Building damage early-stage detection method and system based on multi-field coupling and related equipment

Abstract

The invention provides a building damage early-stage detection method and system based on multi-field coupling and related equipment, and relates to the technical field of house safety monitoring. Respectively acquiring excitation acoustic response, selected wave band spectral reflection and transient heat conduction response signals of the target building facade; performing time-frequency domain transformation processing on the signal to obtain a multi-physical field feature vector; obtaining target building facade material attribute parameters and environmental action historical data, and constructing a degradation evolution knowledge graph according to the attribute parameters and the environmental action historical data; performing similarity matching on the multi-physical field feature vector and each degradation mode feature template, and judging whether a potential degradation mode exists or not; if yes, time sequence correlation analysis is carried out on the potential degradation mode, physical field characteristic evolution trends in continuous M collection periods are determined, and a degradation development rate index is determined accordingly; and when the index is greater than a preset rate threshold, generating degradation early warning information.

Description

Technical Field

[0001] This application relates to the field of building safety monitoring technology, and in particular to a method, system and related equipment for early detection of building damage based on multi-field coupling. Background Technology

[0002] With the rapid development of urbanization and the continuous growth of building stock, structural health monitoring of buildings, especially old buildings, has become an important part of urban safety management. Early identification of their deterioration status is of great significance for protecting people's lives and property and optimizing the allocation of maintenance resources.

[0003] In related technologies, a common approach is to employ intelligent building inspection methods based on multi-sensor fusion from drones. Specifically, the process begins with the drone-mounted LiDAR scanning the building facade to construct a high-precision 3D point cloud model as a spatial positioning reference. Next, a high-resolution visible light camera captures images of the building surface along a pre-defined flight path, and image processing algorithms identify surface defects such as cracks, peeling, and dirt. Simultaneously, an infrared thermal imager collects the building surface temperature distribution, detecting abnormal temperature areas caused by water seepage, insulation layer detachment, etc. The LiDAR point cloud data, visible light images, and infrared thermal images are then registered and aligned in a unified spatial coordinate system. Finally, a deep learning network extracts and fuses features from the multi-source data to automatically identify various building defects and generate an inspection report. This multi-sensor collaborative approach can acquire surface information from different dimensions, improving the comprehensiveness and accuracy of defect detection.

[0004] However, when using the multi-sensor fusion detection method described above, the data collected by each sensor are independent in the spatiotemporal dimension. The data fusion only performs a simple combination at the feature level, ignoring the inherent correlation between different physical signals in the early deterioration process. This may result in only being able to detect building damage that has already been characterized as macroscopic defects, thus missing the critical early stage of building deterioration development, and consequently leading to low detection timeliness of building deterioration in related technologies. Summary of the Invention

[0005] This application provides a method, system, and related equipment for early detection of building damage based on multi-field coupling, which can improve the timeliness of building deterioration detection.

[0006] Firstly, this application provides a method for early detection of building damage based on multi-field coupling, applied to the aforementioned early detection system for building damage. The method includes: acquiring, via a UAV-borne acoustic field sensing sensor group, spectral sensing sensor group, and thermal field sensing sensor group, the excitation acoustic response signal, selected band spectral reflection signal, and transient thermal conduction response signal of the target building facade, respectively; performing time-frequency domain transformation processing on the excitation acoustic response signal, the selected band spectral reflection signal, and the transient thermal conduction response signal to obtain a multi-physics field feature vector; acquiring the material property parameters and historical environmental impact data of the target building facade, and constructing a multi-physics field feature vector based on the material property parameters and the historical environmental impact data. A degradation evolution knowledge graph is constructed, which includes physical field response feature templates for multiple degradation modes. The physical field feature vectors are matched with the physical field response feature templates of the multiple degradation modes to determine whether a potential degradation mode exists on the target building facade. When a potential degradation mode is determined to exist on the target building facade, a temporal correlation analysis is performed on the potential degradation mode to determine the evolution trend of the physical field features of the potential degradation mode over M consecutive collection periods. A degradation development rate index is determined based on the evolution trend of the physical field features. When the degradation development rate index is greater than a preset rate threshold, early warning information for early degradation is generated, where M is a positive integer.

[0007] By employing the above technical solution, three heterogeneous sensor groups mounted on a UAV simultaneously collect response signals from the target building facade in different physical fields. The acoustic field sensing sensor group acquires the excitation acoustic response signal, which reflects changes in the internal microstructure of the material; the spectral sensing sensor group acquires the selected band spectral reflectance signal, which reveals changes in surface chemical composition; and the thermal field sensing sensor group records the transient thermal conduction response signal, which characterizes the degradation of the material's thermophysical properties. These three signals complement and verify each other. Time-frequency domain transformation processing converts the original signals into feature vectors, allowing different physical quantities to be expressed in a unified mathematical space. A degradation evolution knowledge graph jointly constructed from material property parameters and historical environmental data provides prior knowledge support. Similarity matching between multi-physics feature vectors and physical field response feature templates enables accurate identification of degradation patterns. Temporal correlation analysis captures the evolutionary laws of features, and threshold judgment of degradation development rate indicators triggers an early warning mechanism. The combination of multi-physics collaborative sensing and knowledge-driven pattern recognition allows the time window for degradation detection to be moved forward to the early stage of damage. This solves the technical problem of low detection timeliness of building deterioration in related technologies, and achieves the technical effect of improving the detection timeliness of building deterioration.

[0008] Secondly, embodiments of this application provide a building damage early detection system, which includes: one or more processors and a memory; the memory is coupled to one or more processors, and the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the building damage early detection system to perform the method described in the first aspect and any possible implementation thereof.

[0009] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a building damage early detection system, cause the building damage early detection system to perform the method described in the first aspect and any possible implementation thereof.

[0010] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a building damage early detection system, cause the building damage early detection system to perform the method described in the first aspect and any possible implementation thereof. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating a method for early detection of building damage based on multi-field coupling in an embodiment of this application. Figure 2 This is a schematic diagram of the physical device structure of a building damage early detection system in the embodiments of this application. Detailed Implementation

[0012] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0013] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0014] This application provides a method for early detection of building damage based on multi-field coupling, see reference. Figure 1 , Figure 1 This is a flowchart illustrating a method for early detection of building damage based on multi-field coupling, as described in this application, including the following steps: Step S101: The acoustic field sensing sensor group, spectral sensing sensor group and thermal field sensing sensor group carried by the UAV are used to collect the excitation acoustic response signal, selected band spectral reflection signal and transient thermal conduction response signal of the target building facade, respectively. Step S102: Perform time-frequency domain transformation on the excitation acoustic response signal, the selected band spectral reflection signal, and the transient thermal conduction response signal to obtain a multi-physics feature vector; Step S103: Obtain the material property parameters and environmental impact history data of the target building facade, and construct a degradation evolution knowledge graph based on the material property parameters and environmental impact history data. The degradation evolution knowledge graph includes physical field response feature templates for multiple degradation modes. Step S104: Perform similarity matching between the multiphysics feature vector and the physical field response feature template of the multiple degradation modes to determine whether the target building facade has a potential degradation mode. Step S105: When it is determined that the potential deterioration mode exists on the target building facade, a time-series correlation analysis is performed on the potential deterioration mode to determine the evolution trend of the physical field characteristics of the potential deterioration mode within M consecutive collection cycles, and a deterioration development rate index is determined based on the evolution trend of the physical field characteristics. When the deterioration development rate index is greater than a preset rate threshold, early warning information for early deterioration is generated, where M is a positive integer.

[0015] Among them, "drone" refers to an unmanned aerial vehicle equipped with multiple sensors for aerial operations, such as a quadcopter drone equipped with acoustic, spectral, and thermal sensors; "sound field sensing sensor group" refers to a sensor array composed of multiple sound wave transmitters and receivers, used to emit frequency-sweeping sound waves and receive reflected and transmitted sound wave signals, such as acoustic detection equipment composed of ultrasonic transducers and acoustic microphone arrays; "spectral sensing sensor group" refers to a set of sensors capable of collecting spectral information in different bands, used to acquire reflectance spectral data in ultraviolet, visible, and near-infrared bands, such as hyperspectral cameras and multispectral imagers; "thermal field sensing sensor group" refers to a sensor system capable of applying thermal excitation and synchronously acquiring temperature field changes, used to record material... The evolution of material surface temperature over time, such as by combining infrared thermal imagers and pulsed thermal excitation devices; the target building facade refers to the exterior wall surface of a building that needs to be inspected for deterioration, such as concrete exterior walls, brick and stone curtain walls, and other vertical structural surfaces; the excitation acoustic response signal refers to the reflected and transmitted wave signals generated by the building material under acoustic excitation, used to characterize changes in the material's acoustic properties; the selected band spectral reflection signal represents the light reflection characteristics of the building surface within a specific band range, used to reveal changes in the material's surface and chemical composition; the transient thermal conduction response signal refers to the time-series change data of the material's temperature field after pulsed thermal excitation, used to reflect the material's internal thermal conductivity and defect distribution; time-frequency domain transformation processing refers to converting the time-domain signal to the frequency domain or... Mathematical processing methods for time-frequency joint domain analysis, such as Fourier transform and wavelet transform; multi-physics eigenvectors refer to high-dimensional eigenvectors formed by fusing characteristic parameters from multiple physical domains such as acoustics, optics, and thermodynamics, used to comprehensively characterize the material degradation state; material property parameters represent the inherent physicochemical properties of building materials, such as strength grade, porosity, and chemical composition; historical environmental data refers to recorded data on historical environmental conditions in the area where the building is located, such as historical records of temperature and humidity changes, acid rain erosion, and freeze-thaw cycles; a degradation evolution knowledge graph represents a structured knowledge representation describing the degradation development law of building materials under different environmental effects, including entities such as degradation patterns, driving factors, and characteristic responses, as well as their relationships; degradation patterns This refers to the typical types and manifestations of building material degradation, such as carbonization, chloride ion corrosion, and freeze-thaw damage; the physical field response feature template represents the standard feature patterns of each degradation mode in the physical domains such as acoustics, optics, and thermal, used for matching and comparison with measured data; potential degradation modes refer to the types of degradation that may be occurring, identified through feature matching; time series correlation analysis represents the time series analysis of data from multiple consecutive time periods, used to reveal the evolution law of feature parameters over time; the acquisition period refers to the time interval for a complete multi-physics data acquisition, such as once a month or once a quarter; the physical field feature evolution trend represents the development direction and speed of multi-physics feature parameters changing over time, used to predict the degradation development trend;The degradation rate index is a comprehensive indicator that quantifies the speed of degradation progress. It is obtained through weighted calculation of the evolution rates of multiple domain features. The preset rate threshold represents a pre-set warning value for the degradation rate; when the actual rate exceeds this threshold, an early warning needs to be issued. Early degradation warning information refers to warning notifications generated in the early stages of degradation development, containing key information such as degradation type, location, and development speed, used to guide early maintenance intervention.

[0016] In the above embodiment, a multi-story residential building in an old urban residential area, built more than 30 years ago, needs to undergo a facade health status assessment. The building is a 7-story brick-concrete structure with a cement mortar plaster layer on the exterior facade, coated with latex paint. Due to long-term exposure to the natural environment, some areas have shown slight signs of surface degradation. First, an octocopter UAV platform equipped with a multi-sensor payload is deployed. The UAV carries an acoustic field sensing sensor group including four piezoelectric ultrasonic transducers (center frequencies of 20kHz, 40kHz, 60kHz, and 80kHz, respectively) and a 16-channel MEMS microphone array; a spectral sensing sensor group is equipped with a hyperspectral imager covering the 350-2500nm band, with a spectral resolution of 5nm and a spatial resolution of centimeters; and a thermal field sensing sensor group uses a cooled mid-wave infrared thermal imager with a temperature resolution of 0.02K, combined with a 2kW halogen lamp array as a pulse heat source. When performing the testing task, based on the building's material characteristics (cement mortar plastering, design strength grade M10) and environmental exposure conditions (average annual temperature 15℃, relative humidity 65%, annual rainfall 800mm, and freeze-thaw cycles in winter), the multi-physics field excitation parameters were automatically configured: the acoustic field excitation selected a 10-50kHz sweep frequency signal, with a focus on the 20-30kHz resonant frequency band; the spectral detection focused on acquiring key bands such as 450nm (detecting surface carbonization), 970nm (moisture absorption peak), and 1450nm (OH group characteristic peak); the thermal field excitation used a power density of 500W / m², a heating time of 5 seconds, and a cooling monitoring time of 60 seconds.

[0017] In the above embodiment, the UAV, following a preset gridded scanning path, set 48 hovering points at a distance of 3 meters from the building facade, covering a detection area of ​​approximately 300 square meters on the entire south facade. At hovering point 23 (corresponding to the area between the windows on the west side of the 4th floor of the building, numbered S4-W3), the following raw signals were collected: the acoustic response showed an abnormal resonance peak at 25.3 kHz, with the peak attenuation coefficient increasing by 15% compared to adjacent areas; the reflectivity at 970 nm was 8% lower than the baseline value, indicating the presence of moisture accumulation in this area; the thermal response curve showed that the cooling time constant was 1.3 times the standard value, suggesting the possible presence of voids or stratification inside. After time-frequency domain transformation, the multiphysics feature vector of the detection point was extracted, including 12 acoustic domain parameters (resonance frequency 25.3kHz, Q value 18.5, energy distribution entropy 0.76, etc.), 10 spectral domain parameters (970nm reflectance 0.42, first derivative extremum at 985nm, second derivative zero-crossing density 3.2 / 100nm, etc.), and 8 thermal domain parameters (thermal diffusivity 0.85×10⁻). 6 The parameters include m² / s, surface heat transfer coefficient of 12.5 W / (m²·K), and thermal resistance gradient of 0.15 K·m² / W. Relevant information was retrieved from a pre-constructed knowledge graph of degradation evolution. This knowledge graph, built upon historical testing data and material test results of similar buildings, includes physical field response feature templates for 12 typical degradation modes, such as carbonization degradation, moisture erosion, interface debonding, and freeze-thaw damage. After similarity matching calculations, the S4-W3 region showed a comprehensive similarity of 0.82 with the early hollowing mode of the mortar layer and 0.75 with the moisture erosion density reduction mode, both exceeding the preset threshold of 0.7, indicating a risk of complex degradation in this region.

