An intelligent evaluation method and system for construction quality of a thermal insulation layer

Abstract

The application relates to the technical field of building construction quality detection, and particularly relates to a thermal insulation layer construction quality intelligent evaluation method and system, measured thermal parameters and designed thermal parameters of each detection sub-region of a target thermal insulation region are acquired, a parameter deviation value is calculated, when the deviation value does not exceed a threshold value, the overall quality grade is directly evaluated, when the deviation value exceeds the threshold value, multi-modal defect detection data of an abnormal sub-region are acquired and a defect feature vector is extracted, the defect feature vector is input into a reverse solving neural network, a construction behavior deviation vector of each construction process is inverted, construction behavior compliance degree scores of each process are calculated based on the deviation vector, and overall quality grade evaluation is performed by comprehensively considering entity quality detection results and construction behavior compliance degree scores. The application realizes reverse tracking from entity defects to construction process negligence, solves the problems that a traditional method cannot locate defect process root causes and cannot distinguish multi-cause defects, and achieves the effect of double-dimension comprehensive evaluation of entity quality and construction behavior.

Description

Technical Field

[0001] This application relates to the field of building construction quality testing technology, and in particular to an intelligent evaluation method and system for the construction quality of thermal insulation layers. Background Technology

[0002] With the continuous improvement of building energy efficiency standards, external wall insulation systems are widely used in new buildings and energy-saving renovations of existing buildings. The construction quality of the insulation layer is directly related to the building's energy-saving effect and safety. Currently, the evaluation of the construction quality of the insulation layer mainly relies on the physical quality inspection after completion. Defect information of the insulation layer is obtained through methods such as infrared thermal imaging and ultrasonic testing. Then, the physical quality of the insulation layer is graded according to the area, type, and severity of the defects.

[0003] However, existing evaluation methods have the following drawbacks: First, they can only assess the current physical quality of the insulation layer. When defects such as hollowness or cracking are detected, they cannot trace the specific construction procedures that caused the defects or the degree of deviation of the operation parameters of each procedure. In other words, they can only know that the defects exist but cannot pinpoint the faults in the procedures. Second, the same apparent defect may be caused by a variety of different construction violations. For example, hollowness may be caused by uneven application of adhesive, improper installation and compression of insulation boards, or improper installation torque of anchors. Existing methods cannot distinguish the root cause of the defects and make it difficult to conduct quantitative assessments and trace responsibilities of construction teams. Summary of the Invention

[0004] To facilitate the reverse tracing of construction process deviations from defects in the insulation layer and to achieve a comprehensive evaluation of both the physical quality and construction behavior, this application provides an intelligent evaluation method and system for the construction quality of insulation layers.

[0005] Firstly, this application provides an intelligent evaluation method for the construction quality of thermal insulation layers, which adopts the following technical solution: A method for intelligent evaluation of the construction quality of thermal insulation layers includes: Obtain the measured thermal parameters and design thermal parameters corresponding to different detection sub-regions within the target insulation area; Based on the measured thermal parameters and the design thermal parameters, the parameter deviation value is obtained; Determine whether the parameter deviation value exceeds a preset deviation threshold; If the parameter deviation value does not exceed the preset deviation threshold, the overall quality level of the target insulation area is evaluated based on the preset evaluation rules. If the parameter deviation value exceeds the preset deviation threshold, then the multimodal defect detection data of the abnormal sub-region is obtained, and feature extraction is performed on the multimodal defect detection data to obtain the defect feature vector. The defect feature vector is input into the inverse solving neural network to obtain the construction behavior deviation vector of each construction process corresponding to the abnormal sub-region; Based on the construction behavior deviation vector, the compliance score of construction behavior for each construction procedure is obtained; Based on the construction behavior compliance score and the preset evaluation rules, the overall quality level of the target insulation area is evaluated.

