A flat lock roofing quality detection system based on deep learning

Abstract

The present application relates to the technical field of construction quality, and more particularly to a flat lock roofing quality detection system based on deep learning, which comprises a feature parameter module that determines a thermal inertia anomaly index based on the temperature difference changes of a plurality of roofing tile lock areas in different time periods in a roofing area, and determines the lock acoustic characteristics of a key frequency band based on the pick-up signals of the lock areas; a region type module that determines the single feature deviation of each feature based on a plurality of feature parameters, and determines the region type of the roofing area in combination with the joint feature deviation; and a negative feedback module that estimates the actual wind lifting resistance of the lock based on the combination of risk, the stiffness decline ratio and the wind speed correction coefficient, so as to determine the abnormal threshold of the single feature deviation for adjusting the wind lifting resistance of the flat lock roofing. The present application realizes non-contact rapid screening through the thermal inertia anomaly index and the lock acoustic fingerprint, and reversely adjusts the detection threshold in combination with the wind lifting resistance evaluation, thereby improving the accuracy and reliability of the flat lock roofing quality detection.

Description

Technical Field

[0001] This invention relates to the field of construction quality technology, and in particular to a deep learning-based flat-lock roof quality inspection system. Background Technology

[0002] Flat-lock metal roofing systems are increasingly widely used in modern high-end buildings such as airports, stadiums, museums, commercial complexes, and historical building renovations due to their superior waterproofing performance, smooth and aesthetically pleasing appearance, and ability to adapt to various complex shapes. This system typically uses high-performance materials such as aluminum-magnesium-manganese alloys, forming a seamless roof through precise, concealed interlocking panels. This achieves a non-perforated, concealed installation, significantly reducing the risk of leaks.

[0003] This complex connection method demands extremely high construction precision. Currently, in the construction and acceptance stages of flat-lock roofs, quality inspection mainly relies on manual visual inspection and tactile examination. The most common method involves quality inspectors or construction workers closely observing the roof surface or controlling the process during construction, touching the locking parts to check for flatness, using simple tools like feeler gauges to check gaps, and testing various performance aspects of the installed metal roof tiles. This method is not only inefficient and labor-intensive, but the test results are also highly susceptible to subjective factors and lack objective and unified standards. Furthermore, the tests are often destructive or sampling-based, making it impossible to achieve full inspection of the locking quality on large roof areas. For large roof areas or high-altitude locations, personnel have difficulty accessing them, resulting in blind spots in the inspection.

[0004] The limitations of conventional measuring tools mean that while theodolites, levels, and vernier calipers are used during construction layout and acceptance to check purlin elevations and panel dimensions, these methods are primarily used to control the geometry of the base layer and panels. They cannot directly and effectively assess whether the core connection node, the "locking joint," is fully and tightly engaged.

[0005] In summary, existing technologies lack an efficient, objective, and non-destructive method for quality inspection of flat-lock roof systems that rely on precise mechanical interlocking. Hidden defects such as incomplete locking or excessive gaps in the interlocking mechanism are difficult to detect in a timely manner, often leading to serious quality problems later on, such as leaks and reduced wind uplift resistance. Furthermore, once the roof is sealed, rework and repairs are extremely costly.

[0006] Therefore, how to provide an automated testing system that can quickly, accurately, and non-destructively test the engagement quality of flat-lock roof locks, and promptly identify problems and guide on-site rectification during construction has become a technical challenge that urgently needs to be solved by those skilled in the art.

[0007] Chinese Patent Publication No. CN120495279A discloses an AI-based metal roofing construction quality inspection system. The system includes: an angle abrupt change identification module, a tangential anomaly extraction module, a curvature stability identification module, a path fitting offset analysis module, and a defect type determination module. In this invention, by constructing a sequence of angle change rates at boundary nodes, abrupt change points at metal roofing splicing points are accurately identified, effectively revealing local structural change trends. Combining the tangential displacement sequence and angle abrupt change characteristics between adjacent boundary nodes, sections with unbalanced directional continuity are captured, allowing abnormal directional variations to exhibit a dynamic evolution trajectory. Furthermore, the stability of the curvature distribution state of boundary nodes is determined, and differential analysis identifies concentrated curvature fluctuation areas and extreme points, enhancing the extraction of geometric representations of abnormal laying accuracy. The system also possesses the ability to extract residuals of abnormal trends during temporal evolution. However, the AI-based metal roofing construction quality inspection system exhibits the following problems: Existing technologies cannot simultaneously utilize thermal response characteristics and acoustic fingerprint features to quickly screen and accurately identify hidden defects such as gaps, insufficient engagement depth, and macroscopic loosening in the roof locking area of ​​flat locking systems through non-contact methods. Summary of the Invention

[0008] To address this, the present invention provides a deep learning-based flat-lock roof quality inspection system to overcome the problem in existing technologies that cannot quickly screen and accurately identify hidden defects such as gaps, insufficient engagement depth, and macroscopic loosening in the locking area of ​​flat-lock roofs using thermal response characteristics and acoustic fingerprint features.

[0009] To achieve the above objectives, the present invention provides a deep learning-based flat interlocking roof quality inspection system, comprising: The feature parameter module is used to determine the thermal inertia anomaly index based on the temperature difference changes of several roof tile locking areas at different times within the roof area, and to determine the locking acoustic characteristics of key frequency bands based on the sound pickup signals of the locking areas. The region type module is used to determine the single feature deviation of each feature based on several feature parameters, and to determine the region type of the roof region by combining the joint feature deviation. The defect type module is used to determine the defect type of the roof tiles in the roof area based on each characteristic parameter in response to the area type of the roof area, determine the defect level of the locking area by obtaining any defect characteristic parameter, mark the roof tiles, or determine the degree of impact of the defect on the mechanical performance of the locking mechanism. Combined with the risk module, it is used to determine the bonding risk of the roof tile locking area by obtaining the core mechanical parameters and defect characteristic parameters, and to adjust the re-inspection cycle. The core mechanical parameters include the initial core stiffness and residual displacement. The negative feedback module is used to estimate the actual wind uplift resistance of the latch based on the combined risk stiffness reduction ratio and wind speed correction coefficient, so as to determine the abnormal threshold of the single feature deviation of the wind uplift resistance adjustment of the flat latch roof.

[0010] Furthermore, the characteristic parameters include the thermal inertia anomaly index and the latch acoustic characteristics; The thermal inertia anomaly index is within the normal range of thermal inertia, the characteristic parameter module indicates that the heat conduction is normal, and the locking mechanism is tightly engaged; The thermal inertia anomaly index exceeds the normal range of thermal inertia. The characteristic parameter module judges that the thermal inertia has increased abnormally, indicating that there is a gap or looseness in the latch.

[0011] Furthermore, the key frequency bands include low-frequency characteristic bands and high-frequency characteristic bands, and the latching acoustic features include spectral energy ratio and spectral centroid shift; The spectrum energy ratio of the health lock is within the energy ratio threshold range, and the absolute value of the spectrum centroid offset is less than the standard centroid offset threshold. The spectral energy ratio of the loosened latch is less than the lower limit of the threshold, the spectral centroid shift is less than the low-frequency threshold of the centroid shift, and the centroid shifts towards the lower frequency. If the spectral energy ratio of the overtight latch is greater than the upper limit of the threshold, the spectral centroid shift is greater than the high-frequency centroid shift threshold, and the centroid shifts towards higher frequencies.