[0018] In the above embodiment, to assess the degradation trend, historical data from the past 6 months (M=6) for this detection point were retrieved. Time-series analysis showed that the resonant peak frequency shifted to lower frequencies at a rate of 0.8% per month, indicating a continuous degradation of material stiffness; the 970nm reflectivity showed an accelerated downward trend, with the average monthly decrease increasing from 1% initially to the current 2.5%; and the degradation rate of the thermal diffusivity reached 3% per month. Combining the evolution rates of the three physical domains, and weighted according to the weights corresponding to the hollow deterioration mode (acoustic 0.5, spectral 0.2, thermal 0.3), the degradation development rate index was calculated to be 2.3% / month. Because the rate of degradation exceeded the preset threshold of 1.5% / month, an early warning of deterioration was generated: location coordinates (facade height 12.5m, horizontal position 18.3m), degradation type: mortar layer hollowing accompanied by moisture erosion, current development rate level: medium speed, expected to develop into visible hollowing within 8-10 months, recommendation to implement grouting reinforcement measures within 3 months, and waterproofing treatment above the area. The warning information was pushed to the building maintenance management platform in real time via wireless network, displaying the risk area in orange at the corresponding location in the 3D building model, and generating an assessment report containing detailed detection data, degradation mechanism analysis, and maintenance recommendations.

[0019] Through the above steps, three heterogeneous sensor groups mounted on a UAV simultaneously collect response signals from the target building facade in different physical fields. The acoustic field sensing sensor group acquires the excitation acoustic response signal, which reflects changes in the internal microstructure of the material; the spectral sensing sensor group acquires the selected band spectral reflectance signal, which reveals changes in surface chemical composition; and the thermal field sensing sensor group records the transient thermal conduction response signal, which characterizes the degradation of the material's thermophysical properties. These three signals complement and verify each other. Time-frequency domain transformation processing converts the original signals into feature vectors, allowing different physical quantities to be expressed in a unified mathematical space. A degradation evolution knowledge graph jointly constructed from material property parameters and historical environmental data provides prior knowledge support. Similarity matching between multi-physics feature vectors and physical field response feature templates enables accurate identification of degradation patterns. Temporal correlation analysis captures the evolutionary laws of features, and threshold judgment of the degradation development rate index triggers an early warning mechanism. The combination of multi-physics collaborative sensing and knowledge-driven pattern recognition allows the time window for degradation detection to be moved forward to the early stage of damage.

[0020] The entity performing the above steps can be a system, such as a building damage early detection system, or equipment, such as building damage early detection equipment, or a controller or processor in the equipment or system, or a separate controller or processor, or other processing equipment or processing units with similar processing functions, but is not limited to these.

[0021] In an optional embodiment, the acoustic response signal, selected band spectral reflection signal, and transient thermal conduction response signal of the target building facade are collected by an acoustic field sensing sensor group, a spectral sensing sensor group, and a thermal field sensing sensor group carried by the UAV. Specifically, this includes: determining multi-physics field excitation parameters based on the material type, construction year, and environmental exposure conditions of the target building facade; the multi-physics field excitation parameters include the acoustic field excitation frequency band range, the spectral detection band combination, and the thermal field excitation power density; and setting up a UAV hovering reference point array at a predetermined distance outside the target building facade, so that the UAV hovers at each of the hovering reference point arrays. The detection field of view at the hovering reference point covers a pre-defined detection unit on the facade of the target building. The UAV is controlled to hover sequentially at each of the hovering reference points in the array. At each hovering reference point, a sweeping acoustic wave signal is emitted to the pre-defined detection unit covered by the current hovering reference point via an acoustic field sensing sensor array. Simultaneously, reflected and transmitted acoustic wave signals from the current pre-defined detection unit under the excitation of the sweeping acoustic wave signal are acquired. These reflected and transmitted acoustic wave signals are used as excitation acoustic response signals. The frequency range of the sweeping acoustic wave signal covers the acoustic field excitation frequency band. A preset time interval is elapsed after the excitation acoustic response signal acquisition is completed. Inside the enclosure, the spectral sensing sensor group sequentially performs multi-band spectral scanning on the current preset detection unit according to the spectral detection band combination. This collects the reflectance spectral data of the current preset detection unit in the ultraviolet, visible, and near-infrared bands, and uses the reflectance spectral data as the selected band spectral reflectance signal. The ultraviolet band is used to detect changes in scattering characteristics caused by microcracks on the surface of the current preset detection unit, the visible band is used to identify abnormal color difference areas on the surface of the current preset detection unit, and the near-infrared band is used to detect abnormal moisture distribution inside the current preset detection unit. After completing the acquisition of the selected band spectral reflectance signal, the thermal field sensing sensor... The sensor array applies pulsed thermal excitation to the current preset detection unit, and simultaneously collects the temporal evolution data of the surface temperature field of the current preset detection unit during the heating and cooling stages. The temporal evolution data of the surface temperature field is used as the transient thermal conduction response signal. The excitation acoustic response signal, the selected band spectral reflection signal and the transient thermal conduction response signal collected at the same hovering reference point are timestamped to obtain a multiphysics signal group with time stamps. The three-dimensional spatial coordinates of the corresponding hovering reference point and the number information of the corresponding preset detection unit are added to the multiphysics signal group to generate a multiphysics original dataset with spatiotemporal consistency.

[0022] Among them, multi-physics excitation parameters refer to the configuration of physical field parameters used to excite building materials to produce a measurable response, including the setting of key excitation parameters such as sound field, light field, and thermal field; the sound field excitation frequency band range refers to the frequency range used for sound wave detection, which is determined according to the material type and the expected detection depth. For example, the ultrasonic frequency band of 20kHz-100kHz is commonly used for concrete structures; the spectral detection band combination represents the set of wavelength ranges for spectral scanning, which usually includes a combination of multiple bands such as ultraviolet (200-400nm), visible light (400-700nm), and near-infrared (700-2500nm); the thermal field excitation power density refers to the intensity of thermal excitation energy applied to the material surface, with the unit of W / m², which needs to be determined according to the material's heat capacity and... The required detection depth is defined; the material type indicates the type of materials used in the building facade, such as concrete, brick, and metal sheeting; the construction year refers to the time of the building's construction or most recent major overhaul, used to estimate the service life of the materials; the environmental exposure conditions indicate the type of environment in which the building is located, such as coastal high-salt-fog environments, industrial pollution environments, and high-altitude freeze-thaw environments; the hovering reference point array refers to the set of UAV hovering positions arranged according to rules on the outer side of the building facade, used to ensure the integrity and consistency of the detection coverage; the detection field of view indicates the spatial range that the sensor can effectively detect at a specific hovering position, which depends on the sensor's viewing angle and distance; the preset detection unit refers to the standardized detection grid unit into which the building facade is divided, such as a 1m×1m square. The terms "region" and "frequency sweep" refer to different types of acoustic signals. Frequency sweep signals are acoustic excitation signals with continuously varying frequencies, used to detect the frequency response characteristics of materials over a wide frequency range. Reflected acoustic signals represent the signals reflected back from the material surface, carrying surface and shallow layer information. Transmitted acoustic signals are acoustic signals received after penetrating the material, carrying information about the material's internal structure. Multi-band spectral scanning refers to the process of sequentially performing spectral measurements within multiple different wavelength ranges. The ultraviolet band refers to the electromagnetic wave range of 200-400 nm, sensitive to scattering phenomena caused by microcracks on the material surface. The visible light band refers to the electromagnetic wave range of 400-700 nm, used to detect color difference changes perceptible to the human eye. The near-infrared band refers to the electromagnetic wave range of 700-2500 nm. Wavelength range, capable of penetrating to a certain depth to detect internal moisture distribution; changes in scattering characteristics indicate changes in the scattering pattern of light on the material surface as the material deteriorates; anomalous color regions refer to areas with significant color differences from the surrounding areas, which may indicate changes in chemical composition or contamination; abnormal moisture distribution indicates non-uniform distribution of moisture content within the material, which may indicate the development of cracks or pores; pulsed thermal excitation refers to a transient thermal load applied for a short period of time, typically lasting from a few seconds to tens of seconds; the heating phase represents the time period during which thermal excitation is applied, during which the material absorbs heat and its temperature rises; the cooling phase refers to the time period during which the material cools naturally after the thermal excitation is stopped, during which the temperature gradually recovers; the temporal evolution data of the surface temperature field represents a continuous record of the changes in the material's surface temperature distribution over time;Timestamp annotation refers to adding precise data acquisition time markers for time-series correlation analysis; multiphysics signal sets represent the collection of acoustic, optical, and thermal signals acquired at the same spatiotemporal location; three-dimensional spatial coordinates refer to the X, Y, and Z position information of the hovering point in the building coordinate system; numbering information represents the unique identifier of the detection unit; spatiotemporal consistency refers to the correspondence between data in time and space dimensions, ensuring that data from different physics fields can be correlated; the multiphysics raw dataset represents the collection of all multiphysics detection data with completed spatiotemporal annotation.

[0023] In the above embodiment, facade deterioration detection was conducted on an office building in an industrial park that had been in service for 25 years. The building is a frame structure with an exterior facade made of fair-faced concrete coated with an anti-carbonation paint. It is 32 meters high and has a total facade area of ​​approximately 2400 square meters. Because the area has a temperate monsoon climate with significant annual temperature variations and acid rain erosion, a detailed early-stage deterioration detection of the building facade is necessary. First, the multiphysics excitation parameters were determined based on the specific characteristics of the building. Considering the material properties of fair-faced concrete (C30 concrete, designed protective layer thickness 25mm) and the environmental exposure conditions of the area (annual average acid rain pH value 5.2, winter-summer temperature difference of up to 45℃, annual average relative humidity 70%), the following parameters were automatically configured: the acoustic field excitation frequency band was set to 15-80kHz, with a focus on the 25-45kHz range to detect microcracks and carbonation fronts within the concrete; the spectral detection band combination selected five key bands: 380nm, 550nm, 970nm, 1450nm, and 2200nm, corresponding to surface microcrack scattering, carbonation indication, moisture absorption, hydroxide ion and carbonate characteristics, respectively; the thermal field excitation power density was set to 600W / m², with continuous heating for 8 seconds to ensure that the heat wave could penetrate to the depth of the protective layer. At a distance of 5 meters from the outer side of the east facade of the building, 120 hovering reference points were set at a grid spacing of 2m × 2m, forming a detection array covering the entire facade. Each hovering point corresponds to a detection field of view of 2.5 meters × 2.5 meters, with a 0.5-meter overlap between adjacent detection units to ensure no blind spots. The UAV is equipped with a centimeter-level positioning system, and the hovering accuracy is controlled within ±10cm.

[0024] In the above embodiment, the UAV begins its operation from the first hovering point in the lower left corner. At hovering point E06-05 (corresponding to the central area of ​​the 3rd floor of the building), the following detection process is performed: First, the acoustic field sensing sensor group emits a linear sweep frequency signal of 15-80kHz with a sweep period of 100ms, repeated 10 times and averaged to improve the signal-to-noise ratio. The transmitting transducer power is set to 50W, and the receiving array synchronously records reflected and transmitted signals. The acquired acoustic response shows an abnormal resonance peak at 32.5kHz, with a peak amplitude 12dB lower than the adjacent area and a frequency shift of 0.15kHz; a new resonant mode is detected at 48kHz, which may indicate the presence of internal layering or voids. The time-domain envelope of the signal shows that the arrival time of the first wave is delayed by 0.3ms, indicating a decrease in sound velocity of approximately 5%. After completing the acoustic acquisition, wait 2 seconds for the material vibration to completely decay, and then activate the spectral sensing sensor group. The hyperspectral camera sequentially imaged in five preset bands, with the exposure time for each band automatically adjusted according to ambient light (500ms for 380nm, 200ms for 550nm, 300ms for 970nm, 400ms for 1450nm, and 600ms for 2200nm). Spectral data showed that the scattering intensity in the 380nm band increased by 18% compared to the reference value, indicating an increased density of surface microcracks; the reflectance in the 550nm band decreased by 6%, suggesting possible carbonization; a significant absorption peak appeared at 970nm, with a reflectance of only 0.38, indicating a high moisture content in this region; the intensity of the characteristic peak of the OH group at 1450nm increased by 15%; and the characteristic peak of the carbonate at 2200nm weakened by 8%.

[0025] In the above embodiment, after spectral acquisition, the thermal field sensing sensor group begins operation. A halogen lamp array heats the detection area for 8 seconds at a power density of 600 W / m², while an infrared thermal imager continuously records temperature field changes at a frame rate of 30 Hz. After heating, a 90-second cooling process is recorded. Temperature data shows that the highest temperature rise during the heating phase reached 8.2 K, 1.5 K higher than adjacent areas; the temperature rise time constant was 3.8 seconds, 0.6 seconds slower than normal; the temperature decay during the cooling phase exhibited a double-exponential characteristic, with a fast decay component time constant of 2.5 seconds and a slow decay component time constant of 18 seconds, with the slow component accounting for 35%, significantly higher than the normal 20%; spatially, a low-temperature anomaly zone with a diameter of approximately 30 cm was detected, with a temperature difference of 1.2 K. After acquiring the signals from the three physical fields, a unified timestamp (accurate to milliseconds for the start time of acquisition) was added to the data set, along with spatial information annotations: the three-dimensional coordinates of the hovering point (X=10.0m, Y=10.0m, Z=6.0m), corresponding to the detection unit number E-3F-C05, and the location of the center point in the building coordinate system (east facade, 7.5m above ground, 18.0m from the north end). The drone sequentially completed data acquisition for all 120 hovering points according to a preset path, taking approximately 4 hours in total. Throughout the process, data quality was monitored in real time, and data with a signal-to-noise ratio below the threshold were automatically marked and reacquired. The final multiphysics raw dataset includes: 360 acoustic signal files (3 directions per hovering point), 600 spectral image files (5 bands per hovering point), 120 thermal video files, and accompanying spatiotemporal annotation metadata files. The total data size is approximately 48GB, and it was transmitted in real time to the cloud processing platform via a 5G network. In the E06-05 inspection unit, subsequent time-frequency domain transformation and feature extraction identified a composite degradation mode of concrete carbonation accompanied by microcrack development. The degradation depth was estimated at 8-10 mm, approaching 40% of the protective layer thickness. Based on historical data analysis, the degradation rate in this area is 1.8 mm / year, and it is projected to pose a risk of steel reinforcement corrosion within 3-4 years. The resulting maintenance recommendations include: immediate repair of the anti-carbonation coating and installation of long-term monitoring sensors in the area, with quarterly re-inspections.