[0006] Optionally, if the parameter deviation value does not exceed the preset deviation threshold, then based on the preset evaluation rules, the overall quality level assessment of the target insulation area includes: If the parameter deviation value does not exceed the preset deviation threshold, the thermal parameter difference between adjacent detection sub-regions in the target insulation area is obtained; Determine whether the difference in the thermal parameters exceeds the local difference threshold; If the difference in thermal parameters does not exceed the local difference threshold, the overall quality level of the target insulation area is evaluated based on the preset evaluation rules. If the difference in thermal parameters exceeds the local difference threshold, the boundary position of the adjacent detection sub-region is obtained; Determine whether the boundary position coincides with the board seam position or the inside / outside corner position; If the boundary position coincides with the board seam position or the internal and external corner position, the difference in thermal parameters is determined to be a structural temperature difference, and the overall quality level of the target insulation area is evaluated based on the preset evaluation rules. If the boundary position does not coincide with either the board seam position or the internal / external corner position, then the quality level of each adjacent detection sub-region is assessed, and the lowest assessed quality level is taken as the overall quality level of the target insulation area. Optionally, the multimodal defect detection data includes visible light images, infrared thermal imaging spectra, array ultrasonic data, and laser three-dimensional point cloud data; the feature extraction of the multimodal defect detection data to obtain the defect feature vector includes: Cracks are identified in the visible light image to obtain crack width, crack length, and crack distribution direction; Temperature field analysis is performed on the infrared thermal imaging spectrum to obtain the infrared temperature difference amplitude and the boundary of the hollow area; Impedance analysis was performed on the array ultrasonic data to obtain the debonding depth and internal void distribution; Deformation analysis is performed on the laser three-dimensional point cloud data to obtain the deformation amplitude and thickness deviation of the plate surface; Obtain the design layout information of the target insulation area, which includes the position of the board joint, the position of the anchor bolt, and the position of the inside and outside corners; Based on the design layout information, the relative spatial distances between the abnormal sub-region and the board seam position, the anchor bolt position, and the inside and outside corner position are obtained; The crack width, crack length, crack distribution direction, infrared temperature difference amplitude, hollow area boundary, debonding depth, internal void distribution, plate surface deformation amplitude, thickness deviation value, and relative spatial distance are combined to form the defect feature vector.

[0007] Optionally, the step of inputting the defect feature vector into the inverse solving neural network to obtain the construction behavior deviation vector of each construction procedure corresponding to the abnormal sub-region includes: Obtain the defect decoupling features output by the encoder of the inverse solving neural network after performing dimensionality reduction analysis on the defect feature vector; The initial construction behavior deviation vectors of each construction procedure are obtained by the decoder of the reverse solving neural network after it reverse solves the defect decoupling features. The reverse solving process of the decoder is constrained by physical constraint equations. Determine whether the deviation value of each process in the initial construction behavior deviation vector is less than a preset deviation threshold; If the deviation value of each process is less than the preset deviation threshold, then the initial construction behavior deviation vector is output as the construction behavior deviation vector. If at least one process has a deviation value not less than the preset deviation threshold, then the construction and maintenance environment record of the abnormal sub-region is obtained. The initial construction behavior deviation vector is corrected based on the construction and maintenance environment records, and the corrected construction behavior deviation vector is output as the construction behavior deviation vector.

[0008] Optionally, the correction of the initial construction behavior deviation vector based on the construction and maintenance environment record includes: Obtain the ambient temperature and humidity sequences of the abnormal sub-region during construction and maintenance; Based on the ambient temperature sequence and the ambient humidity sequence, the maintenance environment deviation index is obtained; Determine whether the maintenance environment deviation index exceeds the environment deviation threshold; If the maintenance environment deviation index does not exceed the environment deviation threshold, then the initial construction behavior deviation vector is used as the corrected construction behavior deviation vector. If the maintenance environment deviation index exceeds the environment deviation threshold, then the maintenance process deviation value and the non-maintenance process deviation value are extracted from the initial construction behavior deviation vector. Based on the maintenance environment deviation index, the deviation values ​​of the maintenance process are attenuated and the deviation values ​​of the non-maintenance process are amplified. The deviation values ​​of the maintenance procedures after attenuation treatment and the deviation values ​​of the non-maintenance procedures after amplification treatment are combined to form the corrected construction behavior deviation vector.