[0012] Furthermore, the region type module determines the single feature deviation of each feature based on the mean and standard deviation of each feature parameter; If the deviation of any single feature parameter is greater than or equal to the first anomaly threshold, the region type module determines that the region type of the roof area is a defect region. If the deviation of any single feature parameter is less than the first abnormal threshold and greater than or equal to the second abnormal threshold, the area type module determines that the area type of the roof area is a sub-healthy area. If the deviation of a single feature parameter is less than the second anomaly threshold, the region type module obtains the joint feature deviation of several feature parameters to determine the region type.

[0013] Furthermore, if the joint feature deviation is greater than or equal to the joint anomaly threshold, the region type module determines that the region type of the roof area is a sub-healthy region. If the deviation of the joint features is less than the joint anomaly threshold, the region type module determines that the region type of the roof area is a healthy region.

[0014] Furthermore, the defect characteristic parameters include the maximum actual clearance, the engagement depth deviation, and the relative stiffness deviation; The area type of the roof area is a sub-healthy area. If the thermal inertia anomaly index exceeds the normal range of thermal inertia, the defect type module determines the defect type as thermal anomaly. There are gaps between the roof tiles corresponding to the locking area. The maximum value of the actual gap between the locking areas of the roof tiles is obtained. If the spectral energy ratio is less than the lower limit of the threshold, or the spectral centroid offset is less than the low-frequency threshold of the centroid offset, the defect type module determines the defect type as acoustic anomaly, indicating insufficient interlocking depth between roof tiles, and obtains the actual interlocking depth inside the locking area to determine the interlocking depth deviation. If the defect type is thermal anomaly or acoustic anomaly, the defect type module determines that the locking area between the roof tiles is macroscopically loose, applies a fixed load to obtain the initial stiffness, and determines the relative stiffness deviation.

[0015] Furthermore, the defect type module determines the defect level of the latching area based on the defect feature parameters in response to the defect type; If the defect feature parameters are within the defect threshold range, the defect type module determines that the defect level of the locking area exceeds the standard and marks the roof tiles as needing to be reworked. If the defect characteristic parameters are not within the defect threshold range, the defect type module determines that the defect level of the latch area has not exceeded the standard and determines the degree of impact of the defect on the mechanical performance of the latch.

[0016] Furthermore, the core mechanical parameters include initial core stiffness and residual displacement; If the defect characteristic parameters meet the conditions for minor defects, and the initial core stiffness exceeds the stiffness threshold range and the residual displacement is lower than the displacement threshold range, then the locking area of ​​the roof tile is deemed qualified in conjunction with the risk module. If the defect characteristic parameters meet the conditions for a medium defect, and the initial core stiffness is within the stiffness threshold range and the residual displacement is within the displacement threshold range, then the risk module determines that there is a risk of connection in the locking area of ​​the roof tile, issues a repair signal, and shortens the re-inspection cycle.

[0017] Furthermore, the slight defect condition is that the defect feature parameter is not in the critical threshold range, and the moderate defect condition is that the defect feature parameter is in the critical threshold range.

[0018] Furthermore, the negative feedback module determines the stiffness reduction ratio based on the initial core stiffness; If the ratio of the actual wind uplift resistance to the designed wind uplift bearing capacity is less than the first evaluation threshold but greater than the second evaluation threshold, the negative feedback module determines that the flat-locked roof has a risk of wind uplift and adjusts the second abnormal threshold. If the ratio of the actual wind uplift resistance to the designed wind uplift bearing capacity is less than the second evaluation threshold, the negative feedback module determines that the wind uplift resistance of the flat-locked roof is insufficient and issues a roof quality warning signal.

[0019] Compared with existing technologies, the advantages of this invention are as follows: The flat-locking system, as a traditional metal roofing system, uses a flat-locking interlocking system for the roof tiles, featuring a concealed installation design with no perforations or punctures on the surface. This invention achieves rapid, non-contact, and non-destructive macroscopic screening by integrating the thermal inertia anomaly index and the acoustic fingerprint of the lock through a feature parameter module. A region type module accurately divides healthy, sub-healthy, and defective regions based on single-feature deviation and joint-feature deviation. A defect type module quantitatively verifies gap values, interlocking depth deviations, and relative stiffness deviations in sub-healthy regions, identifying hidden defects such as incomplete lock placement and excessive interlocking gaps. A risk module obtains initial core stiffness and residual displacement, assessing the impact of defects on mechanical performance and adjusting the re-inspection cycle. A negative feedback module estimates actual wind uplift resistance based on stiffness reduction ratio and wind speed correction coefficient, and adjusts the abnormal threshold of single-feature deviation in reverse, forming a self-optimizing closed-loop system for detection parameters. This provides precise quantitative and continuously optimized decision support for roof quality and safety assessment.

[0020] Furthermore, this invention introduces acoustic fingerprinting into latch quality detection, utilizing ambient wind as a natural excitation source to achieve non-contact, non-destructive, and artificially induced state perception. The acoustic features are extremely sensitive to the microscopic contact state of the latch, enabling the detection of anomalies in the early stages of loosening. This invention calculates the thermal inertia anomaly index by acquiring the latch area's temperature in the afternoon and early morning, quantifying the latch's thermal response characteristics and accurately determining whether the latch has defects such as gaps, looseness, or excessive tightness. By acquiring the acoustic signal from the latch area, the spectral energy ratio and spectral centroid shift are extracted as the latch's acoustic features, allowing for the detection of anomalies in the early stages of loosening. Through the region type module, the roof area is divided into healthy, sub-healthy, and defective regions based on single-feature deviation and joint-feature deviation, achieving rapid macroscopic screening and providing precise guidance for subsequent targeted testing, significantly improving the efficiency and accuracy of roof quality inspection.

[0021] Furthermore, this invention targets sub-healthy areas and uses a defect type module to accurately identify defect types based on thermal inertia anomaly index and latch acoustic fingerprint. For thermal anomalies, laser ranging is used to obtain the actual gap value; for acoustic anomalies, line-of-sight detection is used to obtain the interlocking depth deviation; and for both anomalies, micro-force loading is used to obtain the relative stiffness deviation. This achieves quantitative verification of hidden defects such as latch misalignment and excessive interlocking gap. By comparing defect characteristic parameters with defect threshold ranges, it quickly determines whether the defect level exceeds the standard. For latches that do not exceed the standard, core mechanical parameters such as initial core stiffness and residual displacement are further obtained. Combined with the critical threshold range where the defect characteristic parameters are located, the latch area is divided into qualified connection or connection risk, and corresponding repair signals and re-inspection cycle suggestions are issued. This invention constructs a complete verification chain from defect type identification to defect level determination to mechanical performance evaluation, providing accurate and quantitative decision-making basis for roof quality inspection.

