unique and fine response features of penetration sources in specific scenarios, thereby improving the recognition accuracy of complex penetration sources.

[0089] Further, the response spectrum features enhanced by multi-scale detail fusion are jointly sparsely decomposed on the general dictionary and the scene-specific sub-dictionaries, which aims to represent the input signal using both general features and scene-specific features. Joint sparse decomposition can be implemented using algorithms such as orthogonal matching pursuit or basis pursuit, etc., to find the sparsest representation on these dictionaries. Thus, joint sparse coefficients can be obtained, which reflect the projection intensity of the input response spectrum features on the basis vectors of each dictionary.

[0090] Among them, the joint sparse coefficients are weighted fused according to the preliminary position information of the penetration source and the similarity of the response spectrum features, which aims to dynamically adjust the contribution of different dictionaries in the final feature representation. For example, when the preliminary position information indicates that the penetration source is located in the deep air cavity and the response spectrum features are highly similar to a certain specific structural gap sub-dictionary, the weight of this sub-dictionary can be increased, while the weight of the general dictionary can be appropriately reduced. This dynamic adjustment mechanism enables the sparse representation process to better adapt to the complexity and diversity of the actual detection environment, thereby improving the environmental adaptability of detection.

[0091] Finally, the response spectrum features with higher discriminability are obtained by reconstructing the weighted fused joint sparse coefficients and the general dictionary and the scene-specific sub-dictionaries, which aims to recover the response spectrum features with higher discriminability from the sparse representation. The reconstruction process is essentially a linear combination of sparse coefficients and corresponding dictionary basis vectors, thereby generating a purer and more representative feature representation that can more clearly reflect the essential characteristics of the penetration source and effectively suppress noise and irrelevant disturbances.

[0092] As a specific embodiment, assume that the mobile detection platform detects abnormal acoustic signals in a certain region inside the air cavity of a double-skin facade building during the detection process. After multi-scale detail fusion enhancement processing, a group of response spectrum features is obtained. At this time, the system will use the pre-constructed multi-dictionary collaborative sparse representation model. The model contains a general dictionary, which covers the typical response spectrum features of common types of facade penetration (such as sealant aging, profile connection loosening, glass breakage, etc.). At the same time, for the specific structure of this double-skin facade, multiple scene-specific sub-dictionaries are also constructed, for example, one sub-dictionary is specifically for the micro-crack penetration at the glass and profile connection, and another sub-dictionary is for the penetration of the thermal insulation layer material in the deep part of the air cavity.

[0093] In the sparse representation of the detected response spectrum features, first, joint sparse decomposition is performed on the general dictionary and all scene-specific sub-dictionaries to obtain a group of joint sparse coefficients. Subsequently, based on the preliminary position information provided by the speed and attitude information of the mobile detection platform (for example, this region is located in the lower part of the facade, close to the glass and profile connection) and the preliminary similarity of the response spectrum features and each dictionary template, the system dynamically adjusts the contribution weights of these dictionaries. For example, if the preliminary analysis shows that the response spectrum features have a high similarity with the scene-specific sub-dictionary of the micro-crack penetration at the glass and profile connection, the weight of this sub-dictionary will be significantly increased, while the weights of other dictionaries will be correspondingly reduced. Through this weighted fusion, the final joint sparse coefficients can more accurately reflect the essential features of the penetration source. Finally, using the weighted fusion joint sparse coefficients and the corresponding dictionaries, a highly differentiated response spectrum feature is obtained, which can clearly indicate that the abnormal signal is caused by the micro-crack penetration at the glass and profile connection, rather than other types of penetration or background noise.

[0094] Specifically, in the above building facade detection method, the step of obtaining the acoustic signals and / or tracer gas concentration signals generated by air penetration through sensors can be further refined.

[0095] Specifically, the step of obtaining the acoustic signals and / or tracer gas concentration signals generated by air penetration through sensors includes: obtaining the acoustic signals generated by air penetration through acoustic sensors; obtaining the tracer gas concentration signals generated by air penetration through gas sensors; When both the acoustic signals generated by air penetration and the tracer gas concentration signals generated by air penetration need to be obtained, the acoustic signals generated by air penetration and the tracer gas concentration signals generated by air penetration are collected in a synchronous manner to ensure that the time stamps of the acoustic signals generated by air penetration and the tracer gas concentration signals generated by air penetration are consistent.

[0096] The acoustic sensor can be understood as a device capable of converting sound waves into electrical signals, for example, it can be a high-sensitivity microphone array or an ultrasonic sensor, and its purpose is to capture weak acoustic features generated during air infiltration. The gas sensor is a device capable of detecting changes in the concentration of a specific gas, for example, it can be a tracer gas (such as SF6 or helium) detector, and its purpose is to quantify the amount of tracer gas leakage caused by air infiltration. In practical applications, the synchronous acquisition means that the acoustic signal and the tracer gas concentration signal are collected with high consistency in time. This can be achieved by using a unified clock source, a time stamp synchronization protocol, or an integrated multi-channel data acquisition system. Ensuring that the time stamps of the acoustic signal generated by air infiltration and the tracer gas concentration signal generated by air infiltration are consistent, the purpose is to ensure that the two different types of signals have comparability in the time dimension, so as to accurately associate acoustic events with gas concentration changes, and provide accurate time reference for subsequent data processing and infiltration point identification.