[0009] Optionally, obtaining the construction behavior compliance score for each construction procedure based on the construction behavior deviation vector includes: Obtain the deviation values ​​of adhesive powder ratio and water addition ratio in the adhesive preparation process from the construction behavior deviation vector; Determine whether the deviation of the adhesive powder ratio is greater than a first target threshold or whether the deviation of the water addition ratio is greater than a second target threshold; If the deviation of the adhesive powder ratio is greater than the first target threshold or the deviation of the water addition ratio is greater than the second target threshold, then the adhesive preparation process will be marked as a target violation process. If the deviation of the adhesive powder ratio is not greater than the first target threshold and the deviation of the water addition ratio is not greater than the second target threshold, then the anchor bolt torque deviation value of the anchoring construction process is obtained. If the deviation value of the anchor bolt torque is greater than the third target threshold, the anchoring construction process will be marked as a target violation process; Obtain the deviation value of each process in the construction behavior deviation vector, and calculate the initial compliance score of each process based on the deviation value of each process and the preset deviation weight; If the target non-compliant process exists, the initial compliance score of the target non-compliant process will be penalized to obtain the construction behavior compliance score of each process; If no such violation process exists, the initial compliance score will be used as the compliance score for the construction behavior of each process.

[0010] Optionally, the initial compliance score of the target non-compliant process is penalized to obtain the compliance score of the construction behavior for each process, including: Obtain the deviation value of the target non-compliant process, and determine the penalty coefficient based on the deviation value. The larger the deviation value, the larger the penalty coefficient. Obtain the defect type and defect area corresponding to the target violation process, and obtain the penalty intensity base based on the defect type and defect area; Based on the penalty coefficient and the penalty intensity base, calculate the score deduction value for the target violation process; The initial compliance score of the target non-compliant process is subtracted from the score deduction value to obtain the construction behavior compliance score of the target non-compliant process; Obtain the initial compliance score of non-target non-compliance procedures as the compliance score of construction behavior for non-target non-compliance procedures; The compliance scores of the construction behavior of the target non-compliant process and the compliance scores of the construction behavior of the non-target non-compliant process are combined to form the compliance score of the construction behavior of each process.

[0011] Optionally, the step of assessing the overall quality level of the target insulation area based on the construction behavior compliance score and the preset evaluation rules includes: Obtain the parameter deviation values ​​of each detection sub-region in the target insulation area, and obtain the initial quality level of the entity based on the parameter deviation values ​​of each detection sub-region and the preset evaluation rules; Obtain the compliance score of the construction behavior of each construction process corresponding to each detection sub-region, and obtain the behavior quality correction coefficient based on the compliance score of the construction behavior of each construction process; Determine whether there is a target non-compliant process in the target insulation area; If the target non-compliant process exists, the initial quality level of the entity will be lowered by at least one level based on the behavior quality correction coefficient to obtain the overall quality level of the target insulation area; If the target violation process does not exist, then determine whether the behavior quality correction coefficient is less than the correction threshold; If the behavior quality correction coefficient is less than the correction threshold, the initial quality level of the entity is downgraded based on the behavior quality correction coefficient to obtain the overall quality level of the target insulation area. If the behavior quality correction coefficient is not less than the correction threshold, then the initial quality level of the entity is taken as the overall quality level of the target insulation area.

[0012] Optional, also includes: Based on the construction behavior deviation vector, the target process with the largest deviation value and the corresponding construction period of the target process are obtained; Based on the construction period, retrieve the corresponding video clips from the construction site monitoring video. Behavior recognition is performed on the video clips to obtain actual construction operation parameters; Determine whether the deviation values ​​of the actual construction operation parameters and the corresponding procedures in the construction behavior deviation vector point to the same direction of violation; If they all point to the same direction of violation, the defect feature vector, the construction behavior deviation vector, and the video clip will be associated and stored as evidence chain data. If the error does not point to the same direction, the multimodal defect detection data of the abnormal sub-region is reacquired and the reverse solution is performed again.