[0022] Furthermore, this invention addresses the locking areas with potential bonding risks by using a negative feedback module to estimate the actual wind uplift resistance based on the stiffness reduction ratio and wind speed correction coefficient. This quantifies geometric defects into functional impact, achieving a deep assessment from "whether a defect exists" to "whether it affects safe use." Based on the ratio of the actual wind uplift resistance to the designed wind uplift bearing capacity, it determines whether the wind uplift resistance is sufficient or if there is a risk of wind uplift and automatically adjusts the second anomaly threshold to optimize the sensitivity of the first-layer screening. This invention constructs a closed-loop feedback mechanism for defect detection and wind uplift resistance assessment. By using the actual wind uplift resistance to calibrate the detection threshold, the system continuously optimizes itself during use, providing a quantitative and dynamic decision-making basis for roof quality and safety assessment. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of a unit in the flat-lock roof quality inspection system of this invention. Figure 2 This is a plan view of the tile arrangement in an embodiment of the present invention; Figure 3 This is a purlin arrangement diagram in an embodiment of the present invention; Figure 4 This is a diagram showing the arrangement of roof tiles in an embodiment of the present invention; In the diagram: 1-Main roof purlin, 2-Secondary roof purlin, 3-Roof tile. Detailed Implementation

[0024] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0025] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0026] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0027] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0028] Please see Figures 1-4 As shown, Figure 1 This is a schematic diagram of a unit in the flat-lock roof quality inspection system of this invention. Figure 2 This is a plan view of the tile arrangement in an embodiment of the present invention; Figure 3 This is a purlin arrangement diagram in an embodiment of the present invention; Figure 4 This is a diagram showing the arrangement of roof tiles in an embodiment of the present invention.

[0029] This invention provides a deep learning-based quality inspection system for flat interlocking roofs, comprising: The feature parameter module is used to determine the thermal inertia anomaly index based on the temperature difference changes of several roof tile locking areas at different times within the roof area, and to determine the locking acoustic characteristics of key frequency bands based on the sound pickup signals of the locking areas. The region type module is used to determine the single feature deviation of each feature based on several feature parameters, and to determine the region type of the roof region by combining the joint feature deviation. The defect type module is used to determine the defect type of the roof tiles in the roof area based on each characteristic parameter in response to the area type of the roof area, determine the defect level of the locking area by obtaining any defect characteristic parameter, mark the roof tiles, or determine the degree of impact of the defect on the mechanical performance of the locking mechanism. Combined with the risk module, it is used to determine the bonding risk of the roof tile locking area by obtaining the core mechanical parameters and defect characteristic parameters, and to adjust the re-inspection cycle. The core mechanical parameters include the initial core stiffness and residual displacement. The negative feedback module is used to estimate the actual wind uplift resistance of the latch based on the combined risk stiffness reduction ratio and wind speed correction coefficient, so as to determine the abnormal threshold of the single feature deviation of the wind uplift resistance adjustment of the flat latch roof.

[0030] Specifically, the flat-locking system, as a traditional metal roofing system, uses a flat-locking interlocking system for the roof tiles, featuring a concealed installation design with no perforations or punctures on the surface. This invention achieves rapid, non-contact, and non-destructive macroscopic screening by integrating the thermal inertia anomaly index and the acoustic fingerprint of the lock through a feature parameter module. A region type module accurately divides healthy, sub-healthy, and defective regions based on single-feature deviation and combined-feature deviation. A defect type module quantifies and verifies gap values, interlocking depth deviations, and relative stiffness deviations in sub-healthy regions, identifying hidden defects such as incomplete lock placement and excessive interlocking gaps. A risk module obtains initial core stiffness and residual displacement, assessing the impact of defects on mechanical performance and adjusting the re-inspection cycle. A negative feedback module estimates actual wind uplift resistance based on stiffness reduction ratio and wind speed correction coefficient, and adjusts the abnormal threshold of single-feature deviation in reverse, forming a self-optimizing closed-loop system for detection parameters. This provides precise quantitative and continuously optimized decision support for roof quality and safety assessment.

[0031] As a traditional metal roofing system, the flat-locking system uses a flat-locking interlocking system for the roof tiles. It features a concealed installation design with no perforations or punctures on the surface and is fixed with stainless steel fasteners.

[0032] In this embodiment, the metal roofing tiles used in this project are 0.9mm thick flat-locking aluminum-magnesium-manganese alloy roofing tiles. Tile A has a metal tile length of 351mm and an exposed length of 330mm; tile B has a metal tile length of 271mm and an exposed length of 250mm; the width of the roofing metal tiles is 245mm, and metal tiles with an exposed length of 250mm are used at the roof corners, which is achieved by overlapping and locking the tiles.

[0033] The purlins are installed before the roof tiles 3 are installed; the main roof purlin 1 is... Rectangular galvanized steel pipe, roof secondary purlin 2 is Galvanized square tubing; the main roof purlins are welded and fixed to the purlin brackets, and the secondary roof purlins are welded to the main roof purlins.

[0034] During implementation, the roof is divided into several roof areas, and each roof tile area is divided into locking areas and non-locking areas. The locking areas are the locking parts at the edges of the roof tiles, and the non-locking areas are the central area of ​​the roof tiles and the parts away from the locking parts.

[0035] In areas where the latches are tightly engaged, there is sufficient metal contact, resulting in high heat conduction efficiency and temperature changes synchronized with the main structure. In areas where the latches are loose or have gaps, the air layer increases thermal resistance, causing temperature changes to lag. The thermal response characteristics of the latches are quantified by the ratio of the daytime and nighttime temperature differences.

[0036] The feature parameter module uses thermal sensors to obtain the temperature of the locking area of ​​several roof tiles in any roof area during the afternoon and early morning to determine the thermal inertia anomaly index. Specifically, the thermal inertia anomaly index is used to reflect the difference in thermal response between the locking area and the non-locking area of ​​the roof panel under the day-night temperature difference conditions. It quantitatively assesses the tightness of the interlocking area of ​​the flat locking roof, judges whether there are defects such as gaps, looseness or overtightness in the locking, and can discover internal gaps and incomplete interlocking problems that cannot be directly observed by the naked eye, and issue early warnings in the early stage of defect development. The principle behind determining the thermal inertia anomaly index lies in the fact that metallic materials have excellent thermal conductivity. When the latches are tightly engaged, heat conduction between the latching area and the main body of the roof panel is smooth, and their temperature changes remain synchronized. When there are gaps or looseness in the latches, the air layer, whose thermal conductivity is much lower than that of metal, increases thermal resistance, causing the temperature change in the latching area to lag behind that of the main body of the roof panel. Using a ratio rather than an absolute temperature difference can eliminate the interference of external factors such as solar radiation intensity, ambient temperature, and roof orientation, making the index comparable across time periods and regions. Selecting the two extreme temperature times of afternoon and early morning can maximize the temperature difference signal and improve the detection sensitivity of thermal response differences. The non-latching area is used as a benchmark. The non-latching area of ​​the roof panel, such as the core part of the panel, is far from the latching interface, and its thermal response is not affected by the latching state, so it can be used as a health reference.