[0097] The scheme of the present application can more comprehensively and accurately capture the physical characteristics of air infiltration by refining the collection of acoustic signals and tracer gas concentration signals and emphasizing synchronous acquisition. The acoustic sensor can directly perceive the turbulent noise or resonance sound generated when air flows through a small gap, while the gas sensor can quantify the amount of infiltration through the diffusion of tracer gas. By synchronously collecting these two signals, a direct correlation between acoustic features and gas leakage can be established, for example, a specific acoustic pattern may correspond to a certain type of infiltration source and its leakage rate. This synchronization enables accurate matching of acoustic events and gas concentration changes occurring at the same time point when analyzing data, thereby avoiding misjudgment or information loss due to different time synchronization.

[0098] Through the above technical scheme, the accuracy and reliability of air infiltration detection can be significantly improved. Specifically, through the cooperative work of the acoustic sensor and the gas sensor, multi-dimensional perception of air infiltration phenomena can be achieved, making up for the limitations of a single sensor in some scenarios. For example, acoustic signals have an advantage in quickly locating small cracks, while tracer gas concentration signals can provide more accurate leakage quantification. In addition, ensuring the consistency of the time stamps of the two signals collected makes subsequent data fusion and analysis more accurate, which helps to more accurately identify the infiltration point, judge the type of infiltration, and provide high-quality raw data for subsequent quantitative evaluation and repair report generation, thereby improving the effectiveness of the entire detection method.

[0099] The above disclosed content is only the preferred feasible embodiment of the present application, and does not limit the protection scope of the present application, so any equivalent technical changes made according to the content of the specification and drawings of the present application are included in the protection scope of the present application, and in addition, the elements can be updated as technology develops.

Claims

1. A method for detecting a building curtain wall, characterized by, The method comprises the following steps: deploying a mobile detection platform carrying sensors; acquiring acoustic signals and / or tracer gas concentration signals generated by air infiltration through the sensors; processing the acoustic signals and / or tracer gas concentration signals to identify air infiltration points and determine the locations of the air infiltration points; quantitatively evaluating the air infiltration points and generating a repair report.

2. The method of claim 1, wherein the method further comprises: The method further comprises the following steps: deploying a swept-frequency pneumatic device that generates differential pressure pulses with frequency modulation characteristics; acquiring acoustic signals and / or tracer gas concentration signals generated by air infiltration through the sensors under the action of the differential pressure pulses; performing time-frequency spectrum analysis on the acoustic signals and / or tracer gas concentration signals to extract response spectrum features related to the frequency of the differential pressure pulses; performing type identification on the extracted response spectrum features according to a pre-set response spectrum feature library, and attributing the features to infiltration sources; determining the locations of the infiltration sources.

3. The method of claim 2, wherein the method further comprises: The steps of quantitatively evaluating the air infiltration points and generating a repair report comprise: estimating the air leakage amount of the air infiltration points under standard differential pressure conditions according to the types of the infiltration sources and the response spectrum intensities of the infiltration sources; acquiring indoor and outdoor pressure difference data, adaptive ventilation module operating states, local wind speed, and environmental temperature and humidity; dynamically correcting the estimated air leakage amount based on the indoor and outdoor pressure difference data, adaptive ventilation module operating states, local wind speed, and environmental temperature and humidity, and calculating the long-term energy consumption contribution of the air infiltration points under current and predicted environmental conditions; predicting the leakage amount variation trend of the air infiltration points in a future period of time according to the types of the infiltration sources, material aging rate data, and historical repair records; comprehensively calculating the repair priority of the air infiltration points and generating a report containing adaptive repair strategies based on the dynamically corrected air leakage amount, long-term energy consumption contribution prediction, types of the infiltration sources, locations of the infiltration sources, and predicted leakage amount variation trend.

4. The method of claim 2, wherein the method further comprises: The steps of performing time-frequency spectrum analysis on the acoustic signals and / or tracer gas concentration signals to extract response spectrum features related to the frequency of the differential pressure pulses comprise: performing multi-channel synchronous acquisition on the acoustic signals and / or tracer gas concentration signals; using adaptive filtering technology to separate response components synchronized with the excitation frequency of the differential pressure pulses from the acoustic signals and / or tracer gas concentration signals according to the excitation frequency characteristics of the swept-frequency pneumatic device, and suppressing background noise and signals generated by non-infiltration air flow disturbances; performing multi-resolution time-frequency decomposition on the separated response components to obtain time-frequency spectrograms under different frequency resolutions; identifying and extracting resonance peaks, energy envelopes, and frequency shift patterns generated by different infiltration sources or non-infiltration air flow disturbances in the inner and outer layers of the double-layer curtain and the air cavity under specific swept-frequency in the time-frequency spectrograms under different frequency resolutions as response spectrum features; performing feature enhancement processing on the response spectrum features.