[0013] Secondly, this application also discloses an intelligent evaluation system for the construction quality of thermal insulation layers, which adopts the following technical solution: A smart evaluation system for the construction quality of thermal insulation layers, comprising: The parameter acquisition module is used to acquire the measured thermal parameters and design thermal parameters corresponding to different detection sub-regions in the target insulation area; The deviation calculation module is used to obtain the parameter deviation value based on the measured thermal parameters and the design thermal parameters; The deviation judgment module is used to determine whether the parameter deviation value exceeds a preset deviation threshold; The first evaluation module is used to evaluate the overall quality level of the target insulation area based on the preset evaluation rules if the parameter deviation value does not exceed the preset deviation threshold. The feature extraction module is used to acquire multimodal defect detection data of the abnormal sub-region if the parameter deviation value exceeds the preset deviation threshold, and to extract features from the multimodal defect detection data to obtain a defect feature vector. The reverse solving module is used to input the defect feature vector into the reverse solving neural network to obtain the construction behavior deviation vector of each construction procedure corresponding to the abnormal sub-region; The scoring acquisition module is used to obtain the compliance score of each construction process based on the construction behavior deviation vector. The second evaluation module is used to evaluate the overall quality level of the target insulation area based on the compliance score of the construction behavior and the preset evaluation rules.

[0014] In summary, this application includes the following beneficial technical effects: By first obtaining the deviation values ​​between the measured thermal parameters and the design thermal parameters of each detection sub-region in the target insulation area, the quality level is directly evaluated when the deviation value does not exceed the threshold. When the deviation value exceeds the threshold, multimodal defect detection data of abnormal sub-regions are collected and defect feature vectors are extracted. The reverse solving neural network is used to deduce the construction behavior deviation vector of each construction process from the defect features, thereby obtaining the compliance score of the construction behavior of each process. Finally, the overall quality level of the target insulation area is evaluated by combining the two dimensions of physical quality detection results and construction behavior evaluation results. On the one hand, it realizes the reverse tracing of the specific construction process and the degree of deviation of the operation parameters that caused the defects from the physical defects after completion, which solves the problem that the traditional method can only detect defects but cannot locate the root cause of the process. On the other hand, it expands the quality evaluation from a single physical quality dimension to a comprehensive evaluation of physical quality and construction behavior, providing quantitative data support for the behavior assessment and quality traceability of construction teams. Attached Figure Description

[0015] Figure 1This is a flowchart of the main process of an intelligent evaluation method for the construction quality of thermal insulation layer according to an embodiment of this application; Figure 2 This is a flowchart illustrating the steps involved in assessing the overall quality level of the target insulation area. Figure 3 This is a flowchart of the steps to extract features from multimodal defect detection data and obtain defect feature vectors. Figure 4 This is a flowchart showing the steps to obtain the construction behavior deviation vectors for each construction process corresponding to the abnormal sub-region. Figure 5 This is a module diagram of an intelligent evaluation system for the construction quality of thermal insulation layers according to an embodiment of this application.

[0016] Explanation of reference numerals in the attached figures: 1. Parameter acquisition module; 2. Deviation calculation module; 3. Deviation judgment module; 4. First evaluation module; 5. Defect feature extraction module; 6. Reverse solving module; 7. Score acquisition module; 8. Second evaluation module. Detailed Implementation

[0017] Firstly, this application discloses an intelligent evaluation method for the construction quality of thermal insulation layers.

[0018] Reference Figure 1 A method for intelligent evaluation of the construction quality of thermal insulation layers, comprising steps S101 to S108: Step S101: Obtain the measured thermal parameters and design thermal parameters corresponding to different detection sub-regions in the target insulation area.

[0019] Specifically, the target insulation area refers to the entire coverage area of ​​the insulation layer on the building's exterior walls where the insulation construction quality needs to be evaluated. The testing sub-areas are several adjacent testing units divided from the target insulation area according to set dimensions, with each sub-area independently collecting thermal parameters. Measured thermal parameters refer to parameters such as surface temperature and heat flux density of each testing sub-area, actually measured using testing equipment such as infrared thermal imagers and heat flux meters after the target insulation area is completed. Design thermal parameters refer to the theoretical thermal performance indicators that each testing sub-area should achieve under standard operating conditions, determined according to the insulation layer design drawings and energy-saving calculations, including design surface temperature and design heat flux density. By comparing the measured values ​​with the design values, a preliminary judgment can be made as to whether the insulation effect of each testing sub-area meets the design requirements.