[0026] In an optional embodiment, the excitation acoustic response signal, the selected band spectral reflection signal, and the transient thermal conduction response signal are subjected to time-frequency domain transformation processing to obtain a multi-physics feature vector. Specifically, this includes: performing short-time Fourier transform processing on the excitation acoustic response signal in the original multi-physics dataset to obtain a time-spectrum matrix; extracting the resonant peak frequency position, resonant peak amplitude attenuation coefficient, and spectral energy distribution entropy from the time-spectrum matrix as acoustic domain feature parameters. The resonant peak frequency position characterizes the material stiffness degradation characteristics of the target building facade; the resonant peak amplitude attenuation coefficient characterizes the material damping loss evolution characteristics of the target building facade; and the spectral energy distribution entropy characterizes the discrete distribution characteristics of internal material damage in the target building facade. The selected band spectral reflection signal in the original multi-physics dataset is subjected to continuous wavelet transform processing to obtain a spectral reflectance curve; extracting spectral reflectance eigenvalues, the position of the first derivative extremum point of the spectral curve, and the density of zero crossover points of the second derivative from the spectral reflectance curve as spectral domain feature parameters. The spectral reflectance eigenvalues ​​characterize the material stiffness degradation characteristics of the target building facade; and the spectral reflectance eigenvalues ​​characterize the material stiffness degradation characteristics of the target building facade. The weathering characteristics of the target building facade's material surface are analyzed. The location of the extreme points of the first derivative of the spectral curve represents the variation characteristics of the chemical composition of the target building facade's material. The density of zero-crossing points of the second derivative represents the complexity of the microstructure of the target building facade's material surface. The transient thermal conduction response signal in the original multiphysics dataset is processed by Laplace transform to obtain the complex frequency domain temperature transfer function. The thermal diffusion time constant, surface heat transfer coefficient, and internal thermal resistance distribution gradient are extracted from the complex frequency domain temperature transfer function as thermal domain feature parameters. The thermal diffusion time constant represents the degradation characteristics of the target building facade's material thermal conductivity, the surface heat transfer coefficient represents the variation characteristics of the target building facade's material surface density, and the internal thermal resistance distribution gradient represents the characteristics of the internal layering defects of the target building facade's material. The acoustic domain feature parameters, spectral domain feature parameters, and thermal domain feature parameters are arranged and combined according to a preset feature order to generate an initial multiphysics feature vector. Cross-physics coupling strength analysis is performed on the initial multiphysics feature vector to obtain the multiphysics feature vector.

[0027] Among them, the short-time Fourier transform is a time-frequency analysis method that performs piecewise Fourier transform on the signal through a sliding time window, enabling the simultaneous acquisition of the signal's time and frequency information; the time-frequency spectrum matrix represents the distribution of signal energy in a two-dimensional time-frequency plane, with time on the horizontal axis and frequency on the vertical axis, and the matrix element values ​​representing the energy intensity at the corresponding time-frequency point; the resonant peak frequency position refers to the frequency point where energy is concentrated in the time-frequency spectrum, corresponding to the material's natural frequency, and a decrease in material stiffness will cause the resonant peak to shift to a lower frequency; the resonant peak amplitude attenuation coefficient represents the rate attenuation of the resonant peak energy over time, reflecting the material's damping characteristics, and increased damping in deteriorated materials leads to faster attenuation; the spectral energy distribution entropy is a measure of the uniformity of the spectral energy distribution, and the energy concentration of healthy materials is higher than that of unhealthy materials. Low entropy indicates increased energy dispersion and higher entropy due to increased internal damage; acoustic domain characteristic parameters represent the set of parameters extracted from the acoustic response signal to characterize the acoustic properties of the material; material stiffness degradation characteristics refer to the trend of decreasing elastic modulus with deterioration, reflected by changes in resonant frequency; material damping loss evolution characteristics represent the changing law of energy dissipation capacity due to internal friction in the material, with deterioration leading to increased damping due to porous microstructure; discrete distribution characteristics of internal damage in the material refer to the degree of dispersion of defects such as cracks and pores within the material, with more dispersed defects resulting in more severe energy scattering; continuous wavelet transform refers to the method of multi-scale decomposition of signals using scalable wavelet basis functions, which can capture local singular features of the signal; spectral reflectance curves represent the material's response to different wavelengths of light. The reflectance of long wavelengths varies with wavelength; the characteristic value of spectral reflectance refers to the reflectance value at a specific wavelength or the average reflectance of a characteristic band; material weathering leads to an overall decrease in reflectance; the position of the extreme point of the first derivative of the spectral curve indicates the wavelength position where the slope of the reflectance curve is the largest or smallest, corresponding to the absorption edge or characteristic peak of the material; changes in chemical composition cause the displacement of the extreme point; the density of zero-crossing points of the second derivative refers to the number of times the curvature of the reflectance curve changes sign within a unit wavelength range; increased surface roughness leads to more curve fluctuations and more zero-crossing points; spectral domain characteristic parameters represent the set of parameters extracted from the spectral reflectance signal that characterize the optical properties of the material; the degree of weathering of the material surface refers to the degree of physicochemical changes that occur on the material surface due to environmental factors. The changes in overall reflectivity reflect changes in the material's chemical composition; for example, carbonization leads to an increase in calcium carbonate, which is reflected by the shift of characteristic absorption peaks; the microscopic morphology of the material's surface refers to the surface roughness and texture complexity, which is reflected by the fine structure of the spectral curve; the Laplace transform is an integral transform method that transforms a time-domain function to the complex frequency domain, often used to analyze the transfer characteristics of a system; the complex frequency domain temperature transfer function represents the ratio of the material's surface temperature response to the thermal excitation input in the complex frequency domain, reflecting the material's heat transfer kinetics; the thermal diffusion time constant is the characteristic time required for a material to reach thermal equilibrium, inversely proportional to the material's thermal diffusivity, and degradation leads to slower thermal diffusion and an increased time constant;The surface heat transfer coefficient represents the convective heat transfer capacity between the material surface and the air; decreased surface density leads to an increase in the heat transfer coefficient. The internal thermal resistance distribution gradient refers to the rate of change of thermal resistance at different depths within the material; the gradient changes abruptly when delamination defects exist. Thermal domain characteristic parameters represent the set of parameters characterizing the material's thermal properties extracted from the thermal conduction response signal. Material thermal conductivity degradation characteristics refer to the trend of decreasing thermal conductivity with degradation; increased porosity leads to a decrease in thermal conductivity. Surface density variation characteristics represent changes in the pore structure of the material surface; weathering and erosion lead to surface porosity. Internal delamination defect characteristics refer to defects such as interlayer debonding and voids within the material, hindering heat conduction and forming thermal resistance interfaces. The preset feature order represents the arrangement rule when combining characteristic parameters from different physical domains into a vector; consistency is required for subsequent processing. The initial multiphysics feature vector refers to feature vectors arranged in order but without considering interdomain coupling. Cross-physics coupling strength analysis represents the process of analyzing the correlation between features from different physical domains; for example, acoustic and thermal features may be correlated due to internal defects.

[0028] In the above embodiments, the raw multiphysics data collected from the aforementioned industrial park office building underwent in-depth time-frequency domain transformation processing to extract characteristic parameters that can accurately characterize the material degradation state. For the acoustic response signal of detection unit E06-05, preprocessing was first performed, including 50Hz notch filtering to remove power frequency interference, bandpass filtering to retain the 15-80kHz effective frequency band, and noise reduction based on wavelet thresholding. Then, short-time Fourier transform was used for time-frequency analysis, with a Hanning window function set, a window length of 1024 points (corresponding to a 4ms time resolution at a sampling rate of 256kHz), and an overlap rate of 75%. The generated time-spectrum matrix showed a continuous energy concentration band at 32.5kHz within the 0-50ms time axis range, forming a distinct resonance peak structure. Using a peak detection algorithm, the resonance peak frequency was accurately located at 32.48kHz, a 2.2% decrease compared to the standard value of 33.2kHz for this material type, indicating a decrease in the concrete's elastic modulus of approximately 4.3%. The 3dB bandwidth of the resonance peak is 1.85kHz, and the calculated quality factor Q = 17.6 corresponds to an amplitude attenuation coefficient of 0.089 / ms, which is 37% higher than the normal value of 0.065 / ms, reflecting the increased damping caused by the increase in microcracks inside the material. Shannon entropy calculation of the entire spectrum yields a spectral energy distribution entropy value of 0.82, significantly higher than the 0.65 of healthy concrete, indicating an increased dispersion of energy in the frequency domain and a discrete distribution of internal damage.

[0029] In the above embodiments, for the spectral reflectance signal, Morlet wavelet was used as the mother wavelet for continuous wavelet transform, with the scale parameter range set to 2-128, corresponding to an analysis frequency range of 0.5-64 periods / wavelength. Near the 970nm moisture absorption peak, the wavelet coefficients showed obvious singularities, with reflectance decreasing from the baseline of 0.52 to 0.38, a decrease of 26.9%. After obtaining a smooth spectral reflectance curve through spline interpolation, its first derivative was calculated. It was found that the derivative maximum point, originally located at 985nm, shifted to 992nm, a shift of 7nm. This redshift phenomenon indicates a change in the internal hydrogen bonding environment of the material, possibly due to chemical structural changes caused by moisture infiltration. The second derivative of the reflectance curve was calculated, and zero-crossing points were statistically analyzed in the 400-700nm visible light range, with a density of 4.3 per 100nm, which is 1.8 times that of a normal surface. This quantitatively reflects a significant increase in surface micro-roughness due to weathering. Particularly around 550 nm, the reflectivity characteristic value is 0.42. Considering the sensitivity of this wavelength to carbonization products, the carbonization depth is inferred to be approximately 8-10 mm. The transient thermal conduction response signal was processed using the Laplace transform technique. First, the temperature-time curve was subjected to exponential fitting to separate the fast surface response and the slow volume response components. Data from the temperature rise phase showed: T(t) = T∞[1 - 0.65exp(-t / τ1) - 0.35exp(-t / τ2)], where τ1 = 2.3 s is the fast time constant, and τ2 = 8.7 s is the slow time constant. After the Laplace transform, in the complex frequency domain s = σ + jω plane, the temperature transfer function G(s) = ΔT(s) / Q(s) exhibits two distinct poles, located at s = -0.43 and s = -0.115, respectively. The thermal diffusion time constant τ was calculated from the pole locations. d =5.8s, 38% higher than the 4.2s of normal concrete, corresponding to a thermal diffusivity α = 0.62 × 10^-6 m² / s, indicating a significant degradation in the material's thermal conductivity. The surface heat transfer coefficient h = 14.2 W / (m²·K) extracted from the slope of the high-frequency asymptote of the transfer function is higher than the 10⁻¹² W / (m²·K) of dense concrete, indicating enhanced convective heat transfer due to increased surface porosity. By inverting the temperature response at different depths and calculating the internal thermal resistance distribution, a sudden change in the thermal resistance gradient was detected at a depth of 12 mm, with a gradient value of 0.18 K·m² / W·mm, three times that of homogeneous materials, strongly indicating the presence of delamination or void defects at this depth.

[0030] In the above embodiment, the extracted acoustic domain features (resonance frequency 32.48kHz, Q value 17.6, attenuation coefficient 0.089 / ms, spectral entropy 0.82, secondary resonance 48.3kHz, peak spacing 15.82kHz, low-frequency energy ratio 0.23, high-frequency attenuation rate -3.2dB / 10kHz, phase delay 2.8ms) and spectral domain features (970nm reflectance 0.38, 550nm reflectance 0.42, first derivative extremum 992nm, second derivative zero crossover density 4.3 / The initial eigenvectors were formed by arranging the acoustic, spectral, and thermal parameters (100 nm, spectral slope -0.0008 / nm, red edge position 712 nm, vegetation index 0.15, moisture index 1.35) and seven characteristic parameters (thermal diffusion time constant 5.8 s, surface heat transfer coefficient 14.2 W / (m²·K), thermal resistance gradient 0.18 K·m² / W·mm, peak temperature rise 8.2 K, cooling half-life 12.3 s, thermal inertia index 850 J / (m²·K·s^0.5), phase lag 18°) in the order of acoustic-spectral-thermal. To capture the coupling effect between different physical fields, cross-physical field coupling strength analysis was performed. By calculating the mutual information matrix between the features, it was found that the mutual information value between the acoustic attenuation coefficient and the thermal diffusion time constant reached 0.73, indicating a high correlation between the two, jointly reflecting changes in the internal pore structure of the material. The correlation coefficient between the spectral absorption at 970 nm and the thermal surface heat transfer coefficient was 0.68, revealing that moisture content affects both optical and thermal properties. Based on principal component analysis, three coupled features were extracted: the first principal component (contribution rate 42%) mainly consists of acoustic Q value, thermal diffusivity constant, and spectral reflectivity, representing the overall degree of material degradation; the second principal component (contribution rate 28%) is dominated by spectral entropy, second derivative density, and thermal resistance gradient, reflecting defect distribution characteristics; and the third principal component (contribution rate 18%) is strongly correlated with moisture-related features.

[0031] In the above embodiment, after constructing the initial 24-dimensional feature vector, an in-depth cross-physics coupling strength analysis is performed to uncover the intrinsic correlations between features from different physical domains and construct a more representative multi-physics feature vector. First, a 24×24 feature correlation matrix is ​​constructed. Pearson correlation coefficient, Spearman rank correlation coefficient, and mutual information are used to comprehensively evaluate the linear and nonlinear correlations between features. The analysis revealed several significant cross-domain coupling relationships: the acoustic resonance peak frequency shift (-0.72 kHz) showed a strong positive correlation with the thermal diffusion time constant increment (+1.6 s) (r=0.81), indicating that the increased microcracks and porosity within the material simultaneously reduced the elastic modulus and thermal conductivity; the mutual information between the spectral absorption depth at 970 nm (reflectivity reduction of 26.9%) and the acoustic spectral energy distribution entropy (0.82) reached 0.76 bits, suggesting that moisture intrusion not only altered optical properties but also affected sound wave propagation by changing material homogeneity; the correlation coefficient between the internal thermal resistance gradient (0.18 K·m² / W·mm) and the acoustic attenuation coefficient (0.089 / ms) was 0.73, both pointing to potential interface defects at a depth of 12 mm. Based on the correlation analysis results, kernel principal component analysis (KPCA) was used to extract nonlinear coupling features. A radial basis function (RBF) kernel was selected, and the kernel parameter γ was determined to be 0.05 through cross-validation. After mapping to a high-dimensional feature space, the first five principal components were extracted, and the cumulative variance contribution rate reached 87.3%. In the above embodiments, the first coupling principal component (contribution rate 38.5%) is composed of: PC1 = 0.42 × (acoustic Q value normalization) + 0.38 × (thermal diffusion time constant normalization) + 0.35 × (reciprocal of 970nm reflectance) + 0.28 × (spectral entropy) + ... This component comprehensively reflects the overall degree of material density degradation; the larger the value, the more severe the deterioration. In the E06-05 detection unit, the PC1 value is 2.83, significantly higher than the benchmark value of 1.0 for healthy concrete. The second coupling principal component (contribution rate 24.6%) is mainly composed of defect distribution-related characteristics: PC2 = 0.45 × (second derivative zero cross density) + 0.41 × (thermal resistance gradient) + 0.38 × (acoustic phase delay) + 0.31 × (spectral slope change rate) + ..., reflecting the spatial dispersion and non-uniformity of defects. The PC2 value of the E06-05 unit is 1.92, indicating the existence of locally concentrated defect regions. The third principal component of the coupling (contributing 15.2%) is related to chemical degradation: PC3 = 0.48 × (550nm carbonization-indicating reflectance) + 0.43 × (first derivative extremum shift) + 0.36 × (surface heat transfer coefficient change) + 0.29 × (low-frequency acoustic energy ratio) + ..., mainly characterizing the degree of carbonization and chemical erosion. This unit has a PC3 value of 1.65, indicating moderate carbonization. Conditional mutual information was calculated using information theory to identify higher-order coupling relationships between features. A ternary coupling mode was discovered: when the acoustic Q value is below 18 and the thermal diffusion time constant exceeds 5.5 s, the conditional probability of a decrease in 970nm reflectance reaches 0.85. This ternary coupling strongly indicates complex degradation caused by moisture intrusion. Based on this discovery, four ternary coupling features were constructed: TC1 = (acoustic Q value) × (thermal diffusion time constant) / (970nm reflectivity), with a value of 251.3, far exceeding the health threshold of 100; TC2 = exp[-(spectral entropy-0.6)] × (thermal resistance gradient) × (second derivative density), with a value of 0.92; TC3 and TC4 respectively capture the interaction effects of other physical quantities.