[0037] The thermal inertia anomaly index is within the normal range of thermal inertia, the characteristic parameter module indicates that the heat conduction is normal, and the locking mechanism is tightly engaged; The thermal inertia anomaly index exceeds the normal range of thermal inertia. The characteristic parameter module judges that the thermal inertia anomaly has increased, there is a gap or looseness in the latch, and the temperature difference change is more drastic than the normal area. The thermal inertia anomaly index is lower than the normal range of thermal inertia. The characteristic parameter module judges that the thermal inertia anomaly has decreased, which may lead to stress concentration or excessive contact due to overtight locking. During implementation, sampling points are evenly distributed in both the locking and non-locking areas, and the average temperature of the sampling points in both the locking and non-locking areas is taken as the detection temperature and the reference temperature. The thermal inertia anomaly index is the absolute value of the temperature difference between the same locking area in the afternoon and early morning divided by the absolute value of the temperature difference between the reference temperature of the same non-locking area of ​​the roof panel in the afternoon and early morning.

[0038] The normal range of thermal inertia is [0.8, 1.2], consisting of the upper and lower limits of normal thermal inertia. The theoretical ideal value of the abnormal thermal inertia index is 1.0, indicating that the temperature changes of the locking area and the non-locking area of ​​the roof panel are completely synchronized, that is, the thermal response behavior of the two is consistent under the diurnal temperature difference. At this time, the locking is tightly engaged and the heat conduction is smooth. In actual engineering, due to material differences, local environment, measurement errors and other factors, the thermal inertia index of a healthy locking will show a certain fluctuation range around 1.0. When the thermal inertia index is lower than 0.8, the stress concentration problem caused by the locking being too tight begins to appear, and long-term service may cause fatigue cracks at the root of the locking. Taking 1.2 as the upper limit ensures both sensitivity to early loosening defects and avoids frequent false alarms caused by environmental noise.

[0039] When a latch is in a locked state, it will produce a characteristic acoustic spectrum under wind excitation. When a latch is tightly locked, its fixed boundary conditions are clear, the acoustic energy is concentrated in a specific frequency band, and the spectrum shape is stable. When a latch is loose, due to changes in boundary conditions, it will produce broadband noise, and the characteristic peaks will drift or split. Essentially, this is a manifestation of the latch as a modal characteristic of a vibration system in acoustic radiation.

[0040] The feature parameter module determines the latching acoustic features of the key frequency band, which include the spectral energy ratio and the spectral centroid shift. The key frequency band includes a low-frequency feature band and a high-frequency feature band.

[0041] Specifically, the spectral energy ratio is sensitive to the microscopic state of the contact surface, while the spectral centroid shift is sensitive to changes in overall stiffness. Using both together can more accurately determine the type of defect.

[0042] The spectral energy ratio is used to determine whether the overall stiffness of the latch is normal by comparing the acoustic energy distribution of the low-frequency characteristic band and the high-frequency characteristic band. When the latch is loose, the overall stiffness decreases and the proportion of low-frequency energy decreases. At the same time, the micro gaps on the contact surface increase and the high-frequency friction noise increases. A decrease in the spectral energy ratio is a typical acoustic fingerprint of a loose latch. When the latch is too tight, the compressive stress on the contact surface increases, the high-frequency vibration energy is suppressed, and the spectral energy ratio increases. The spectral centroid shift reflects the average frequency distribution of sound energy. When the latch is loose or too tight, the boundary conditions of the vibration system change, the natural frequency changes, and the spectral centroid shifts. Loose latch leads to a decrease in overall stiffness, a decrease in the natural frequency of the vibration system, and a shift of the spectral centroid towards lower frequencies. Too tight latch leads to an increase in compressive stress on the contact surface, an increase in local stiffness, and a shift of the spectral centroid towards higher frequencies.

[0043] Understandably, the denominator of the spectral energy ratio is high-frequency energy, which mainly originates from the microscopic friction and collision of the contact surfaces. When the latch loosens, the gap between the contact surfaces increases, the relative motion intensifies, the high-frequency energy rises, and the denominator increases. The numerator of the spectral energy ratio is low-frequency energy, which mainly originates from the vibration response of the latch as an integral structure. When the latch loosens, the connection stiffness decreases, the overall vibration mode frequency decreases, and the amplitude may change. However, more importantly, the relatively significant increase in high-frequency energy leads to a decrease in the ratio. Using an energy ratio instead of an absolute energy value can eliminate the influence of factors such as wind speed changes and differences in excitation location, making the features comparable across time periods and locations. The centroid of the spectrum essentially reflects the location of concentrated sound energy. The natural frequency of a vibrating system is proportional to the square root of its stiffness. When the latch is loose, the stiffness decreases, the natural frequency decreases, the sound energy concentrates at lower frequencies, and the centroid of the spectrum shifts downward. When the latch is too tight, the local stiffness increases, the natural frequency increases, and the centroid of the spectrum shifts upward. Using the offset relative to a healthy latch rather than the absolute centroid value can eliminate the influence of inherent differences such as latch type, plate shape, and installation method, making the test results universal across projects.

[0044] The spectrum energy ratio of the health lock is within the energy ratio threshold range, and the absolute value of the spectrum centroid offset is less than the standard centroid offset threshold. The spectral energy ratio of the loosened latch is less than the lower limit of the threshold, the spectral centroid shift is less than the low-frequency threshold of the centroid shift, and the centroid shifts towards the lower frequency. If the spectral energy ratio of the overtight latch is greater than the upper limit of the threshold, the spectral centroid shift is greater than the high-frequency centroid shift threshold, and the centroid shifts towards higher frequencies.

[0045] Specifically, the feature parameter module acquires the acoustic signals of the locking areas of several roof tiles on the roof panel through a microphone array, performs bandpass filtering of 20-2000Hz to remove low-frequency wind noise and high-frequency irrelevant noise, performs frame-by-frame windowing with a frame length of 0.1 seconds and a frame shift of 0.05 seconds, and determines the locking acoustic characteristics of key frequency bands, including a low-frequency feature band of 150-350Hz and a high-frequency feature band of 800-1200Hz. During implementation, the spectral energy ratio ; In the formula and These represent the upper and lower limits of the low-frequency characteristic band. and These represent the upper and lower limits of the high-frequency characteristic band. The power spectral density is denoted as . The spectral energy ratio reflects the stiffness distribution characteristics of the latch. When the latch is loosened, the low-frequency energy decreases and the high-frequency energy increases, resulting in a decrease in the spectral energy ratio. Specifically, the acoustic energy in the low-frequency characteristic band reflects the overall stiffness of the latch; the higher the stiffness, the greater the proportion of low-frequency energy. The acoustic energy in the high-frequency characteristic band reflects the microscopic state of the latch's contact surface; the tighter the contact, the smaller the proportion of high-frequency energy.

[0046] The characteristic parameter module reflects the average frequency distribution of acoustic energy through the spectral centroid. Actual Spectral Centroid The spectral centroid offset is determined based on the difference between the actual spectral centroid and the standard spectral centroid of the latch, where the standard spectral centroid of the latch is obtained through laboratory calibration.

[0047] Specifically, the laboratory calibration method for the standard spectral centroid is as follows: acoustic signals are collected for roof tile specimens under three conditions: wind speeds of 3 m / s, 4 m / s, and 5 m / s. Ten sets of data are collected under each condition, with each set lasting 5 seconds, a sampling rate of ≥44.1 kHz, and a resolution of ≥16 bits. Each set of data is bandpass filtered, divided into frames, and windowed. The power spectral density of each frame is calculated to determine the spectral centroid of each frame. The average value of the spectral centroids of all frames is taken as the spectral centroid of the specimen under that wind speed. The average value of the spectral centroids of five ideal compact specimens under all wind speeds is taken as the standard spectral centroid.