[0020] Step S102: Obtain the parameter deviation value based on the measured thermal parameters and the design thermal parameters.

[0021] Specifically, parameter deviation refers to the degree of deviation between measured thermal parameters and design thermal parameters. It is calculated by subtracting the measured surface temperature from the design surface temperature of each test sub-region, or by subtracting the measured heat flux density from the design heat flux density, to obtain the parameter deviation value for each test sub-region. A larger parameter deviation value indicates a greater gap between the insulation effect of that region and the design expectation, potentially indicating defects in the insulation layer.

[0022] Step S103: Determine whether the parameter deviation value exceeds the preset deviation threshold.

[0023] Specifically, the preset deviation threshold is an upper limit of allowable deviation that is pre-set based on building energy-saving design standards, thermal insulation material performance parameters, and engineering practice experience.

[0024] Step S104: If the parameter deviation value does not exceed the preset deviation threshold, the overall quality level of the target insulation area is evaluated based on the preset evaluation rules.

[0025] Specifically, the preset evaluation rule refers to a pre-defined evaluation standard used to map the parameter deviation values ​​of each detection sub-region to a quality level. This evaluation rule sets the range of parameter deviation values ​​corresponding to different quality levels. When the parameter deviation values ​​of all detection sub-regions do not exceed the preset deviation threshold, the parameter deviation values ​​of each detection sub-region are mapped to the corresponding quality level according to the preset evaluation rule. Then, the overall quality level of the target insulation area is determined by combining the quality levels of each detection sub-region. For example, the lowest quality level of each detection sub-region can be taken as the overall quality level, or the weighted average of the quality levels of each detection sub-region can be taken as the overall quality level.

[0026] Step S105: If the parameter deviation value exceeds the preset deviation threshold, the multimodal defect detection data of the abnormal sub-region is obtained, and the multimodal defect detection data is used to extract features to obtain the defect feature vector.

[0027] Specifically, an abnormal sub-region refers to a detection sub-region where the parameter deviation value exceeds a preset deviation threshold. This region may contain quality defects such as hollowness, cracking, and debonding in the insulation layer. Multimodal defect detection data refers to the detection data of the same abnormal sub-region collected by various types of detection equipment, including visible light images, infrared thermal imaging spectra, array ultrasonic data, and laser 3D point cloud data. Visible light images are acquired by industrial cameras and used to identify cracks and damage on the surface of the insulation layer. Infrared thermal imaging spectra are acquired by infrared thermal imagers and used to detect hollow areas inside the insulation layer. Because the thermal conductivity of hollow areas differs from that of dense areas, they will appear as temperature anomalies in the infrared image. Array ultrasonic data is acquired by an ultrasonic phased array detector, which uses the reflection and attenuation characteristics of ultrasonic waves inside the insulation layer to detect the debonding depth and the distribution of internal voids. Laser 3D point cloud data is acquired by a 3D laser scanner and used to obtain the 3D morphological information of the insulation layer surface, including the deformation amplitude and thickness deviation value. Feature extraction of multimodal defect detection data refers to extracting quantitative indicators that can characterize defect attributes from various types of detection data and combining these quantitative indicators to form a defect feature vector.

[0028] Step S106: Input the defect feature vector into the reverse solving neural network to obtain the construction behavior deviation vector of each construction process corresponding to the abnormal sub-region.

[0029] Specifically, the inverse solving neural network is a pre-trained deep learning network model that takes a defect feature vector as input and outputs a construction behavior deviation vector for each construction process. The training process of the inverse solving neural network is as follows: First, in a laboratory setting, the changes in a single construction behavior parameter are controlled, and the corresponding defect features are recorded, forming a sample library containing the deviation values ​​of the construction behavior parameters and the corresponding defect feature labels. Then, this sample library is used to train the neural network, enabling the network to learn the inverse mapping relationship from defect features to construction behavior deviations.

[0030] The construction behavior deviation vector is a multi-dimensional vector, where each dimension corresponds to the degree to which the operational parameters of a construction process deviate from the standard value. Construction processes include adhesive preparation, adhesive application, insulation board laying and extrusion, anchor installation, mesh fabric laying, finishing mortar application, and finished product curing. The value of each dimension represents the direction and magnitude of deviation of the actual operational parameters of that process relative to the standard construction parameters. For example, if the standard anchor torque is 4 N·m, and the inverted torque deviation value is 2.1 N·m, it indicates that the actual torque is lower than the standard value, and the anchoring force is insufficient.