[0032] In the above embodiments, a graph neural network method is also used to model the topological relationships between features. The 24 original features are used as nodes, and edge connections are established between feature pairs with a correlation coefficient greater than 0.5 to form a feature association graph. Through a two-layer graph convolution operation, each node aggregates neighbor node information to generate a graph embedding vector containing the local topological structure. The feature map of cell E06-05 shows three tightly connected subgraphs: an acoustic-thermal coupling subgraph (6 nodes, 9 edges), a spectral feature subgraph (5 nodes, 7 edges), and an interface defect correlation subgraph (4 nodes, 5 edges). Through graph pooling, the graph structure information is compressed into an 8-dimensional embedding vector. To verify the physical meaning of the coupled features, sensitivity analysis was performed. By calculating the partial derivatives of each coupled feature with respect to the degree of degradation, it was found that PC1 is sensitive to concrete strength loss at -3.2 MPa / unit PC1 value, and TC1 is sensitive to the permeability coefficient at 0.8 × 10^-12 m / s / unit TC1 value. These quantitative relationships establish a mapping between the feature space and the physical performance space. Finally, the original 24-dimensional feature vector, 5 kernel principal components, 4 ternary coupling features, and 8-dimensional graph embedding vector are combined to form a 41-dimensional enhanced multiphysics feature vector. Through the recursive feature elimination (RFE) algorithm, redundant features are removed, retaining 27 of the most discriminative features to form the final multiphysics feature vector: [6 core features in the acoustic domain, 5 core features in the spectral domain, 4 core features in the thermal domain, 3 coupled principal components, 3 ternary coupling features, and 6 graph embedding features]. This multi-level cross-physics coupling analysis not only preserves the independent information of each physical domain but, more importantly, captures the synergistic effects between different physical mechanisms during the degradation process.

[0033] In an optional embodiment, material property parameters and historical environmental impact data of the target building facade are obtained, and a degradation evolution knowledge graph is constructed based on the material property parameters and historical environmental impact data. The method further includes: obtaining material composition information, mechanical performance indicators, and microstructure parameters of the target building facade from a building archive database as material property parameters; and obtaining historical temperature change data, humidity cycle data, acid rain erosion data, freeze-thaw cycle count, and cumulative ultraviolet radiation in the area where the target building facade is located from an environmental monitoring station as historical environmental impact data; and establishing a set of material degradation response functions based on the material property parameters. The set of material degradation response functions includes the functional relationship between carbonization depth and carbonization sensitivity coefficient, the functional relationship between chloride ion diffusion depth and chloride ion penetration sensitivity coefficient, and the functional relationship between freeze-thaw damage degree and freeze-thaw damage sensitivity coefficient. The study establishes a functional relationship between the cumulative fatigue damage and the fatigue cumulative damage sensitivity coefficient. Based on the material degradation response function set and historical environmental data, it determines the cumulative carbonization driving force, chloride ion erosion driving force, freeze-thaw cycle driving force, and fatigue load driving force experienced by the target building facade. Weighted fusion of these driving forces yields the cumulative degradation driving force of the target building facade. A degradation pattern clustering analysis is performed on the original multi-physics dataset to obtain the degradation pattern classification results and the distribution range of physical field characteristic parameters corresponding to each degradation pattern at different degradation stages. Finally, a degradation evolution knowledge graph is constructed based on material property parameters, cumulative degradation driving force, degradation pattern classification results, and the distribution range of physical field characteristic parameters.

[0034] Among them, the building archive database refers to an information system that stores historical archives of building design, construction, and materials, such as a Building Information Modeling (BIM) database and a construction drawing database; material composition information refers to the chemical components and proportions of building materials, such as the types and proportions of cement, aggregates, and admixtures in concrete; mechanical performance indicators represent mechanical parameters such as strength, modulus of elasticity, and toughness of materials, such as compressive strength and splitting tensile strength; microstructure parameters refer to the structural characteristics of materials at the microscale, such as porosity, pore size distribution, and grain size; environmental monitoring stations refer to observation facilities that monitor environmental parameters over a long period of time, such as meteorological stations and air quality monitoring stations; historical temperature change data refers to temperature records of the area where the target building is located over many years. This includes daily average temperature, extreme temperatures, etc.; humidity cycle data represents the periodic changes in relative humidity, reflecting the number of times the material experiences wet-dry cycles; acid rain erosion data refers to the pH value and concentration of acidic components in rainfall, reflecting the intensity of chemical erosion; freeze-thaw cycle count represents the number of times the material experiences freezing and thawing due to temperature fluctuations around the freezing point, which is a major deterioration driver in cold regions; cumulative ultraviolet radiation refers to the total ultraviolet radiation energy received by the material surface, leading to aging of organic materials and weathering of inorganic material surfaces; the set of material deterioration response functions represents a set of mathematical functions describing the performance degradation law of materials under different deterioration mechanisms, establishing a quantitative relationship between deterioration driving force and deterioration degree; carbonization depth refers to the reaction between carbon dioxide in the air and cement hydration products. The depth to which chloride ions penetrate into the material leads to a decrease in concrete alkalinity; the carbonation sensitivity coefficient indicates the material's sensitivity to carbonation and is related to the material's porosity and alkalinity reserve; chloride ion diffusion depth refers to the depth to which external chloride ions (such as sea salt or de-icing agents) penetrate into the material, potentially causing steel corrosion; the chloride ion penetration sensitivity coefficient indicates the material's resistance to chloride ion intrusion and is related to the material's density and protective layer thickness; freeze-thaw damage degree refers to the cumulative damage caused by repeated freezing and swelling of moisture, manifested as surface spalling and internal cracks; the freeze-thaw damage sensitivity coefficient indicates the material's resistance to freeze-thaw cycles and is related to the material's saturation and pore structure; cumulative fatigue damage refers to the cumulative damage of the material under cyclic loading, following the principle of fatigue accumulation. The cumulative criterion; the fatigue cumulative damage sensitivity coefficient represents the sensitivity of a material to cyclic loading, and is related to the material's strength reserve and stress amplitude; the cumulative carbonization driving force represents the total carbonization effect caused by carbon dioxide concentration and humidity conditions during the building's service life; the cumulative chloride ion erosion driving force refers to the total erosion effect caused by chloride ion exposure concentration and time; the cumulative freeze-thaw cycle driving force represents the total damage driven by the number and intensity of freeze-thaw cycles experienced; the cumulative fatigue load driving force refers to the total fatigue driving force caused by the amplitude and number of cyclic loads; weighted fusion represents a calculation method that linearly combines multiple quantities according to different weighting coefficients; the cumulative degradation driving force refers to the total amount of driving forces from various degradation mechanisms, reflecting the overall degradation intensity experienced by the building;Degradation pattern clustering analysis refers to the unsupervised classification of multi-physics data using clustering algorithms to automatically identify degradation pattern categories within the data. The degradation pattern classification results represent the degradation type division obtained through clustering, such as carbonization, chlorination, and freeze-thaw cycles. Degradation stages refer to the different periods from the initiation to the development of degradation, typically divided into initial, development, and acceleration stages. The distribution range of physical field characteristic parameters represents the value range of acoustic, optical, and thermal characteristic parameters corresponding to each degradation pattern at different stages.

[0035] In the above embodiment, a complete knowledge graph of the degradation evolution of the aforementioned industrial park office building is established. By integrating multi-source historical data and theoretical models, a knowledge representation system for the degradation mechanism is constructed. First, detailed material information of the building is extracted from the building archive database. According to the construction records, the main structure of the building uses C30 concrete with the following mix proportions: cement (P.O42.5) 320 kg / m³, fine aggregate 730 kg / m³, coarse aggregate 1180 kg / m³, water 175 kg / m³, fly ash content 15%, and water-reducing agent content 0.8%. Mechanical performance indicators include: 28-day compressive strength design value 30 MPa, measured value 34.5 MPa, and elastic modulus 3.0 × 10⁻⁶. 4 MPa, tensile strength 2.85MPa. Microstructure parameters were obtained from historical test reports: initial porosity 12.3%, average pore size 38nm, capillary pores 65%, gel pores 30%, macropores 5%. Cement hydration degree reached 85%, Ca(OH)2 content accounted for 22% of the cement stone weight. Environmental data for the region over the past 25 years were obtained from environmental monitoring stations. Historical temperature records show: annual average temperature 15.3℃, extreme maximum temperature 42℃, extreme minimum temperature -13℃, annual temperature fluctuation 55℃, daily average temperature difference 12℃. Humidity cycle data shows: annual average relative humidity 68%, average humidity during the rainy season (April-September) 82%, average humidity during the dry season 54%, and approximately 48 wet-dry cycles per year. Acid rain monitoring data shows: annual average rainfall pH value 5.2, minimum value 4.8, annual average SO4²⁻ concentration 12.5mg / L, NO3⁻ concentration 8.3mg / L, and annual average acid rain frequency 35%. Freeze-thaw records show an average of 18 freeze-thaw cycles per year, with a maximum freezing depth of 150 mm. Ultraviolet irradiation data shows an average annual UV index of 4.5 and a cumulative UV-B irradiance of 3850 MJ / m².

[0036] In the above embodiments, a set of material degradation response functions was established based on material property parameters: the carbonation depth function adopted a modified square root model: X_c = k_c × √t × (1 + 0.15RH) × (1.2 - 0.01T), where k_c is the carbonation coefficient 3.2 mm / year^0.5 (calculated based on material porosity and alkalinity reserve), t is time (years), RH is relative humidity (%), and T is temperature (°C). After 25 years of service, the calculated carbonation depth is: X_c = 3.2 × √25 × (1 + 0.15 × 0.68) × (1.2 - 0.01 × 15.3) = 19.8 mm. The carbonation sensitivity coefficient S_c = 0.73, indicating that the concrete has moderate sensitivity to carbonation. Chloride ion diffusion was solved using Fick's second law: C(x,t)=C_s×[1-erf(x / 2√(D_a×t))], diffusion coefficient D_a=D_0×(t_0 / t)^m, where D_0=8.5×10^-12m² / s, m=0.3. Although the area is not coastal, the use of de-icing agents resulted in a surface chloride ion concentration C_s of 0.18% (by weight of cement). The chloride ion concentration at a depth of 15mm after 25 years was calculated to be 0.08%, close to the critical value of 0.1% for steel corrosion. The chloride ion penetration sensitivity coefficient S_cl=0.45, indicating that the material has good resistance to chloride ion penetration. Freeze-thaw damage was assessed using a fatigue damage model: D_f=Σ(n_i / N_i)^β, where n_i is the actual number of freeze-thaw cycles, N_i is the material's resistance to freeze-thaw cycles (300 cycles), and β=1.5. After 450 freeze-thaw cycles, the damage degree D_f = (450 / 300)^1.5 = 1.84, indicating that the design freeze resistance has been exceeded. The freeze-thaw damage sensitivity coefficient S_f = 0.82, reflecting the material's high sensitivity to freeze-thaw effects. Fatigue damage was assessed using the Miner criterion: D_fatigue = Σ(n_i / N_i), considering wind load and temperature stress cycles, with an average annual equivalent stress cycle of 3650 cycles and a stress level of 0.3 times the ultimate strength. The cumulative fatigue damage over 25 years, D_fatigue = 0.28, has not yet reached fatigue failure. The fatigue cumulative damage sensitivity coefficient S_fatigue = 0.35. Based on the combined effects of various deterioration driving forces, the cumulative amounts are calculated as follows: cumulative amount of carbonization driving force L_c = X_c × S_c = 19.8 × 0.73 = 14.45; cumulative amount of chloride ion corrosion driving force L_cl = C(15mm) × S_cl / C_cr = 0.08 × 0.45 / 0.1 = 0.36; cumulative amount of freeze-thaw cycle driving force L_f = D_f × S_f = 1.84 × 0.82 = 1.51; cumulative amount of fatigue load driving force L_fatigue = D_fatigue × S_fatigue = 0.28 × 0.35 = 0.098.