[0048] The energy ratio threshold range is [0.8, 1.2], the lower threshold is 0.6, the upper threshold is 1.5, the standard threshold for centroid offset is 15Hz, the low-frequency threshold for centroid offset is -20Hz, and the high-frequency threshold for centroid offset is 20Hz.

[0049] Understandably, the theoretical ideal value for the energy ratio is 1.0, with a normal fluctuation range of approximately ±20% of the mean. When the energy ratio is below 0.8, high-frequency energy begins to rise relatively significantly, indicating that the microscopic gaps at the contact surface are starting to have an impact. When the energy ratio is above 1.2, low-frequency energy is relatively dominant, indicating that the contact surface may be too tight. At the same time, a lower threshold of 0.6 indicates that the latch has entered a significantly loose state, with high-frequency energy reaching more than 1.67 times that of low-frequency energy, indicating that microscopic friction and collision at the contact surface have become dominant, and the restraining function of the latch has been significantly weakened. Its wind uplift resistance has usually dropped to below 70% of the design value, posing a safety hazard that requires immediate attention. An upper threshold of 1.5 indicates that the latch has entered a significantly overtight state, with low-frequency energy reaching more than 1.5 times that of high-frequency energy. Excessive compressive stress on the latch contact surface inhibits high-frequency microscopic vibration, leading to stress concentration at the latch root. Under long-term wind vibration and temperature cycling, the risk of fatigue cracks at the latch root increases significantly, requiring immediate attention.

[0050] A centroid offset threshold of 15Hz can detect early loosening and early overtightening. This threshold avoids frequent false alarms caused by factors such as wind speed changes and environmental noise. A centroid offset exceeding ±15Hz indicates that the boundary conditions of the latch have undergone measurable changes and require further testing and confirmation. A centroid offset <-20Hz indicates that the centroid has shifted more than 20Hz towards lower frequencies, indicating a significant loosening defect in the latch. A 20Hz frequency drop corresponds to approximately an 8-10% drop in the natural frequency, indicating a significant reduction in stiffness. A centroid offset >+20Hz indicates that the centroid has shifted more than 20Hz towards higher frequencies, indicating a significant overtightening defect in the latch. Overtightening of the latch leads to increased compressive stress on the contact surface and an increase in local contact stiffness, which may cause the frequency of some high-frequency modes to rise by 20-40Hz, causing the centroid to shift upwards.

[0051] The region type module measures thermal inertia and acoustic fingerprints under different engagement states of standard locking specimens to obtain the characteristic thresholds of characteristic parameters. The characteristic parameters include thermal inertia anomaly index, spectral energy ratio, and spectral centroid shift. Different engagement states include tight, 10% loose, 20% loose, and too tight. The region type module determines the average and standard deviation of various characteristic parameters of several roof tiles within a roof region; For any roof tile, the region type module determines the single feature deviation of each feature based on the mean and standard deviation of each feature parameter; In practice, the single feature deviation is determined based on the absolute value of the ratio of the feature parameter difference to the corresponding standard deviation, where the feature parameter difference is the difference between any feature parameter and its corresponding average value.

[0052] Specifically, the single feature deviation represents the degree to which the buckle under test deviates from the mean of a healthy buckle in each feature parameter, expressed in standard deviation, and is used to determine whether there is an abnormality in the corresponding feature.

[0053] If the deviation of any single feature parameter is greater than or equal to the first anomaly threshold, the region type module determines that the region type of the roof area is a defect region. If the deviation of any single feature parameter is less than the first abnormal threshold and greater than or equal to the second abnormal threshold, the area type module determines that the area type of the roof area is a sub-healthy area. If the single feature deviation of any feature parameter is less than the second anomaly threshold, the region type module determines the joint feature deviation of several feature parameters. It is understandable that when the deviation of all individual features does not reach the abnormal threshold, there may be a combined anomaly, that is, each feature does not exceed the standard, but multiple features deviate in the abnormal direction at the same time.

[0054] Specifically, the joint feature deviation is the square root of the sum of the squares of the deviations of several individual features, and the joint feature deviation is the Euclidean distance between the deviations of each feature, reflecting the overall degree of joint deviation of multiple features.

[0055] It is understandable that when all individual feature deviations do not reach the abnormal threshold, there may be a combined anomaly. Each feature does not exceed the standard, but multiple features deviate in the abnormal direction at the same time. The joint feature deviation can effectively identify this hidden combined anomaly pattern. The joint feature deviation comprehensively reflects the overall deviation of the latch in the multi-dimensional feature space, avoiding the one-sidedness of single feature judgment.

[0056] If the deviation of the joint features is greater than or equal to the joint anomaly threshold, the region type module determines that the region type of the roof area is a sub-healthy region. If the deviation of the joint features is less than the joint anomaly threshold, the region type module determines that the region type of the roof area is a healthy region. Wherein, the first abnormal threshold is 3.0, the second abnormal threshold is 2.5, and the combined abnormal threshold is 3.5.

[0057] Understandably, when characteristic parameters such as thermal inertia anomaly index, spectral energy ratio, and spectral centroid shift deviate within the range of 2.5σ-3.0σ, the latch usually has slight defects. The first anomaly threshold of 3.0 means that only when the characteristic deviation is extremely significant will it be directly judged as a defect area, ensuring that the judgment of defect areas has a very high confidence level and avoiding misjudging healthy latches as defect areas. The interval between the first anomaly threshold of 3.0 and the second anomaly threshold of 2.5 provides a sufficient buffer space, making the classification clearer. Setting the 2.5-3.0 interval as the sub-healthy area ensures sufficient sensitivity while avoiding the risk of false alarms caused by directly judging it as a defect. For three independent standard normal distribution variables, their joint deviation, i.e., the Euclidean distance, follows a chi-square distribution with 3 degrees of freedom. When the deviation of all individual features is less than 2.5 but the joint deviation is ≥3.5, a typical combination is [2.0, 2.0, 2.0], and its Euclidean distance is about 3.46, close to 3.5. At this time, each feature is not out of bounds when viewed alone, but the three features deviate from the abnormal direction at the same time, and the joint deviation indicates that there is a combined anomaly. The joint anomaly threshold is set to 3.5 rather than a lower value to avoid a large number of healthy areas being misjudged as sub-healthy areas due to slight combined deviation.

[0058] Specifically, this invention introduces acoustic fingerprinting into latch quality detection, utilizing ambient wind as a natural excitation source to achieve non-contact, non-destructive, and artificially induced state perception. The acoustic features are extremely sensitive to the microscopic contact state of the latch, enabling the detection of anomalies in the early stages of loosening. This invention calculates the thermal inertia anomaly index by acquiring the latch area's temperature in the afternoon and early morning, quantifying the latch's thermal response characteristics and accurately determining whether the latch has defects such as gaps, looseness, or excessive tightness. By acquiring the acoustic signal from the latch area, the spectral energy ratio and spectral centroid shift are extracted as the latch's acoustic features, allowing for the early detection of anomalies in the early stages of loosening. Through a region type module, the roof area is divided into healthy, sub-healthy, and defective regions based on single-feature deviation and joint-feature deviation, achieving rapid macroscopic screening and providing precise guidance for subsequent targeted testing, significantly improving the efficiency and accuracy of roof quality inspection.