[0031] Step S107: Based on the construction behavior deviation vector, obtain the construction behavior compliance score for each construction process.

[0032] Specifically, the construction behavior compliance score is a numerical value used to quantitatively evaluate the degree to which the work behavior of each construction procedure conforms to standards and specifications. It is obtained by: for each construction procedure, obtaining the deviation value corresponding to that procedure in the construction behavior deviation vector, and calculating the compliance score for that procedure based on the magnitude of the deviation value.

[0033] Step S108: Based on the compliance score of construction behavior and the preset evaluation rules, conduct an overall quality level assessment of the target insulation area.

[0034] Specifically, the initial quality level of the entity is first obtained based on the parameter deviation values ​​of each detection sub-region and the preset evaluation rules. Then, the behavior quality correction coefficient is calculated based on the compliance score of each construction process. Finally, the overall quality level of the target insulation area is determined by combining the initial quality level of the entity and the behavior quality correction coefficient.

[0035] Reference Figure 2 In one embodiment of this example, if the parameter deviation value does not exceed a preset deviation threshold, then based on preset evaluation rules, the overall quality level of the target insulation area is assessed, including steps S201 to S207: Step S201: If the parameter deviation value does not exceed the preset deviation threshold, then obtain the thermal parameter difference between adjacent detection sub-regions in the target insulation area.

[0036] Specifically, when the parameter deviation values ​​of all detection sub-regions do not exceed the preset deviation threshold, although each sub-region is within the allowable deviation range when viewed individually, there may be significant differences in thermal parameters between adjacent sub-regions. Such local differences may indicate the presence of local quality problems. The thermal parameter difference refers to the difference between the measured thermal parameters of two adjacent detection sub-regions, such as the difference in surface temperature or heat flux density between adjacent sub-regions.

[0037] Step S202: Determine whether the difference in thermal parameters exceeds the local difference threshold.

[0038] Specifically, the local difference threshold is a pre-set upper limit value used to determine whether the difference in thermal parameters between adjacent detection sub-regions is abnormal.

[0039] Step S203: If the difference in thermal parameters does not exceed the local difference threshold, the overall quality level of the target insulation area is evaluated based on the preset evaluation rules.

[0040] Specifically, when the difference in thermal parameters between adjacent test sub-regions does not exceed the local difference threshold, it indicates that the thermal performance of each sub-region within the target insulation area is uniform and there are no local anomalies. The overall quality level of the target insulation area is directly evaluated according to the preset evaluation rules.

[0041] Step S204: If the difference in thermal parameters exceeds the local difference threshold, then obtain the boundary position of the adjacent detection sub-region.

[0042] Specifically, when the difference in thermal parameters between adjacent detection sub-regions exceeds the local difference threshold, it is necessary to determine the specific spatial coordinates of the boundary between these two adjacent sub-regions within the target insulation area in order to determine whether the boundary position coincides with the location of the structural node of the insulation layer. The boundary position refers to the location of the intersection line between two adjacent detection sub-regions on the surface of the insulation layer.

[0043] Step S205: Determine whether the dividing position coincides with the board seam position or the inside and outside corner position.

[0044] Specifically, the joint location refers to the planar coordinate position of the splice seam between all insulation boards marked on the design drawings on the surface of the insulation layer, while the corner location refers to the planar coordinate position of the corner line at the exterior wall corner marked on the design drawings on the surface of the insulation layer. Because there are material discontinuities and geometric abrupt changes at the joints between insulation boards and at wall corners, these locations will inevitably produce a certain degree of difference in thermal parameters, i.e., structural temperature difference.

[0045] Step S206: If the boundary position coincides with the plate seam position or the internal and external corner position, the thermal parameter difference is determined to be a structural temperature difference, and the overall quality level of the target insulation area is evaluated based on the preset evaluation rules.