[0037] In the above embodiments, the weights were determined using the analytic hierarchy process (AHP): carbonization 0.25, chloride ions 0.15, freeze-thaw cycles 0.45, and fatigue 0.15. Weighted fusion yielded the total cumulative deterioration driving force: L_total = 0.25 × 14.45 + 0.15 × 0.36 + 0.45 × 1.51 + 0.15 × 0.098 = 4.36, indicating that the building is in a moderate deterioration state, with freeze-thaw cycles being the dominant factor. K-means clustering analysis was performed on 1200 sets of multiphysics data, with the cluster number k = 8, and converged after 200 iterations. The following deterioration modes were identified: Mode 1 - Surface carbonization (23%): Acoustic resonance peak frequency 32.8-33.5 kHz, spectral reflectance at 550 nm 0.38-0.44, and thermal diffusion time constant 4.5-5.2 s. The characteristics of this model are not obvious in the early stage of degradation (0-5 years), the spectral characteristics are significant in the development stage (5-15 years), and the characteristics in all three domains deviate from the normal value by more than 20% in the accelerated stage (after 15 years). Model 2 - Freeze-thaw exfoliation type (accounting for 31%): Acoustic spectral entropy 0.75-0.88, surface roughness index 4.0-5.2 / 100nm, surface heat transfer coefficient 13.5-16.2W / (m²·K). In the early stage, only the surface roughness increases slightly, the acoustic entropy rises rapidly in the development stage, and obvious thermal anomalies appear in the accelerated stage. Model 3 - Local water loss type (accounting for 18%): 970nm reflectance 0.32-0.40, acoustic Q value 15-18, thermal resistance gradient 0.15-0.22K·m² / W·mm. Moisture intrusion leads to a multi-physics coordinated response, and the characteristic parameters in the three stages evolve exponentially. Mode 4 - Interface Debonding (12%): A 48kHz resonance peak appears, exhibiting thermal double-exponential cooling characteristics, with a spectral first derivative extremum shift of 8-12 nm. This mode shows significant abrupt changes and a short transition period from normal to failure. Mode 5 - Composite Deterioration (16%): Simultaneously possessing characteristics of two or more individual modes, with large dispersion in various physical field parameters, making prediction difficult. Based on the above analysis results, a degradation evolution knowledge graph with 486 nodes and 1250 edges was constructed. Nodes include: 15 material type nodes (concrete, mortar, etc. of different strength grades), 28 environmental factor nodes (subcategories such as temperature, humidity, and chemical corrosion), 8 degradation mode nodes, and 435 physical field response nodes (covering feature combinations of each mode at different stages). Edge weights were determined based on statistical correlation and physical causality, and confidence levels were calculated through cross-validation, with an average confidence level of 0.82. The knowledge graph also embeds temporal evolution paths, describing the typical development trajectory of each degradation mode. For example, the evolution path of freeze-thaw spalling is as follows: surface capillary saturation (1-3 years) → initiation of surface microcracks (3-8 years) → interconnection of cracks to form a network (8-15 years) → surface spalling (15-20 years) → deep damage (after 20 years). Each stage transition is associated with a corresponding threshold change in physical field characteristics.By constructing this comprehensive knowledge graph of degradation and evolution, we can not only identify the current state of degradation, but also predict future evolution paths, providing a scientific basis for formulating preventive maintenance strategies.

[0038] In an optional embodiment, a degradation evolution knowledge graph is constructed based on material property parameters, cumulative degradation driving forces, degradation mode classification results, and the distribution range of physical field characteristic parameters. Specifically, this includes: converting the material property parameters, cumulative degradation driving forces, degradation mode classification results, and the distribution range of physical field characteristic parameters into graph-structured data; establishing material type nodes, environmental factor nodes, degradation mode nodes, and physical field response nodes in the graph-structured data; establishing sensitivity-related edges between material type nodes and environmental factor nodes, driving-related edges between environmental factor nodes and degradation mode nodes, and response-related edges between degradation mode nodes and physical field response nodes; and assigning weight values ​​and confidence parameters to each related edge, with the weight values ​​used to indicate the graph. The correlation strength between nodes in the structural data, and the confidence parameter, are used to indicate the reliability of the correlation between nodes. Based on the response correlation edges, weight values, and confidence parameters, the distribution range of physical field feature parameters corresponding to each degradation mode is statistically analyzed to generate physical field response feature templates for multiple degradation modes. According to the preset degradation stage division requirements, the physical field response feature templates of multiple degradation modes are divided into early degradation feature templates, degradation development feature templates, and degradation acceleration feature templates. Each physical field response feature template of a degradation mode includes a template acoustic feature vector, a template spectral feature vector, and a template thermal feature vector. The early degradation feature templates, degradation development feature templates, and degradation acceleration feature templates are correlated to obtain a degradation evolution knowledge graph.

[0039] The graph structure data represents data organized using a graph theory model, consisting of nodes and edges, capable of expressing entities and their relationships. Material type nodes represent different types of building materials, such as C30 concrete nodes and sintered brick nodes. Environmental factor nodes represent various environmental effects, such as CO2 concentration nodes and chloride ion concentration nodes. Deterioration mode nodes represent different types of deterioration, such as carbonization nodes and steel corrosion nodes. Physical field response nodes represent physical field characteristic parameters, such as resonance frequency nodes and spectral reflectance nodes. Sensitivity-related edges are directed edges connecting material type nodes and environmental factor nodes, representing the material's sensitivity to environmental factors. Driving-related edges are directed edges connecting environmental factor nodes and deterioration mode nodes, representing the relationship between environmental factors and deterioration. Response-related edges are directed edges connecting deterioration mode nodes and physical field response nodes, representing the response characteristics of the deterioration mode in the physical field. Weight values ​​represent the numerical attributes of the edges, quantifying the correlation strength, such as sensitivity coefficients and driving effects. Coefficients, etc.; Confidence parameter refers to the reliability measure of the edge, reflecting the certainty of the association, usually determined based on the amount of supporting data or expert rating; Association strength indicates the closeness of the connection between two nodes, the larger the weight value, the stronger the association; Association reliability refers to the credible level of the association, the higher the confidence, the more reliable; Template acoustic feature vector represents the standard acoustic domain feature parameter combination of a specific degradation mode, as a reference benchmark for matching; Template spectral feature vector represents the standard spectral domain feature parameter combination of a specific degradation mode; Template thermal feature vector represents the standard thermal domain feature parameter combination of a specific degradation mode; Preset degradation stage division requirements refer to the stage division standards determined according to the degradation development law and management needs, such as division according to degradation depth or remaining life ratio; Early degradation feature template represents the physical field feature mode at the initial stage of degradation, usually with weak feature changes; Degradation development stage feature template represents the feature mode of the stable development stage of degradation, with obvious and linear feature changes; Degradation acceleration stage feature template represents the feature mode of the rapid development stage of degradation, with drastic and nonlinear feature changes.

[0040] In the above embodiments, all the data collected and analyzed are converted into a graph structure to construct a multi-level, scalable knowledge graph of degradation evolution. First, the raw data is converted into a format recognizable by the graph database. The Neo4j graph database engine is used to define the data patterns for nodes and edges. Material property parameters are converted into 15 material type nodes, each containing attribute fields such as material ID, name, strength grade, porosity, and water-cement ratio. For example, node M001 represents C30 concrete, with attributes including {compressive strength: 34.5 MPa, porosity: 12.3%, alkali reserve: 22%}. Environmental factor data is converted into 28 environmental nodes, such as node E005 representing acid rain erosion, with attributes including {annual average pH: 5.2, SO4 concentration: 12.5 mg / L, frequency: 35%}. The degradation pattern clustering results were converted into 8 degradation pattern nodes. Node D003 represents freeze-thaw spalling, with attributes including {occurrence probability: 31%, typical depth: 15mm, development rate: 2.3mm / year}. The distribution of physical field characteristic parameters was converted into 435 response nodes, such as node R128 representing resonant frequency shift, with attributes including {normal value: 33.2kHz, shift range: -0.5~-1.2kHz}. When constructing the relationships between nodes, three types of key edge relationships were established: sensitivity-related edges connecting material and environment nodes. For example, the edge from M001 (C30 concrete) to E008 (freeze-thaw cycle) has a weight of 0.82, indicating that the material is highly sensitive to freeze-thaw cycles, with a confidence level of 0.90 based on laboratory test data. The edge from M001 to E005 (acid rain) has a weight of 0.45, indicating moderate sensitivity, with a confidence level of 0.85. A total of 15 × 28 = 420 potential sensitive edges were established. After threshold screening (weight > 0.3 and confidence > 0.7), 186 valid edges were retained. Driving edges connect environmental factors and degradation patterns. Edges from E008 (freeze-thaw cycle) to D003 (freeze-thaw spalling) have a weight of 0.75, indicating a strong driving relationship, and a confidence level of 0.93 based on 25 years of field data. Edges from E005 (acid rain) to D001 (surface carbonization) have a weight of 0.68 and a confidence level of 0.88. These edges also contain time-dependent attributes, such as activation threshold (requiring 18 cumulative freeze-thaw cycles to activate) and saturation value (no further acceleration after 450 cycles). 28 × 8 = 224 potential driving edges were established, and 92 were retained after screening. Response edges connect degradation patterns and physical field responses. The edge from D003 (freeze-thaw spalling type) to R087 (increased spectral entropy) has a weight of 0.71, indicating that this degradation mode strongly affects the spectral entropy, with a confidence level of 0.86. Edge attributes also include response delay (significant only after 3 months of degradation) and sensitivity (each unit of degradation results in a 0.15 increase in entropy). A total of 8 × 435 = 3480 potential response edges were established, and 872 significant edges were retained through correlation analysis.

[0041] In the above embodiments, the calculation of weight values ​​and confidence scores employs a multi-source fusion approach. The weight values ​​comprehensively consider: statistical correlation coefficient (40%), physical mechanism analysis (35%), and expert scores (25%). For example, the weight calculation for the driving edge of freeze-thaw-stripping is as follows: statistical correlation r=0.78 contributes 0.78×0.4=0.312, mechanism analysis score 0.85 contributes 0.85×0.35=0.298, expert score 0.72 contributes 0.72×0.25=0.18, and the final weight = 0.79. Confidence scores are based on the supporting sample size and data quality: when the sample size > 100 and the coefficient of variation < 0.2, the confidence score > 0.9; when the sample size is 50-100, the confidence score is 0.7-0.9; when the sample size < 50, the confidence score is < 0.7. Based on the constructed graph structure, a graph traversal algorithm is used to extract the complete response path for each degradation mode. For the freeze-thaw spalling type (D003), traversing all response edges, the distribution of physical field response node values ​​connected is statistically analyzed: the acoustic domain includes resonant frequency 32.1-32.8kHz (probability 0.73), Q value 14.5-17.2 (probability 0.68), and spectral entropy 0.78-0.91 (probability 0.81); the spectral domain includes 970nm reflectance 0.35-0.42 (probability 0.62) and surface roughness 4.2-5.8 / 100nm (probability 0.75); the thermal domain includes thermal diffusion time 6.2-7.5s (probability 0.69) and surface heat transfer coefficient 14.5-16.8W / (m²·K) (probability 0.72). Based on the degradation timeline, the feature templates are divided into three stages: Early degradation stage (0-30% lifetime loss): The initial template vector for freeze-thaw spalling is [Acoustic: (32.6±0.2kHz, 16.8±0.5, 0.68±0.03), Spectral: (0.46±0.02, 985±3nm, 3.2±0.2), Thermal: (4.8±0.3s, 11.2±0.5, 0.08±0.01)]. The features deviate from normal values ​​by 5-10%, requiring high-precision detection for identification. Degradation development stage (30-70% lifetime loss): The characteristic template is [Acoustic: (32.3±0.3kHz, 15.2±0.8, 0.76±0.05), Spectral: (0.41±0.03, 990±5nm, 4.5±0.4), Thermal: (5.8±0.5s, 13.5±0.8, 0.15±0.03)]. Characteristic deviation of 15-25% is identifiable by conventional testing. Accelerated degradation period (70-100% lifetime consumption): The characteristic template is [Acoustic: (31.8±0.5kHz, 13.5±1.2, 0.85±0.08), Spectral: (0.36±0.05, 998±8nm, 5.5±0.6), Thermal: (7.2±0.8s, 15.8±1.2, 0.22±0.05)].When the characteristic deviation exceeds 30%, the degradation rate increases exponentially.

[0042] In the above embodiments, transition relationships between stages are established, forming a temporal evolution chain. The transition conditions from the initial stage to the development stage are: cumulative freeze-thaw cycles > 150 times and a deviation of > 15% between any two physical domain features; from the development stage to the acceleration stage: surface spalling depth > 5 mm or spectral entropy > 0.8. Each transition edge includes an "average transition time" (average 8 years from the initial stage to the development stage) and a "transition probability" (85% chance of transition under specific conditions). The final constructed degradation evolution knowledge graph includes: 486 nodes (material 15 + environment 28 + degradation 8 + response 435), 1150 edges (sensitivity 186 + driver 92 + response 872), and 24 stage transition paths (8 degradation modes × 3 stages). The graph supports multiple query operations: given material and environment, predict possible degradation modes; given physical field features, infer the degradation type and stage; given a degradation mode, extract the corresponding detection features and evolution path. Through a visual interface, users can intuitively view the topological structure of the knowledge graph. A force-directed layout algorithm is employed, automatically clustering nodes with strong correlations. For example, the freeze-thaw correlation subgraph includes 12 environmental nodes, 3 degradation mode nodes, and 68 response nodes, forming a tightly connected community structure. Clicking on any node displays detailed attributes, and selecting any edge displays the correlation strength and confidence level. The system also provides a path search function; by inputting the starting node (e.g., a material type) and the target node (e.g., a physical field response), the system automatically calculates the most probable propagation path.

[0043] In an optional embodiment, the multiphysics feature vector is matched with the physical field response feature templates of multiple degradation modes to determine whether there are potential degradation modes in the target building facade. Specifically, this includes: decomposing the multiphysics feature vector into measured acoustic feature sub-vectors, measured spectral feature sub-vectors, and measured thermal feature sub-vectors; traversing the physical field response feature templates of each degradation mode, and performing the following similarity calculation operation on the currently traversed physical field response feature template: calculating the acoustic similarity between the measured acoustic feature sub-vectors and the template acoustic feature vectors of the current physical field response feature template to obtain the acoustic domain similarity. The spectral similarity between the measured spectral feature vector and the template spectral feature vector of the current physical field response feature template is calculated to obtain the spectral domain similarity. Similarly, the thermal similarity between the measured thermal feature vector and the template thermal feature vector of the current physical field response feature template is calculated to obtain the thermal domain similarity. Multiple comprehensive similarities are obtained by weighting and summing the acoustic domain similarity, spectral domain similarity, and thermal domain similarity corresponding to the same physical field response feature template using preset multi-domain fusion weighting coefficients. These comprehensive similarities are then compared with preset similarity thresholds to obtain similarity comparison results. Finally, based on the similarity comparison results, the similarity comparison is determined... When one of the multiple comprehensive similarities is greater than a preset similarity threshold, it is determined that the multiphysics feature vector has successfully matched the physical field response feature template of one of the multiple degradation modes, and a degradation mode is marked as a potential degradation mode. When at least two of the multiple comprehensive similarities are greater than the preset similarity threshold based on the similarity comparison results, the at least two comprehensive similarities are sorted in descending order, and the comprehensive similarity ranked first is marked as the highest comprehensive similarity, and the comprehensive similarity ranked second is marked as the second highest comprehensive similarity. The highest and second highest comprehensive similarities are then further analyzed. The overall similarity is calculated by difference to obtain the distinguishability between the highest and second-highest overall similarity. The distinguishability is then compared with a preset distinguishability threshold to obtain the distinguishability comparison result. If the distinguishability is greater than the preset distinguishability threshold based on the distinguishability comparison result, the degradation mode corresponding to the highest overall similarity is marked as a potential degradation mode. If the distinguishability is less than or equal to the preset distinguishability threshold based on the distinguishability comparison result, the degradation modes corresponding to at least two overall similarities are marked as potential degradation modes. When multiple overall similarities are less than the preset similarity threshold, it is determined that there are no potential degradation modes in the target building facade.