5. The method of claim 3, wherein the method comprises: The steps of performing type identification on the extracted response spectrum features according to a pre-set response spectrum feature library, and attributing the features to infiltration sources comprise: performing preliminary similarity calculation on the extracted response spectrum features to obtain preliminary identification results; When the preliminary identification result shows that the similarity between the extracted response spectrum feature and each feature template in the preset response spectrum feature library is higher than the preset threshold, the fine structure, transient change rate and dynamic trajectory of the corresponding response spectrum feature are extracted, and a fine feature comparison result is obtained; In combination with the position of the penetration source and the running state of the adaptive ventilation module, it is analyzed whether the response spectrum feature is affected by the airflow disturbance in the air cavity; According to the fine feature comparison result and the analysis result of the position of the penetration source and the running state of the adaptive ventilation module, the preliminary identification result is corrected to determine the type of the penetration source and the attribution.

6. The method of claim 4, wherein the step of detecting the position of the building curtain wall is performed by using a camera. The steps of feature enhancement processing of the response spectrum feature include: Nonlinear enhancement processing of the response spectrum feature based on an adaptive threshold; Enhancement processing of the response spectrum feature after nonlinear enhancement processing based on multi-scale detail fusion; Sparse representation and reconstruction of the response spectrum feature after multi-scale detail fusion enhancement.

7. The method of claim 6, wherein the method further comprises: The steps of nonlinear enhancement processing of the response spectrum feature based on an adaptive threshold include: Adaptive gamma correction of the response spectrum feature based on local signal-to-noise ratio, and the gamma correction factor is dynamically adjusted according to the local signal-to-noise ratio of the response spectrum feature, providing stronger enhancement in the low signal-to-noise ratio area and keeping the original form of the response spectrum feature in the high signal-to-noise ratio area; During the adjustment of the gamma correction factor, the instantaneous energy change rate of the response spectrum feature is monitored; When the instantaneous energy change rate exceeds the preset change rate threshold, a noise suppression strategy based on frequency segmentation is enabled, additional attenuation is performed on the high-energy disturbance frequency band, and the gamma correction factor of the adjacent frequency band is smoothed; In combination with the speed and attitude information of the mobile detection platform, the spatial position of the response spectrum feature is calibrated to compensate for the estimation deviation of the local signal-to-noise ratio caused by the movement of the mobile detection platform.

8. The method of claim 6, wherein the method further comprises: The steps of enhancement processing of the response spectrum feature after nonlinear enhancement processing based on multi-scale detail fusion include: Multi-scale detail fusion of the response spectrum feature after nonlinear enhancement processing based on feature similarity clustering; Clustering groups similar resonance peaks, energy envelopes and frequency shift patterns; Weighted combination is performed on each group of resonance peaks, energy envelopes and frequency shift patterns to enhance the distinguishability of resonance peaks, energy envelopes and frequency shift patterns.

9. The method of claim 6, wherein the method further comprises: The steps of sparse representation and reconstruction of the response spectrum feature after multi-scale detail fusion enhancement include: Multi-dictionary collaborative sparse representation of the response spectrum feature after multi-scale detail fusion enhancement is performed for the inside of the air cavity of the double-layer curtain; The multi-dictionary collaborative sparse representation includes a general dictionary and one or more scene-specific sub-dictionaries, the general dictionary is used to capture the common features of the penetration source, and the scene-specific sub-dictionaries are constructed for penetration sources of different depths and different structural gap types in the air cavity to capture fine features of the penetration source; Joint sparse decomposition of the response spectrum feature after multi-scale detail fusion enhancement on the general dictionary and the scene-specific sub-dictionary is performed to obtain joint sparse coefficients; According to the initial position information of the penetration source and the similarity of the response spectrum features, the joint sparse coefficients are weighted and fused, and the contribution weights of the general dictionary and the scene-specific sub-dictionary are dynamically adjusted to improve the environmental adaptability of detection; The reconstructed response spectrum features with enhanced distinguishability are obtained through the joint sparse coefficients after weighted fusion, the general dictionary and the scene-specific sub-dictionary.

10. The method for detecting building curtain walls as described in claim 1, characterized in that, The step of obtaining the acoustic signal generated by air penetration and / or the tracer gas concentration signal through a sensor comprises: obtaining the acoustic signal generated by air penetration through an acoustic sensor; obtaining the tracer gas concentration signal generated by air penetration through a gas sensor; When it is necessary to obtain the acoustic signal generated by air penetration and the tracer gas concentration signal generated by air penetration, the acoustic signal generated by air penetration and the tracer gas concentration signal generated by air penetration are collected in a synchronous manner to ensure that the time stamps of the acoustic signal generated by air penetration and the tracer gas concentration signal generated by air penetration are consistent.

Images