[0059] The roof area is classified as a sub-healthy area. Based on various characteristic parameters, the defect type of the roof tiles within the roof area is determined. If the thermal inertia anomaly index exceeds the normal range of thermal inertia, the defect type module determines the defect type as thermal anomaly. There are gaps between the roof tiles corresponding to the locking area. The actual gap values ​​at both ends and in the middle of the locking area between the roof tiles are obtained by a laser rangefinder. If the spectral energy ratio is less than the lower limit of the threshold, or the spectral centroid offset is less than the low-frequency threshold of the centroid offset, the defect type module determines the defect type as acoustic anomaly, indicating insufficient interlocking depth between roof tiles. The actual interlocking depth inside the locking area is obtained through line-of-sight detection to determine the interlocking depth deviation. If the defect type is thermal or acoustic, the defect type module determines that the locking area between the roof tiles is macroscopically loose, applies a fixed load, and uses a displacement sensor to obtain the initial stiffness and determine the relative stiffness deviation.

[0060] Specifically, if the defect type is thermal anomaly, the defect type module obtains the maximum value of the actual gap between the two ends and the middle of the locking area as the maximum actual gap value; if the defect type is acoustic anomaly, the defect type module determines the difference between the designed engagement depth and the actual engagement depth as the engagement depth deviation; the defect type module applies a fixed load, such as 50N, and measures the initial displacement generated in the locking area. The initial stiffness is determined based on the ratio of the fixed load to the initial displacement, and the relative stiffness deviation is determined based on the percentage value of the difference between the initial stiffness and the standard stiffness to the standard stiffness. In practice, the design engagement depth is usually 3-5 times the plate thickness. The selectable value range of the design engagement depth is 2.7mm-4.5mm, preferably 3.5mm. The standard stiffness can be selected from 800N / mm to 1500N / mm depending on the type of latch and the plate thickness.

[0061] The defect type module responds to the defect type by determining the defect level of the locking area based on defect feature parameters, which include the maximum actual clearance, engagement depth deviation, and relative stiffness deviation. It is understandable that the high thermal inertia anomaly index is essentially due to the presence of an air layer in the locking area, leading to increased thermal resistance and abnormal thermal response. The direct source of the air layer is the gap between the locking surfaces; therefore, the gap is the physical essence of the thermal anomaly. When the gap exceeds 0.2mm, the effective engagement depth of the locking is significantly insufficient, resulting in a significant decrease in wind resistance. The low spectral energy ratio or the shift of the spectral centroid to lower frequencies is essentially due to a decrease in the overall stiffness of the locking. For flat locking systems, the most common cause of stiffness reduction is that the locking is not fully engaged, i.e., the actual engagement depth is less than the designed engagement depth. Therefore, engagement depth deviation is a direct physical verification parameter for acoustic anomalies. When the engagement depth deviation exceeds 0.5mm, it indicates... This means that the actual engagement depth is less than 80% of the design value. Taking the design depth of 2.7mm as an example, the effective constraint of the latch has been significantly weakened. When thermal and acoustic abnormalities occur simultaneously, it indicates that the latch has both gap thermal abnormalities and stiffness reduction acoustic abnormalities. This usually means macroscopic loosening. The latch has lost its normal connection constraint, and the overall mechanical performance is damaged. Stiffness is the most direct and comprehensive mechanical indicator for measuring the connection constraint of the latch. The standard stiffness of latches of different specifications is different. The percentage deviation relative to the standard stiffness is used to make the test results universal. A 15% decrease in stiffness is the engineering boundary where the latch goes from slight loosening to obvious loosening. At this time, the latch's wind resistance has usually dropped to below 85% of the design value.

[0062] If the defect feature parameters are within the defect threshold range, the defect type module determines that the defect level of the locking area exceeds the standard and marks the roof tiles as needing to be reworked. If the defect characteristic parameters are not within the defect threshold range, the defect type module determines that the defect level of the latch area has not exceeded the standard and determines the degree of impact of the defect on the mechanical performance of the latch. Among them, the defect threshold ranges for the maximum actual gap, the bite depth deviation, and the relative stiffness deviation are (0.2, +∞) mm, (0.5, +∞) mm, and (-∞, -15%), respectively.

[0063] For locking areas where the defect level does not exceed the standard, the impact of roof tile defects on the mechanical performance of the locking mechanism is determined by combining the core mechanical parameters obtained from the risk module. The core mechanical parameters include the initial core stiffness and residual displacement. Specifically, controllable loads of 0N-150N are gradually applied to the roof tiles and then unloaded. The first and second load displacements with controllable loads of 50N and 100N are obtained by combining the risk module to determine the initial core stiffness. The initial core stiffness during implementation is determined based on the ratio of the controllable load difference and the load displacement difference, wherein the controllable load difference is the difference between 100N and 50N, and the load displacement difference is the difference between the second load displacement and the first load displacement.

[0064] Specifically, the initial core stiffness is used to quantitatively assess the locking area's resistance to deformation during the elastic deformation stage; the greater the decrease in stiffness, the lower the lock's remaining wind uplift resistance; the essential function of the lock is to transfer loads and constrain roof panel displacement, and its core mechanical performance index is connection stiffness. Stiffness reflects the lock's ability to resist deformation and is a direct measure of connection quality; geometric defects such as gaps and insufficient engagement depth will ultimately manifest as a decrease in stiffness; therefore, by measuring stiffness, the actual impact of geometric defects on mechanical performance can be directly assessed.

[0065] The irrecoverable displacement after unloading to 0N obtained by the risk module is recorded as the residual displacement. The quality inspection rating and maintenance recommendations for the locking area are determined by combining the initial core stiffness and defect characteristic parameters. If the defect characteristic parameters meet the conditions for minor defects, and the initial core stiffness exceeds the stiffness threshold range and the residual displacement is lower than the displacement threshold range, then the locking area of ​​the roof tile is deemed qualified in conjunction with the risk module. If the defect characteristic parameters meet the conditions for a medium defect, and the initial core stiffness is within the stiffness threshold range and the residual displacement is within the displacement threshold range, then the risk module is used to determine that there is a risk of connection in the locking area of ​​the roof tile, and a repair signal is issued and the re-inspection cycle is shortened. Specifically, the minor defect condition is that the defect feature parameters are not in the critical threshold range, and the moderate defect condition is that the defect feature parameters are in the critical threshold range.

[0066] The initial stiffness threshold ranges from 0.70 to 0.85 times the initial stiffness reference, the displacement threshold ranges from 0.08 to 0.15 mm, the selectable value range for the initial stiffness reference is 800 N / mm to 1500 N / mm, and the critical threshold ranges for the maximum actual clearance, the engagement depth deviation, and the relative stiffness deviation are 0.1-0.2 mm, 0.2-0.5 mm, and 5%-15%, respectively.