[0044] Among them, the measured acoustic feature subvector represents the acoustic domain feature parameter part separated from the multiphysics feature vector, including resonant frequency, attenuation coefficient, energy entropy, etc.; the measured spectral feature subvector refers to the spectral domain feature parameter part separated from the multiphysics feature vector, including reflectivity, derivative extrema, zero crossover density, etc.; the measured thermal feature subvector represents the thermal domain feature parameter part separated from the multiphysics feature vector, including thermal diffusion time constant, heat transfer coefficient, thermal resistance gradient, etc.; traversal refers to the operation of visiting each element in the set in sequence; similarity calculation operation represents the mathematical operation of calculating the similarity between two vectors, and common methods include cosine similarity, Euclidean distance, etc.; acoustic similarity meter The similarity calculation refers to the process of calculating the similarity between the measured and template acoustic feature vectors, which can be achieved using methods such as cosine similarity and Pearson correlation coefficient. Acoustic domain similarity represents the degree of matching between the measured acoustic features and the template acoustic features, with values ​​ranging from 0 to 1; the closer to 1, the more similar the features. Spectral similarity calculation refers to the process of calculating the similarity between the measured and template spectral feature vectors; spectral domain similarity represents the degree of matching between the measured spectral features and the template spectral features. Thermal similarity calculation refers to the process of calculating the similarity between the measured thermal features and the template thermal features; thermal domain similarity represents the degree of matching between the measured thermal features and the template thermal features. The preset multi-domain fusion weight coefficient represents the weight of each domain when integrating the similarities of multiple physical domains, based on the sensitivity of each domain to degradation. Certainty is determined, for example, the acoustic domain of carbonization degradation may have a higher weight; weighted summation refers to the calculation method of multiplying multiple values ​​by their respective weights and then adding them together; comprehensive similarity represents the overall matching degree after fusing the similarity of multiple physical domains, reflecting the overall similarity between the measured feature and a certain degradation mode template; preset similarity threshold is the critical value for judging whether a match is successful, usually set between 0.7 and 0.85, and a value higher than this is considered a match; similarity comparison refers to the operation of comparing the similarity with the threshold; similarity comparison result refers to the relationship of greater than, less than, or equal to obtained after comparison; a successful match means that the similarity between the measured feature and a certain template exceeds the threshold, and the detected object is considered to have that degradation mode; labeling refers to adding tags to data or objects. The operations for labeling; descending sorting refers to arranging by numerical value from largest to smallest; the highest comprehensive similarity score represents the largest value among all comprehensive similarity scores; the second highest comprehensive similarity score represents the second highest value among all comprehensive similarity scores; similarity difference calculation refers to the operation of calculating the difference between two similarity scores; the discrimination score is the difference between the highest and second highest comprehensive similarity scores, reflecting the degree of distinguishability between the most likely degradation mode and the second most likely degradation mode. A larger difference indicates a more certain judgment; the preset discrimination score threshold represents the critical value for whether a clear distinction can be made, usually set at 0.1-0.2. Values ​​higher than this value are considered to indicate a clear distinction; the discrimination comparison refers to comparing the discrimination score with the threshold; the discrimination comparison result represents the size relationship obtained after the comparison.

[0045] In the above embodiment, degradation pattern matching is performed on the 27-dimensional multiphysics feature vector extracted by the aforementioned E06-05 detection unit to determine the potential degradation type in the region. First, the 27-dimensional feature vector is decomposed according to the physical domain. The measured acoustic feature sub-vector contains 9 elements: [32.48kHz, 17.6, 0.089 / ms, 0.82, 48.3kHz, 15.82kHz, 0.23, -3.2dB / 10kHz, 2.8ms], which correspond to the resonant frequency, Q value, attenuation coefficient, spectral entropy, sub-resonant, peak spacing, low-frequency energy ratio, high-frequency attenuation rate, and phase delay, respectively. The measured spectral feature vector contains 8 elements: [0.38, 0.42, 992nm, 4.3 / 100nm, -0.0008 / nm, 712nm, 0.15, 1.35], corresponding to 970nm reflectance, 550nm reflectance, first derivative extremum, second derivative density, spectral slope, red edge position, vegetation index, and moisture index. The measured thermal feature vector contains 7 elements: [5.8s, 14.2W / (m²·K), 0.18K·m² / W·mm, 8.2K, 12.3s, 850J / (m²·K·s^0.5), 18°], corresponding to thermal diffusion time constant, surface heat transfer coefficient, thermal resistance gradient, peak temperature rise, cooling half-life, thermal inertia index, and phase lag. Eight physical field response feature templates for the development stage of eight degradation modes were retrieved from the knowledge graph, and similarity calculations were performed on each one: Mode 1 - Surface carbonization template: Template acoustic feature vector [32.85kHz, 18.5, 0.072 / ms, 0.68, 47.2kHz, 14.35kHz, 0.25, -2.8dB / 10kHz, 2.2ms], acoustic domain similarity calculation used normalized Euclidean distance d_acoustic=√(Σ((x_i-y_i) / σ_i)² / n)=0.342, similarity S_acoustic=1 / (1+d_acoustic)=0.745; Template spectral feature vector: [0.45, 0.4 [0, 988nm, 3.8 / 100nm, -0.0006 / nm, 708nm, 0.18, 1.12], the spectral domain similarity is expressed as cosine similarity: S_optic=(X·Y) / (||X||×||Y||)=0.832; the template thermal feature vector is expressed as: [5.2s, 12.8W / (m²·K), 0.12K·m² / W·mm, 7.5K, 10.8s, 920J / (m²·K·s^0.5), 15°], the thermal domain similarity is expressed as correlation coefficient: S_thermal=0.756; the overall similarity is expressed as: 0.35×0.745+0.30×0.832+0.35×0.756=0.775.

[0046] In the above embodiments, Mode 2 - Freeze-Thaw Exfoliation Template: Template acoustic feature vector: [32.35kHz, 16.2, 0.095 / ms, 0.81, 48.8kHz, 16.45kHz, 0.22, -3.5dB / 10kHz, 3.0ms], S_acoustic = 0.892; Template spectral feature vector: [0.39, 0.43, 995nm, 4.5 / 100nm, -0.0008 / nm, 71 [5nm, 0.14, 1.32], S_optic = 0.943; Template thermal feature vector: [6.0s, 14.8W / (m²·K), 0.20K·m² / W·mm, 8.5K, 12.8s, 820J / (m²·K·s^0.5), 19°], S_thermal = 0.921; Overall similarity = 0.35×0.892 + 0.30×0.943 + 0.35×0.921 = 0.918. Mode 3 - Local water loss type template: Template feature vector calculated as: S_acoustic = 0.823, S_optic = 0.876, S_thermal = 0.798; Overall similarity = 0.35×0.823 + 0.30×0.876 + 0.35×0.798 = 0.831. Mode 4 - Interface Debonding Formwork: S_acoustic=0.756, S_optic=0.692, S_thermal=0.812; Overall Similarity = 0.35×0.756+0.30×0.692+0.35×0.812=0.756. Mode 5 - Reinforcement Corrosion Formwork: Overall Similarity = 0.523. Mode 6 - Alkali-Aggregate Reaction Formwork: Overall Similarity = 0.412. Mode 7 - Sulfate Erosion Formwork: Overall Similarity = 0.385. Mode 8 - Composite Deterioration Formwork: Overall Similarity = 0.862. The multi-domain fusion weight coefficients are dynamically adjusted based on the building type and environmental characteristics. Considering the significant freeze-thaw action in this region, the acoustic and thermal weights are increased to 0.35, and the spectral weight is 0.30. The weights are determined through analytic hierarchy process and historical data verification to ensure sensitivity to the dominant degradation mechanism.

[0047] In the above embodiment, the eight comprehensive similarities were compared with a preset threshold of 0.70, and four patterns were found to exceed the threshold: freeze-thaw spalling: 0.918 (1st), compound degradation: 0.862 (2nd), local water loss: 0.831 (3rd), and surface carbonization: 0.775 (4th). The four degradation patterns exceeding the threshold were sorted in descending order, with freeze-thaw spalling having the highest similarity (0.918), followed by compound degradation (0.862). The discrimination between the highest and second highest similarities was calculated as: 0.918 - 0.862 = 0.056. This discrimination of 0.056 was compared with the preset discrimination threshold of 0.10. Since 0.056 < 0.10, the difference between the highest and second highest similarities is not significant enough, indicating the possible coexistence of multiple degradation patterns. Therefore, freeze-thaw spalling, compound degradation, local water loss, and surface carbonization were all marked as potential degradation patterns. To further analyze the characteristics of the composite degradation, cross-validation was performed on the four modes. It was found that the feature overlap between the freeze-thaw spalling type and the localized water damage type reached 68%, indicating a synergistic effect: moisture intrusion exacerbates freeze-thaw damage, while microcracks generated by freeze-thaw cycles provide channels for moisture penetration. The high similarity of the composite degradation types confirms this multi-factor coupling judgment. The contribution rate of each degradation mode was also calculated. Through partial least squares regression analysis, it was determined that in the current feature vector: freeze-thaw spalling contributes 42% of the feature variation, localized water damage contributes 28%, surface carbonization contributes 18%, and the remaining 12% is an interaction effect. This quantitative decomposition helps in formulating targeted maintenance strategies. For the other four degradation modes that did not exceed the threshold (steel corrosion 0.523, alkali-aggregate reaction 0.412, sulfate attack 0.385, and a backup mode), it was determined that these degradation types do not currently exist, but their similarity values ​​were still recorded as a benchmark for subsequent monitoring. The generated degradation identification report shows that detection unit E06-05 exhibits composite degradation dominated by freeze-thaw spalling, accompanied by localized water loss and surface carbonization. The confidence level of the dominant degradation mode is 91.8%, indicating significant composite degradation characteristics. Recommended measures include: prioritizing waterproofing to block moisture intrusion paths; performing surface sealing treatment before the freeze-thaw season; and regularly monitoring the carbonization depth to prevent it from reaching the thickness of the reinforcing steel protective layer. Through this multi-level similarity matching and intelligent decision-making mechanism, not only are single degradation modes identified, but the existence and primary / secondary relationship of composite degradation can also be determined. Especially when the degradation modes are close in similarity, the discrimination mechanism effectively avoids misjudgment and improves the reliability of diagnosis.

[0048] In an optional embodiment, when a potential degradation mode is determined to exist on the target building facade, a temporal correlation analysis is performed on the potential degradation mode to determine the evolution trend of the physical field characteristics of the potential degradation mode within M consecutive acquisition cycles. A degradation development rate index is determined based on the evolution trend of the physical field characteristics. When the degradation development rate index exceeds a preset rate threshold, early warning information for the early stage of degradation is generated. The method further includes: acquiring multi-physical field feature vectors of the same preset detection unit within M consecutive acquisition cycles; constructing a temporal multi-physical field feature vector sequence, which includes an acoustic domain feature parameter sequence, a spectral domain feature parameter sequence, and a thermal domain feature parameter sequence corresponding to the M consecutive acquisition cycles; and performing acoustic... Linear regression fitting is performed on the formant frequency positions in the acoustic domain characteristic parameter sequence to obtain the rate of change of formant frequency positions. Exponential fitting is performed on the formant amplitude attenuation coefficient in the acoustic domain characteristic parameter sequence to obtain the amplitude attenuation acceleration coefficient. Monotonicity testing is performed on the spectral energy distribution entropy in the acoustic domain characteristic parameter sequence to obtain the energy dispersion evolution trend index. Time series decomposition is performed on the spectral reflectance characteristic values ​​in the spectral domain characteristic parameter sequence to obtain the decreasing trend component and periodic fluctuation component of spectral reflectance. Drift statistics are performed on the positions of the extreme points of the first derivative of the spectral curve in the spectral domain characteristic parameter sequence to obtain the displacement rate of spectral characteristic peaks. Zero-crossing of the second derivative in the spectral domain characteristic parameter sequence is also performed. The surface roughness growth index is obtained by calculating the growth rate of point density; the thermal diffusion time constant in the thermal domain characteristic parameter sequence is differentially processed to obtain the thermal diffusion performance degradation rate; the surface heat transfer coefficient in the thermal domain characteristic parameter sequence is calculated by the relative change rate to obtain the surface density degradation rate; the internal defect propagation rate is obtained by detecting the gradient change of the internal thermal resistance distribution gradient in the thermal domain characteristic parameter sequence; the resonance peak frequency position change rate, amplitude attenuation acceleration coefficient, and energy dispersion evolution trend index are used as acoustic domain evolution rate parameters; the spectral reflectance decrease trend component, spectral characteristic peak displacement rate, and surface roughness growth index are used as spectral domain evolution rate parameters; and thermal diffusivity... The degradation rate, surface density deterioration rate, and internal defect propagation rate are used as thermal domain evolution rate parameters. Multi-domain evolution rate fusion weights corresponding to potential degradation modes are queried from the degradation evolution knowledge graph. These weights are then used to weight and fuse the acoustic, spectral, and thermal domain evolution rate parameters to obtain a degradation development rate index. This index is compared with a preset rate threshold to obtain the comparison results. When the degradation development rate index exceeds the preset rate threshold based on the comparison results, an early warning message for pre-degradation is generated, including the degradation mode type, degradation location coordinates, degradation development rate level, and suggested maintenance time window.Early warning information on pre-deterioration is pushed to the building maintenance management platform. On the platform's 3D visualization interface, color-coded indicators of the deterioration risk level are displayed at corresponding locations on the 3D model of the target building facade, based on the early warning information. Simultaneously, the spatial distribution range and predicted development trajectory of potential deterioration modes are marked.