[0046] Specifically, if the boundary coincides with the joint of the insulation board or the corner, it indicates that the difference in thermal parameters between adjacent test sub-areas is due to structural factors at the joints of the insulation boards or the corners of the walls. This is a normal phenomenon and does not represent a quality defect. Therefore, this difference in thermal parameters is judged as a structural temperature difference, and no points are deducted. The overall quality level of the target insulation area is directly assessed according to the preset evaluation rules.

[0047] Step S207: If the boundary position does not coincide with the board seam position and the inside and outside corner position, then the quality level of the adjacent test sub-areas shall be evaluated separately, and the lowest evaluated quality level shall be used as the overall quality level of the target insulation area.

[0048] Specifically, if the boundary does not coincide with either the board seam or the corner, it indicates that the difference in thermal parameters between adjacent test sub-regions is not due to structural factors, but rather signifies a genuine difference in insulation quality between the two sub-regions. In this case, the two adjacent test sub-regions are independently assessed for quality according to preset evaluation rules. The lower of the two quality levels is then taken as the overall quality level of the target insulation area, reflecting the principle of strict control in quality assessment.

[0049] Reference Figure 3In one embodiment of this invention, feature extraction of multimodal defect detection data to obtain defect feature vectors includes steps S301 to S307: Step S301: Perform crack identification on the visible light image to obtain crack width, crack length, and crack distribution direction.

[0050] Specifically, visible light images are high-resolution color or grayscale images captured by industrial cameras on the surface of anomaly sub-regions. Crack identification refers to extracting crack regions from visible light images using image processing algorithms (such as edge detection, thresholding, and morphological processing), and then measuring the width, length, and distribution direction of the cracks on the insulation layer surface. Crack width refers to the vertical distance between the two edges of the crack, crack length refers to the total length of the crack along its extension direction, and crack distribution direction refers to the directional angle of the cracks extending on the insulation layer surface.

[0051] Step S302: Perform temperature field analysis on the infrared thermal imaging spectrum to obtain the infrared temperature difference amplitude and the boundary of the hollow area.

[0052] Specifically, an infrared thermal imaging map is an image of the temperature distribution on the surface of anomaly sub-regions acquired by an infrared thermal imager. Temperature field analysis refers to the statistical analysis of the temperature values ​​of each pixel in the infrared thermal imaging map to identify temperature anomaly regions. Infrared temperature difference amplitude refers to the difference between the highest and lowest temperatures within the anomaly sub-region; a larger temperature difference amplitude indicates a more severe degree of uneven heat conduction within the insulation layer. The boundary of the void region refers to the outline of the temperature anomaly region in the infrared thermal imaging map. Because the thermal conductivity of the air inside the void region is much lower than that of the insulation material, the void region appears as a closed area with a significant temperature difference from the surrounding area in the infrared image.

[0053] Step S303: Perform impedance analysis on the array ultrasound data to obtain the debonding depth and internal void distribution.

[0054] Specifically, array ultrasonic data consists of multi-channel ultrasonic echo signals acquired by an ultrasonic phased array detector in anomaly sub-regions. Impedance analysis refers to calculating the acoustic impedance values ​​at different depths within the insulation layer based on the amplitude and phase changes of the ultrasonic echo signals. Debonding depth refers to the depth of the bonding failure interface between the insulation layer and the base wall, i.e., the vertical distance from the debonding area to the surface of the insulation layer. Internal void distribution refers to the coordinates of the locations of voids within the insulation layer and the distribution range of these voids on the plane of the insulation layer.

[0055] Step S304: Perform deformation analysis on the laser 3D point cloud data to obtain the deformation amplitude and thickness deviation values ​​of the plate surface.

[0056] Specifically, laser 3D point cloud data consists of the 3D spatial coordinates of numerous discrete points on the surface of anomaly sub-regions, collected by a 3D laser scanner. Deformation analysis involves comparing the actual 3D coordinates of the scanned board surface with the theoretical coordinates of the designed board surface to calculate the deformation at each point. The board surface deformation amplitude refers to the maximum offset of the insulation board surface relative to the design plane, reflecting the flatness of the insulation board. The thickness deviation value refers to the difference between the actual thickness and the designed thickness of the insulation layer.