[0067] Understandably, the displacement threshold range reflects the irreversible deformation after the latch is unloaded. A healthy latch typically has a residual displacement ≤0.05mm after a 150N load-unload cycle. 0.08mm is taken as the boundary between healthy and slightly damaged, leaving a safety margin. When the residual displacement begins to exceed 0.08mm, it indicates that the latch material has begun to show micro-yield, an early signal of plastic deformation. A residual displacement of 0.15mm is approximately 1 / 6 of the latch plate thickness of 0.9mm. At this deformation level, the latch geometry has undergone observable changes, and the interlocking function has been affected. Latches with a residual displacement >0.15mm, even if their current load-bearing capacity is acceptable, will degrade more rapidly under long-term wind vibration and temperature cycling, and rework is recommended.

[0068] The normal clearance of a healthy locking latch is ≤0.05mm, with 0.1mm being the mean plus 2 standard deviations. Values ​​exceeding this are considered abnormal. 0.2mm is the upper limit of the engineering allowable latch engagement. When the clearance is >0.2mm, the effective engagement depth is significantly insufficient, and the wind resistance is significantly reduced. The designed engagement depth is usually 3.5mm. A deviation of 0.2mm corresponds to a deviation of about 5-6%, which exceeds the normal installation tolerance range. A deviation >0.5mm means that the actual engagement depth is less than 85% of the design value, and the latch's restraint function has been significantly weakened. The normal fluctuation range of the stiffness of a healthy locking latch is about ±5%. Values ​​exceeding this are considered abnormal. When the stiffness decreases by more than 15%, the latch may have entered the plastic stage, and the wind resistance is significantly reduced.

[0069] Specifically, this invention targets sub-optimal areas and uses a defect type module to accurately identify defect types based on thermal inertia anomaly index and latch acoustic fingerprint. For thermal anomalies, laser ranging is used to obtain the actual gap value; for acoustic anomalies, line-of-sight detection is used to obtain the interlocking depth deviation; and for both anomalies, micro-force loading is used to obtain the relative stiffness deviation. This achieves quantitative verification of hidden defects such as latch misalignment and excessive interlocking gap. By comparing defect characteristic parameters with defect threshold ranges, it quickly determines whether the defect level exceeds the standard. For latches that do not exceed the standard, core mechanical parameters such as initial core stiffness and residual displacement are further obtained. Combined with the critical threshold range where the defect characteristic parameters are located, the latch area is divided into qualified connection or connection risk, and corresponding repair signals and re-inspection cycle suggestions are issued. This invention constructs a complete verification chain from defect type identification to defect level determination to mechanical performance evaluation, providing accurate and quantitative decision-making basis for roof quality inspection.

[0070] There is a risk of failure in the locking area of ​​the roof tiles. The negative feedback module estimates the actual wind uplift resistance of the locking mechanism based on the stiffness reduction ratio and wind speed correction coefficient.

[0071] Specifically, the actual wind uplift resistance is determined based on the product of the design wind uplift bearing capacity, stiffness correction factor, and wind speed correction factor; If the ratio of the actual wind uplift resistance to the designed wind uplift bearing capacity is greater than or equal to the first evaluation threshold, the negative feedback module determines that the wind uplift resistance of the flat-locked roof is sufficient. If the ratio of the actual wind uplift resistance to the designed wind uplift bearing capacity is less than the first evaluation threshold but greater than the second evaluation threshold, the negative feedback module determines that the flat-locked roof has a risk of wind uplift and adjusts the second abnormal threshold. During implementation, the negative feedback module reduces the second abnormal threshold based on the ratio of the actual wind uplift capacity to the designed wind uplift bearing capacity and the ratio of the first evaluation threshold.

[0072] If the ratio of the actual wind uplift resistance to the designed wind uplift bearing capacity is less than the second evaluation threshold, the negative feedback module determines that the wind uplift resistance of the flat-locked roof is insufficient and issues a roof quality warning signal.

[0073] Specifically, the actual wind uplift resistance is used to assess whether the ability of interlocking devices with connection risks to resist wind loads under the current condition meets safety requirements. Stiffness reduction is positively correlated with load-bearing capacity reduction. The wind uplift resistance of interlocking devices essentially depends on their connection stiffness. Stiffness reduction means that the interlocking device's ability to resist deformation is weakened, making it more prone to excessive displacement under wind loads, leading to interlocking device detachment or roof tiles being lifted. The stiffness-load-bearing capacity conversion coefficient reflects the correspondence between stiffness and load-bearing capacity, with a value range of 0.5-1.0 corresponding to the differences in different interlocking device types. At the same time, the response of interlocking devices under wind excitation is nonlinear. The wind speed on the day of testing is usually lower than the design wind speed. At this time, the impact of the measured stiffness reduction on the load-bearing capacity may be underestimated. The wind speed correction coefficient extrapolates the test results to the design wind speed conditions, making the estimated results more conservative.

[0074] In this embodiment, the stiffness correction factor is the difference between 1 and the conversion value, where the conversion value is the product of the stiffness bearing capacity conversion factor and the stiffness reduction ratio.

[0075] In the formula, the design wind uplift bearing capacity is obtained based on the roof tile locking area in combination with the specifications or wind tunnel test, and the stiffness bearing capacity conversion coefficient is calibrated by finite element simulation, with a value range of 0.5-1.0, and an initial value of 0.7.

[0076] Among them, the stiffness reduction ratio is the percentage value of the difference between the initial stiffness reference and the initial core stiffness to the initial stiffness reference, and the wind speed correction coefficient is the minimum value between the wind speed ratio and 1.0. The wind speed ratio is the ratio of the daily average wind speed to the design wind speed, and the design wind speed is the maximum value of the daily average wind speed within three years.

[0077] The first evaluation threshold is 0.9, and the second evaluation threshold is 0.7.

[0078] Understandably, the first evaluation threshold retains a 10% margin. In the initial stage of the decrease in lock stiffness (<10%), the decrease in load-bearing capacity is less than the decrease in stiffness. Taking 0.9 as the first threshold means that a decrease in stiffness of about 14-20% is allowed without triggering a risk alarm, covering the normal fluctuation range and slight degradation stage of healthy locks. The second evaluation threshold indicates that when the wind uplift resistance decreases by more than 30%, the ultimate load-bearing capacity of the lock is close to or lower than the design wind load. Considering the randomness and uncertainty of wind speed, retaining a 30% load-bearing capacity margin means that the safety reserve is insufficient.

[0079] The stiffness bearing capacity conversion factor can be obtained by establishing a three-dimensional solid model of the flat interlocking roof tile, including the roof panel, interlocking parts, and connectors to simulate the interlocking geometry of the interlocking; by modifying the interlocking conditions to simulate different degrees of degradation, a gradually increasing uniformly distributed load is applied, from 0 to twice the design load, and the load-displacement curve is recorded. The calibration value of the stiffness bearing capacity conversion factor for each degradation condition is determined and the average value is taken.