[0049] Among them, the temporal multiphysics feature vector sequence represents a set of multiphysics feature vectors from multiple acquisition cycles arranged in chronological order, forming time series data; the acoustic domain feature parameter sequence refers to a sequence of acoustic domain feature parameters from multiple consecutive periods arranged in time, such as [resonance frequency 1, resonance frequency 2, ..., resonance frequency M]; the spectral domain feature parameter sequence represents a time series of spectral domain feature parameters from multiple consecutive periods; the thermal domain feature parameter sequence refers to a time series of thermal domain feature parameters from multiple consecutive periods; linear regression fitting means fitting data points with a linear function y=ax+b, determining the optimal slope a and intercept b to minimize the fitting error; the rate of change of resonance peak frequency position refers to the slope of the resonance frequency changing with time, with units of Hz / cycle, and negative values ​​indicating frequency... The rate of decay decreases; exponential fitting means fitting data with an exponential function y=ae^(bx), which is suitable for describing the decay process; the amplitude decay acceleration coefficient refers to the rate of change of the decay rate, reflecting whether the decay is accelerating or decelerating; monotonicity test is a statistical method to test whether the sequence shows a monotonically increasing or decreasing trend, such as the Mann-Kendall test; the energy dispersion evolution trend index represents the trend of energy entropy change over time, and monotonically increasing indicates that the damage diffusion is intensified; time series decomposition refers to the method of decomposing a time series into trend components, periodic components, and random components, such as seasonal decomposition; the spectral reflectance decreasing trend component represents the long-term decreasing trend of reflectance, eliminating seasonal fluctuations; the periodic fluctuation component refers to the reflectance variation due to seasons or diurnal changes. The periodic components caused by morphology; drift statistics represent the cumulative deviation of the calculated sequence values ​​from their initial positions; the displacement rate of spectral characteristic peaks refers to the speed at which the wavelength positions of spectral characteristic peaks shift, measured in nm / period; the growth rate calculation refers to the method for calculating the relative growth rate of the sequence, typically (current value - initial value) / initial value; the surface roughness growth index represents the growth rate of the surface microstructure complexity; differential processing refers to the operation of calculating the difference between adjacent elements in the sequence, with the first-order difference being x(t) - x(t-1); the thermal diffusivity degradation rate represents the rate at which the thermal diffusivity time constant increases, reflecting the speed of thermal conductivity deterioration; the relative change rate calculation refers to calculating the percentage change relative to a reference value; the surface density degradation rate represents the change in the surface heat transfer coefficient. Speed; gradient change detection refers to the method of detecting the evolution of the gradient field over time; internal defect propagation rate represents the speed of change of the internal thermal resistance gradient, reflecting the speed of propagation of layered defects; acoustic domain evolution rate parameter refers to the set of parameters characterizing the speed of acoustic property evolution over time; spectral domain evolution rate parameter represents the set of parameters characterizing the speed of optical property evolution; thermal domain evolution rate parameter refers to the set of parameters characterizing the speed of thermal property evolution; multi-domain evolution rate fusion weight represents the weight coefficient of each domain when integrating the evolution rates of multiple physical domains, obtained from the knowledge graph; rate comparison refers to comparing the degradation development rate with a threshold; rate comparison result represents the magnitude relationship obtained from the comparison; degradation mode type refers to the identified degradation type, such as carbonization, chlorination, etc.The degradation location coordinates represent the three-dimensional spatial location of the degradation; the degradation development rate level refers to discretizing the rate into low, medium, and high levels; the recommended maintenance time window represents the recommended maintenance intervention time based on rate prediction; the building maintenance management platform refers to an information system used for building operation and maintenance management, integrating detection, early warning, and maintenance decision-making functions; the 3D visualization interface represents a user interface that displays the building and its status in a 3D graphical format; the 3D model refers to the digital 3D geometric representation of the building; the degradation risk level color code refers to visual symbols that use different colors to indicate the level of degradation risk, such as green - low risk, yellow - medium risk, and red - high risk; the spatial distribution range represents the area affected by degradation on the building facade; the predicted development trajectory refers to predicting the path and range of future degradation expansion based on the current rate.

[0050] In the above embodiment, a time-series correlation analysis was performed on the E06-05 detection unit, which was identified as exhibiting composite degradation dominated by freeze-thaw spalling, to assess the degradation trend and determine whether an early warning should be issued. Multiphysics feature vectors for the detection unit were extracted from the historical database for six consecutive acquisition cycles (M=6, acquired monthly). The constructed time-series feature vector sequence showed a clear evolution trajectory: acoustic domain feature parameter sequence: formant frequency position sequence [33.05,32.92,32.78,32.65,32.53,32.48]kHz; formant amplitude attenuation coefficient sequence: [0.068,0.072,0.076,0.081,0.085,0.089] / ms; spectral energy distribution entropy sequence: [0.65,0.68,0.72,0.75,0.79,0.82]. Linear regression fitting of the resonant peak frequency yielded f(t) = 33.08 - 0.115t, with a correlation coefficient R² = 0.987, resulting in a frequency position change rate of -0.115 kHz / month, indicating that the material stiffness degrades at a rate of 0.35% per month. Exponential fitting of the amplitude attenuation coefficient yielded α(t) = 0.067 × exp(0.045t), with a goodness of fit R² = 0.973, resulting in an amplitude attenuation acceleration coefficient of 0.045 / month, reflecting an exponential deterioration of damping characteristics. A Mann-Kendall monotonicity test was performed on the spectral energy distribution entropy, with a test statistic Z = 4.12 and a p-value < 0.001, confirming a significant upward trend. The energy dispersion evolution trend index was calculated: an entropy increase rate of 0.028 / month, indicating continuous dispersion of internal damage.

[0051] In the above embodiment, the spectral domain characteristic parameter sequence is as follows: 970nm reflectance sequence: [0.48, 0.45, 0.43, 0.41, 0.39, 0.38]; first derivative extreme point sequence: [986, 987, 988, 990, 991, 992] nm; second derivative zero-crossing point density sequence: [3.2, 3.5, 3.7, 3.9, 4.1, 4.3] / 100nm. The spectral reflectance is decomposed into a time series using the STL (Season on Earth and Trend decomposition using Loess) method. The trend component shows a linear decrease: R_trend(t) = 0.485 - 0.018t, with a decrease rate of 0.018 / month. The periodic component has an amplitude of 0.012 and a period of approximately 2 months, which may be related to the alternating dry and wet conditions. Statistical analysis of the drift at the extreme points of the first derivative: total drift of 6 nm, average drift rate of 1.0 nm / month, indicating continuous changes in chemical composition. Autocorrelation analysis revealed the persistence of the drift, with an autocorrelation coefficient of 0.82 after a 1-period lag. Calculation of the density growth rate at the zero crossover point of the second derivative: (4.3-3.2) / 3.2 / 6 = 5.7% / month, indicating a rapid increase in surface roughness and predicting intensified weathering of the surface material. Thermal domain characteristic parameter sequences: thermal diffusion time constant sequence: [4.5,4.8,5.1,5.3,5.6,5.8] s; surface heat transfer coefficient sequence: [11.2,11.8,12.4,13.0,13.6,14.2] W / (m²·K); internal thermal resistance distribution gradient sequence: [0.08,0.10,0.12,0.14,0.16,0.18] K·m² / W·mm. Thermal diffusion time constant differential processing: First-order difference sequence [0.3, 0.3, 0.2, 0.3, 0.2] s / month, average degradation rate 0.26 s / month, corresponding to a 4.8% / month decrease in thermal diffusivity. Relative change rate of surface heat transfer coefficient: (14.2-11.2) / 11.2 / 6=4.5% / month, indicating a continuous decrease in surface density and increased porosity leading to enhanced convective heat transfer. Internal thermal resistance gradient change detection uses the CUSUM (cumulative sum) algorithm, detecting a change point in the 3rd month, with accelerated gradient growth. Calculated internal defect propagation rate: 0.02 K·m² / W·mm / month, indicating continuous development of layered defects.

[0052] In the above embodiments, the extracted evolution rate parameters are organized into parameter sets for three domains: acoustic domain evolution rate parameters [-0.115kHz / month, 0.045 / month, 0.028 / month]; spectral domain evolution rate parameters [0.018 / month, 1.0nm / month, 5.7% / month]; and thermal domain evolution rate parameters [0.26s / month, 4.5% / month, 0.02K·m² / W·mm / month]. The multi-domain evolution rate fusion weights for freeze-thaw spalling degradation modes are queried from the degradation evolution knowledge graph. Considering that freeze-thaw degradation mainly affects the material's physical structure, the weights are set as follows: acoustic domain 0.40, spectral domain 0.25, and thermal domain 0.35. After normalizing and weighting the evolution rate parameters of each domain, the acoustic domain contribution is 0.40×(0.115 / 0.15+0.045 / 0.05+0.028 / 0.03) / 3=0.347; the spectral domain contribution is 0.25×(0.018 / 0.02+1.0 / 1.2+0.057 / 0.06) / 3=0.223; the thermal domain contribution is 0.35×(0.26 / 0.30+0.045 / 0.05+0.02 / 0.025) / 3=0.308; the degradation development rate index is 0.347+0.223+0.308=0.878. The degradation development rate index of 0.878 is compared with the preset rate threshold of 0.60. Since 0.878>0.60, the degradation is determined to be in a rapid development stage, and an early warning message needs to be generated. Based on the degradation rate index, the remaining safe service time is further calculated. Using an exponential extrapolation model, the current degradation depth is approximately 15mm. At the current rate, it is projected to reach a 25mm rebar protective layer thickness in 12 months, and visible spalling may occur in 18 months. Detailed early warning information for degradation is generated: the degradation mode is freeze-thaw spalling combined with water damage; the degradation location coordinates are the east facade (X: 10.0m, Y: 18.0m, Z: 7.5m), with an area of ​​approximately 6.25m²; the degradation rate level is high (rate index 0.878, exceeding the threshold by 46.3%); the current degradation depth is estimated at 15±2mm; the recommended maintenance window is within 3-6 months (no later than 9 months); recommended maintenance measures include immediate surface waterproofing and sealing; application of penetrating antifreeze within 3 months; and consideration of local reinforcement or replacement repair within 6 months.

[0053] In the above embodiment, the early warning information is pushed to the building maintenance management platform in real time through an encrypted channel. In the platform's 3D visualization interface, the following display operations are automatically performed: At the E06-05 detection unit location in the building's 3D model, the system displays the corresponding risk color code according to the rate of deterioration. A five-level color code system is adopted: dark red (extremely high risk, rate index > 1.0), orange-red (high risk, 0.8-1.0), orange (medium-high risk, 0.6-0.8), yellow (medium risk, 0.4-0.6), and green (low risk, < 0.4). This unit is displayed as orange-red and flashes at a frequency of 0.5Hz to attract attention. The spatial distribution range of deterioration is marked on the 3D model, and a 6.25m² affected area is outlined using a semi-transparent red overlay. The boundary is smoothed using a Bézier curve to accurately reflect the gradual characteristics of deterioration. A gridded deterioration depth distribution map is displayed on the overlay, with color depth indicating deterioration depth, the central area being the deepest (dark red, 15mm), decreasing towards the edge (light red, 5mm). The predicted development trajectory is displayed via dynamic projection. An animation of degradation expansion over the next 18 months is generated, with each 3-month period serving as a time node. The animation shows the degradation area gradually expanding from the current 6.25 m² to the predicted 12 m², with the depth increasing from 15 mm to 28 mm. The trajectory line uses dashed lines to represent uncertainty, and line thickness indicates probability. Detailed numerical information and trend charts are displayed synchronously in the information panel. The trend chart includes three sub-charts: historical evolution curves and future prediction curves for multiphysics characteristic parameters; a degradation depth-time relationship chart, marking key nodes (such as the time to reach the protective layer thickness); and a maintenance cost-time relationship chart, showing the economic analysis of interventions at different time points. A downloadable detailed report is also generated, including: degradation mechanism analysis (illustrating the freeze-thaw-water damage synergistic mechanism with illustrations), historical data statistics (a complete 6-month characteristic evolution record), a description of the prediction model (the algorithm used and confidence intervals), a comparison of maintenance schemes (technical and economic analysis of 3 schemes), and similar case references (5 similar degradation cases retrieved from the knowledge base and their treatment effects). Through this comprehensive time-series correlation analysis and early warning system, maintenance personnel can accurately grasp the development of degradation and take intervention measures within the optimal time window. It should also be noted that the specific numerical examples described above are merely illustrative embodiments, and the specific numerical values ​​are not limited to the examples given.

[0054] The following describes the building damage early detection system in the embodiments of this invention from the perspective of hardware processing. (See attached document.) Figure 2 , Figure 2 This is a schematic diagram of the physical device structure of a building damage early detection system in the embodiments of this application.

[0055] It should be noted that, Figure 2The structure of the pre-damage detection system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0056] like Figure 2 As shown, the pre-damage detection system for buildings includes a Central Processing Unit (CPU) 201, which can perform various appropriate actions and processes based on a program stored in Read-Only Memory (ROM) 202 or a program loaded from storage section 208 into Random Access Memory (RAM) 203, such as performing the methods described in the above embodiments. The RAM 203 also stores various programs and data required for system operation. The CPU 201, ROM 202, and RAM 203 are interconnected via a bus 204. An Input / Output (I / O) interface 205 is also connected to the bus 204.

[0057] The following components are connected to I / O interface 205: input section 206 including audio input devices, push-button switches, etc.; output section 207 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 208 including a hard disk, etc.; and communication section 209 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 209 performs communication processing via a network such as the Internet. Drive 210 is also connected to I / O interface 205 as needed. Removable media 211, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 210 as needed so that computer programs read from them can be installed into storage section 208 as needed.

[0058] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 209, and / or installed from removable medium 211. When the computer program is executed by central processing unit (CPU) 201, it performs the various functions defined in the present invention.

[0059] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

Claims

1. A method for early detection of building damage based on multi-field coupling, characterized in that, include: The drone carries acoustic field sensing sensor group, spectral sensing sensor group and thermal field sensing sensor group to collect excitation acoustic response signal, selected band spectral reflection signal and transient thermal conduction response signal of the target building facade respectively. The excitation acoustic response signal, the selected band spectral reflection signal, and the transient thermal conduction response signal are subjected to time-frequency domain transformation to obtain a multi-physics feature vector. The material property parameters and environmental impact history data of the target building facade are obtained, and a degradation evolution knowledge graph is constructed based on the material property parameters and environmental impact history data. The degradation evolution knowledge graph includes physical field response feature templates for multiple degradation modes. The multiphysics feature vector is matched with the physical field response feature template of the multiple degradation modes to determine whether the target building facade has potential degradation modes. When it is determined that the target building facade has the potential deterioration mode, a time-series correlation analysis is performed on the potential deterioration mode to determine the evolution trend of the physical field characteristics of the potential deterioration mode within M consecutive collection cycles, and a deterioration development rate index is determined based on the evolution trend of the physical field characteristics. When the deterioration development rate index is greater than a preset rate threshold, early warning information of deterioration is generated, where M is a positive integer.