[0080] Specifically, this invention targets locking areas with potential bonding risks. Through a negative feedback module, it estimates the actual wind uplift resistance based on the stiffness reduction ratio and wind speed correction coefficient, quantifying geometric defects into functional impact levels. This achieves a deep assessment, moving from "whether a defect exists" to "whether it affects safe use." Based on the ratio of actual wind uplift resistance to the designed wind uplift bearing capacity, it determines whether the wind uplift resistance is sufficient or if there is a risk, automatically adjusting the second anomaly threshold and optimizing the sensitivity of the first-layer screening. This invention constructs a closed-loop feedback mechanism for defect detection and wind uplift resistance assessment. By reverse-calibrating the detection threshold through actual wind uplift resistance, the system continuously optimizes itself during use, providing a quantitative and dynamic decision-making basis for roof quality and safety assessment.

[0081] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0082] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A deep learning-based flat interlocking roof quality inspection system, characterized in that, include: The feature parameter module is used to determine the thermal inertia anomaly index based on the temperature difference changes of several roof tile locking areas at different times within the roof area, and to determine the locking acoustic characteristics of key frequency bands based on the sound pickup signals of the locking areas. The region type module is used to determine the single feature deviation of each feature based on several feature parameters, and to determine the region type of the roof region by combining the joint feature deviation. The defect type module is used to determine the defect type of the roof tiles in the roof area based on each characteristic parameter in response to the area type of the roof area, determine the defect level of the locking area by obtaining any defect characteristic parameter, mark the roof tiles, or determine the degree of impact of the defect on the mechanical performance of the locking mechanism. Combined with the risk module, it is used to determine the bonding risk of the roof tile locking area by obtaining the core mechanical parameters and defect characteristic parameters, and to adjust the re-inspection cycle. The core mechanical parameters include the initial core stiffness and residual displacement. The negative feedback module is used to estimate the actual wind uplift resistance of the latch based on the combined risk stiffness reduction ratio and wind speed correction coefficient, so as to determine the abnormal threshold of the single feature deviation of the wind uplift resistance adjustment of the flat latch roof.

2. The deep learning-based flat interlocking roof quality inspection system according to claim 1, characterized in that, Characteristic parameters include thermal inertia anomaly index and latch acoustic characteristics; The thermal inertia anomaly index is within the normal range of thermal inertia, the characteristic parameter module indicates that the heat conduction is normal, and the locking mechanism is tightly engaged; The thermal inertia anomaly index exceeds the normal range of thermal inertia. The characteristic parameter module judges that the thermal inertia has increased abnormally, indicating that there is a gap or looseness in the latch.

3. The deep learning-based flat interlocking roof quality inspection system according to claim 2, characterized in that, Key frequency bands include low-frequency characteristic bands and high-frequency characteristic bands; latching acoustic features include spectral energy ratio and spectral centroid shift. The spectrum energy ratio of the health lock is within the energy ratio threshold range, and the absolute value of the spectrum centroid offset is less than the standard centroid offset threshold. The spectral energy ratio of the loosened latch is less than the lower limit of the threshold, the spectral centroid shift is less than the low-frequency threshold of the centroid shift, and the centroid shifts towards the lower frequency. If the spectral energy ratio of the overtight latch is greater than the upper limit of the threshold, the spectral centroid shift is greater than the high-frequency centroid shift threshold, and the centroid shifts towards higher frequencies.

4. The deep learning-based flat interlocking roof quality inspection system according to claim 3, characterized in that, The region type module determines the single-feature deviation of each feature based on the mean and standard deviation of each feature parameter; If the deviation of any single feature parameter is greater than or equal to the first anomaly threshold, the region type module determines that the region type of the roof area is a defect region. If the deviation of any single feature parameter is less than the first abnormal threshold and greater than or equal to the second abnormal threshold, the area type module determines that the area type of the roof area is a sub-healthy area. If the deviation of a single feature parameter is less than the second anomaly threshold, the region type module obtains the joint feature deviation of several feature parameters to determine the region type.

5. The deep learning-based flat interlocking roof quality inspection system according to claim 4, characterized in that, If the deviation of the joint feature is greater than or equal to the joint anomaly threshold, the region type module determines that the region type of the roof area is a sub-healthy region. If the deviation of the joint features is less than the joint anomaly threshold, the region type module determines that the region type of the roof area is a healthy region.

6. The deep learning-based flat interlocking roof quality inspection system according to claim 5, characterized in that, Defect characteristic parameters include the maximum actual clearance, engagement depth deviation, and relative stiffness deviation; The area type of the roof area is a sub-healthy area. If the thermal inertia anomaly index exceeds the normal range of thermal inertia, the defect type module determines the defect type as thermal anomaly. There are gaps between the roof tiles corresponding to the locking area. The maximum value of the actual gap between the locking areas of the roof tiles is obtained. If the spectral energy ratio is less than the lower limit of the threshold, or the spectral centroid offset is less than the low-frequency threshold of the centroid offset, the defect type module determines the defect type as acoustic anomaly, indicating insufficient interlocking depth between roof tiles, and obtains the actual interlocking depth inside the locking area to determine the interlocking depth deviation. If the defect type is thermal anomaly or acoustic anomaly, the defect type module determines that the locking area between the roof tiles is macroscopically loose, applies a fixed load to obtain the initial stiffness, and determines the relative stiffness deviation.

7. The deep learning-based flat interlocking roof quality inspection system according to claim 6, characterized in that, The defect type module responds to the defect type and determines the defect level of the latching area based on the defect feature parameters. If the defect feature parameters are within the defect threshold range, the defect type module determines that the defect level of the locking area exceeds the standard and marks the roof tiles as needing to be reworked. If the defect characteristic parameters are not within the defect threshold range, the defect type module determines that the defect level of the latch area has not exceeded the standard and determines the degree of impact of the defect on the mechanical performance of the latch.

8. The deep learning-based flat interlocking roof quality inspection system according to claim 7, characterized in that, The core mechanical parameters include initial core stiffness and residual displacement; If the defect characteristic parameters meet the conditions for minor defects, and the initial core stiffness exceeds the stiffness threshold range and the residual displacement is lower than the displacement threshold range, then the locking area of ​​the roof tile is deemed qualified in conjunction with the risk module. If the defect characteristic parameters meet the conditions for a medium defect, and the initial core stiffness is within the stiffness threshold range and the residual displacement is within the displacement threshold range, then the risk module determines that there is a risk of connection in the locking area of ​​the roof tile, issues a repair signal, and shortens the re-inspection cycle.

9. The deep learning-based flat interlocking roof quality inspection system according to claim 8, characterized in that, The condition for a minor defect is that the defect feature parameter is not in the critical threshold range, and the condition for a moderate defect is that the defect feature parameter is in the critical threshold range.

10. The deep learning-based flat interlocking roof quality inspection system according to claim 9, characterized in that, The negative feedback module determines the stiffness reduction ratio based on the initial core stiffness; If the ratio of the actual wind uplift resistance to the designed wind uplift bearing capacity is less than the first evaluation threshold but greater than the second evaluation threshold, the negative feedback module determines that the flat-locked roof has a risk of wind uplift and adjusts the second abnormal threshold. If the ratio of the actual wind uplift resistance to the designed wind uplift bearing capacity is less than the second evaluation threshold, the negative feedback module determines that the wind uplift resistance of the flat-locked roof is insufficient and issues a roof quality warning signal.

Images