2. The method according to claim 1, characterized in that, The method utilizes an acoustic field sensing sensor group, a spectral sensing sensor group, and a thermal field sensing sensor group mounted on the UAV to collect the excitation acoustic response signal, the selected band spectral reflection signal, and the transient thermal conduction response signal of the target building facade, respectively. Specifically, this includes: The multiphysics excitation parameters are determined based on the material type, construction year, and environmental exposure conditions of the target building facade. The multiphysics excitation parameters include the acoustic field excitation frequency range, the spectral detection band combination, and the thermal field excitation power density. An array of hovering reference points for a drone is set at a predetermined distance outside the target building facade, so that the detection field of view of the drone at each hovering reference point in the array covers a predetermined detection unit of the target building facade. The UAV is controlled to hover sequentially at each hovering reference point in the hovering reference point array. At each hovering reference point, the UAV transmits a swept-frequency acoustic wave signal to a preset detection unit covered by the current hovering reference point through the sound field sensing sensor group. Simultaneously, the UAV collects the reflected and transmitted acoustic wave signals of the current preset detection unit under the excitation of the swept-frequency acoustic wave signal. The reflected and transmitted acoustic wave signals are used as the excitation acoustic response signal. The frequency range of the swept-frequency acoustic wave signal covers the sound field excitation frequency band. Within a preset time interval after the acquisition of the excitation acoustic response signal, the spectral sensing sensor group is controlled to sequentially perform multi-band spectral scanning on the current preset detection unit according to the spectral detection band combination, so as to collect the reflectance spectral data of the current preset detection unit in the ultraviolet band, visible light band and near-infrared band, and use the reflectance spectral data as the selected band spectral reflectance signal. The ultraviolet band is used to detect the scattering characteristic changes caused by microcracks on the surface of the current preset detection unit, the visible light band is used to identify abnormal color difference areas on the surface of the current preset detection unit, and the near-infrared band is used to detect abnormal moisture distribution inside the current preset detection unit. After completing the acquisition of the selected band spectral reflection signal, a pulsed thermal excitation is applied to the current preset detection unit through the thermal field sensing sensor group, and the surface temperature field time-series evolution data of the current preset detection unit during the heating and cooling stages are acquired simultaneously, and the surface temperature field time-series evolution data is used as the transient thermal conduction response signal. The excitation acoustic response signal, the selected band spectral reflection signal, and the transient thermal conduction response signal collected at the same hovering reference point are timestamped to obtain a multiphysics signal group carrying time identifiers. The three-dimensional spatial coordinates of the corresponding hovering reference point and the number information of the corresponding preset detection unit are added to the multiphysics signal group to generate a multiphysics original dataset with spatiotemporal consistency characteristics.

3. The method according to claim 2, characterized in that, The step of performing time-frequency domain transformation on the excitation acoustic response signal, the selected band spectral reflection signal, and the transient thermal conduction response signal to obtain a multi-physics feature vector specifically includes: The excitation acoustic response signal in the original multiphysics dataset is processed by short-time Fourier transform to obtain a time-spectrum matrix. The resonant peak frequency position, resonant peak amplitude attenuation coefficient, and spectral energy distribution entropy are extracted from the time-spectrum matrix as acoustic domain feature parameters. The resonant peak frequency position characterizes the material stiffness degradation characteristics of the target building facade, the resonant peak amplitude attenuation coefficient characterizes the material damping loss evolution characteristics of the target building facade, and the spectral energy distribution entropy characterizes the discrete distribution characteristics of internal material damage of the target building facade. Continuous wavelet transform processing is performed on the selected band spectral reflectance signals in the original multiphysics dataset to obtain spectral reflectance curves. Spectral reflectance feature values, the positions of the first derivative extrema of the spectral curve, and the density of zero-crossing points of the second derivative are extracted from the spectral reflectance curves as spectral domain feature parameters. The spectral reflectance feature values ​​characterize the weathering degree of the material surface of the target building facade, the positions of the first derivative extrema of the spectral curve characterize the variation characteristics of the chemical composition of the material of the target building facade, and the density of zero-crossing points of the second derivative characterizes the microscopic morphological complexity of the material surface of the target building facade. The transient thermal conduction response signal in the original multiphysics dataset is processed by Laplace transform to obtain the complex frequency domain temperature transfer function. The thermal diffusion time constant, surface heat transfer coefficient, and internal thermal resistance distribution gradient are extracted from the complex frequency domain temperature transfer function as thermal domain feature parameters. The thermal diffusion time constant characterizes the degradation characteristics of the thermal conductivity of the target building facade material, the surface heat transfer coefficient characterizes the surface density variation characteristics of the target building facade material, and the internal thermal resistance distribution gradient characterizes the internal delamination defect characteristics of the target building facade material. The acoustic domain feature parameters, the spectral domain feature parameters, and the thermal domain feature parameters are arranged and combined according to a preset feature order to generate an initial multiphysics feature vector; The initial multiphysics field feature vector is subjected to cross-physics field coupling strength analysis to obtain the multiphysics field feature vector.

4. The method according to claim 2, characterized in that, The method of acquiring material property parameters and historical environmental impact data of the target building facade, and constructing a degradation evolution knowledge graph based on the material property parameters and historical environmental impact data, further includes: The material composition information, mechanical performance indicators, and microstructure parameters of the target building facade are obtained from the building archive database as material property parameters. Historical temperature change data, humidity cycle data, acid rain erosion data, freeze-thaw cycle number, and cumulative ultraviolet radiation in the area where the target building facade is located are obtained from the environmental monitoring station as historical environmental impact data. Based on the material property parameters, a set of material degradation response functions is established. The set of material degradation response functions includes the functional relationship between carbonization depth and carbonization sensitivity coefficient, the functional relationship between chloride ion diffusion depth and chloride ion penetration sensitivity coefficient, the functional relationship between freeze-thaw damage degree and freeze-thaw damage sensitivity coefficient, and the functional relationship between fatigue damage accumulation and fatigue cumulative damage sensitivity coefficient. Based on the set of material degradation response functions and the historical data of environmental effects, the cumulative amount of carbonization driving force, chloride ion corrosion driving force, freeze-thaw cycle driving force, and fatigue load driving force experienced by the target building facade are determined. The cumulative amount of carbonization driving force, chloride ion corrosion driving force, freeze-thaw cycle driving force, and fatigue load driving force are weighted and fused to obtain the cumulative amount of degradation driving force of the target building facade. Degradation pattern clustering analysis was performed on the original multiphysics dataset to obtain the degradation pattern classification results and the distribution range of physical field characteristic parameters corresponding to each degradation pattern at different degradation stages. The degradation evolution knowledge graph is constructed based on the material property parameters, the cumulative amount of degradation driving force, the degradation mode classification results, and the distribution range of the physical field characteristic parameters.

5. The method according to claim 4, characterized in that, The construction of the degradation evolution knowledge graph based on the material property parameters, the cumulative amount of degradation driving force, the degradation mode classification results, and the distribution range of the physical field characteristic parameters specifically includes: The material property parameters, the cumulative amount of the degradation driving force, the degradation mode classification results, and the distribution range of the physical field characteristic parameters are converted into graph structure data. In the graph structure data, establish nodes for material type, environmental factors, degradation mode, and physical field response; Establish sensitivity correlation edges between material type nodes and environmental factor nodes, establish driving correlation edges between environmental factor nodes and degradation mode nodes, and establish response correlation edges between degradation mode nodes and physical field response nodes. Each associated edge is assigned a weight value and a confidence parameter. The weight value is used to indicate the strength of the association between nodes in the graph structure data, and the confidence parameter is used to indicate the reliability of the association between nodes. Based on the response association edge, the weight value, and the confidence parameter, the distribution range of the physical field feature parameters corresponding to each degradation mode is calculated to generate physical field response feature templates for the multiple degradation modes; According to the preset degradation stage division requirements, the physical field response feature templates of the multiple degradation modes are divided into early degradation feature templates, degradation development feature templates, and degradation acceleration feature templates. Each degradation mode physical field response feature template includes a template acoustic feature vector, a template spectral feature vector, and a template thermal feature vector. By associating the feature templates of the initial stage of degradation, the feature templates of the development stage of degradation, and the feature templates of the accelerated stage of degradation, the degradation evolution knowledge graph is obtained.

6. The method according to claim 5, characterized in that, The step of performing similarity matching between the multiphysics feature vector and the physical field response feature templates of the multiple degradation modes to determine whether the target building facade has potential degradation modes specifically includes: The multiphysics field feature vector is decomposed into measured acoustic feature vector, measured spectral feature vector, and measured thermal feature vector; Iterate through the physical field response feature templates for each degradation mode, and perform the following similarity calculation operation on the currently iterated physical field response feature template: Acoustic similarity is calculated between the measured acoustic feature vector and the template acoustic feature vector of the current physical field response feature template to obtain acoustic domain similarity; spectral similarity is calculated between the measured spectral feature vector and the template spectral feature vector of the current physical field response feature template to obtain spectral domain similarity; and thermal similarity is calculated between the measured thermal feature vector and the template thermal feature vector of the current physical field response feature template to obtain thermal domain similarity. By using preset multi-domain fusion weighting coefficients, the acoustic domain similarity, spectral domain similarity, and thermal domain similarity corresponding to the same physical field response feature template are weighted and summed to obtain multiple comprehensive similarities; The multiple comprehensive similarities are compared with preset similarity thresholds to obtain similarity comparison results; When it is determined from the similarity comparison results that one of the multiple comprehensive similarities is greater than the preset similarity threshold, it is determined that the multi-physics feature vector is successfully matched with the physical field response feature template of one of the multiple degradation modes, and the degradation mode is marked as the potential degradation mode. When it is determined from the similarity comparison results that at least two of the multiple comprehensive similarities are greater than the preset similarity threshold, the at least two comprehensive similarities are sorted in descending order, and the comprehensive similarity that is first after sorting is marked as the highest comprehensive similarity, and the comprehensive similarity that is second after sorting is marked as the second highest comprehensive similarity. The difference between the highest overall similarity and the second highest overall similarity is calculated to obtain the distinguishability between the highest overall similarity and the second highest overall similarity. The discrimination score is compared with a preset discrimination score threshold to obtain the discrimination comparison result. When the discrimination score is determined to be greater than the preset discrimination score threshold based on the discrimination comparison result, the degradation mode corresponding to the highest comprehensive similarity is marked as the potential degradation mode; When the discrimination score is determined to be less than or equal to the preset discrimination score threshold based on the discrimination comparison result, the degradation modes corresponding to the at least two comprehensive similarities are all marked as the potential degradation modes; When all of the multiple comprehensive similarities are less than the preset similarity threshold, it is determined that the target building facade does not have the potential degradation mode.

7. The method according to claim 1, characterized in that, When it is determined that the target building facade has the potential degradation mode, a time-series correlation analysis is performed on the potential degradation mode to determine the evolution trend of the physical field characteristics of the potential degradation mode within M consecutive collection periods, and a degradation development rate index is determined based on the evolution trend of the physical field characteristics. When the degradation development rate index is greater than a preset rate threshold, early warning information for early degradation is generated. The method further includes: The multi-physics feature vectors of the same preset detection unit within the M consecutive acquisition cycles are obtained, and a temporal multi-physics feature vector sequence is constructed. The temporal multi-physics feature vector sequence includes the acoustic domain feature parameter sequence, the spectral domain feature parameter sequence, and the thermal domain feature parameter sequence corresponding to the M consecutive acquisition cycles. Linear regression fitting is performed on the formant frequency positions in the acoustic domain feature parameter sequence to obtain the formant frequency position change rate; exponential fitting is performed on the formant amplitude attenuation coefficient in the acoustic domain feature parameter sequence to obtain the amplitude attenuation acceleration coefficient; and monotonicity test is performed on the spectral energy distribution entropy in the acoustic domain feature parameter sequence to obtain the energy dispersion evolution trend index. The spectral reflectance characteristic values ​​in the spectral domain feature parameter sequence are decomposed into a time series to obtain the decreasing trend component and periodic fluctuation component of spectral reflectance. The drift of the extreme points of the first derivative of the spectral curve in the spectral domain feature parameter sequence is statistically analyzed to obtain the displacement rate of the spectral characteristic peak. The growth rate of the density of zero crossover points of the second derivative in the spectral domain feature parameter sequence is calculated to obtain the surface roughness growth index. The thermal diffusion time constant in the thermal domain characteristic parameter sequence is differentially processed to obtain the thermal diffusion performance degradation rate. The surface heat transfer coefficient in the thermal domain characteristic parameter sequence is relatively changed to obtain the surface compactness degradation rate. The internal thermal resistance distribution gradient in the thermal domain characteristic parameter sequence is gradient changed to obtain the internal defect propagation rate. The resonant peak frequency position change rate, the amplitude attenuation acceleration coefficient, and the energy dispersion evolution trend index are used as acoustic domain evolution rate parameters; the spectral reflectance decrease trend component, the spectral characteristic peak displacement rate, and the surface roughness growth index are used as spectral domain evolution rate parameters; and the thermal diffusion performance degradation rate, the surface compaction deterioration rate, and the internal defect propagation rate are used as thermal domain evolution rate parameters. The deterioration evolution knowledge graph is used to query the multi-domain evolution rate fusion weight corresponding to the potential deterioration mode, and the acoustic domain evolution rate parameter, the spectral domain evolution rate parameter and the thermal domain evolution rate parameter are weighted and fused using the multi-domain evolution rate fusion weight to obtain the deterioration development rate index. The rate of degradation development is compared with the preset rate threshold to obtain the rate comparison result. When the degradation development rate index is determined to be greater than the preset rate threshold based on the rate comparison result, early warning information for degradation is generated, including degradation mode type, degradation location coordinates, degradation development rate level and recommended maintenance time window. The early warning information of the deterioration is pushed to the building maintenance management platform. In the three-dimensional visualization interface of the building maintenance management platform, the deterioration risk level color mark is displayed at the corresponding position of the three-dimensional model of the target building facade according to the early warning information of the deterioration, and the spatial distribution range and predicted development trajectory of the potential deterioration mode are marked.

8. A building damage early detection system, characterized in that, The pre-damage detection system for buildings includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the pre-damage detection system for buildings to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is run on the building damage early detection system, the building damage early detection system performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the building damage early detection system, the building damage early detection system performs the method as described in any one of claims 1-7.

Images