[0057] Step S305: Obtain the design layout information of the target insulation area. The design layout information includes the position of the board joint, the position of the anchor bolt, and the position of the inside and outside corners.

[0058] Specifically, the design layout information is structured data extracted from the insulation layer construction design drawings and construction plans. Anchor bolt locations refer to the planar coordinates of all anchor installation points marked on the design drawings.

[0059] Step S306: Based on the design layout information, obtain the relative spatial distance between the abnormal sub-region and the plate seam position, anchor bolt position, and inside and outside corner position.

[0060] Specifically, the planar coordinates of the abnormal sub-region on the insulation layer surface are spatially superimposed with the planar coordinates of the board joint positions, anchor bolt positions, and internal / external corner positions in the design layout information. The Euclidean distances from the center point of the abnormal sub-region to the nearest board joint, the nearest anchor bolt, and the nearest internal / external corner are calculated. Relative spatial distances are an important basis for judging the cause of defects. For example, if the hollow area is adjacent to the anchor bolt position, the defect is likely caused by improper anchoring construction.

[0061] Step S307: Combine the crack width, crack length, crack distribution direction, infrared temperature difference amplitude, hollow area boundary, debonding depth, internal void distribution, plate surface deformation amplitude, thickness deviation value, and relative spatial distance into a defect feature vector.

[0062] Specifically, crack width, crack length, and crack distribution direction are used as the first type of surface defect characteristic quantity; infrared temperature difference amplitude and hollow area boundary are used as the second type of thermal defect characteristic quantity; debonding depth and internal void distribution are used as the third type of internal defect characteristic quantity; plate deformation amplitude and thickness deviation are used as the fourth type of deformation defect characteristic quantity; and relative spatial distance is used as the fifth type of spatial position characteristic quantity. The first type of surface defect characteristic quantity, the second type of thermal defect characteristic quantity, the third type of internal defect characteristic quantity, the fourth type of deformation defect characteristic quantity, and the fifth type of spatial position characteristic quantity are normalized respectively, so that the numerical range of each type of characteristic quantity is uniformly mapped to the interval [0, 1]. According to the arrangement order of the first type of surface defect characteristic quantity, the second type of thermal defect characteristic quantity, the third type of internal defect characteristic quantity, the fourth type of deformation defect characteristic quantity, and the fifth type of spatial position characteristic quantity, all the normalized characteristic quantities are sequentially concatenated into a one-dimensional vector to obtain the defect characteristic vector.

[0063] Reference Figure 4 In one embodiment of this invention, the defect feature vector is input into the inverse solving neural network to obtain the construction behavior deviation vector of each construction process corresponding to the abnormal sub-region, including steps S401 to S406: Step S401: Obtain the defect decoupling features output by the encoder of the inverse solving neural network after performing dimensionality reduction analysis on the defect feature vector.

[0064] Specifically, the inverse solving neural network employs an encoder-decoder architecture. The encoder consists of multiple stacked convolutional and pooling layers, and its function is to reduce the dimensionality of the input defect feature vector, mapping the high-dimensional original defect features to a low-dimensional latent space. During dimensionality reduction, the encoder separates potentially coupled defect features through feature extraction and transformation, outputting defect decoupling features. Defect decoupling features are low-dimensional feature vectors representing a single defect type or a combination of composite defects in the latent space. For example, the ultrasonic features of a hollow defect are separated from the image features of a crack defect in the latent space, facilitating targeted inverse solving by the decoder for different types of defects.

[0065] Step S402: Obtain the initial construction behavior deviation vector of each construction procedure output by the decoder of the reverse solving neural network after performing reverse solving of the defect decoupling features.

[0066] Specifically, the decoder consists of multiple stacked deconvolutional and fully connected layers. Its function is to inversely map the defect decoupling features output by the encoder into construction behavior deviation vectors for each construction process. The inverse solution process of the decoder is constrained by physical constraint equations, which are pre-constructed explicit physical equations based on adhesive mechanics, adhesive hydration kinetics, and solid heat transfer. The adhesive mechanics equation describes the relationship between stress transfer and bond strength in the adhesive layer and the coating thickness and area ratio. The adhesive hydration kinetics equation describes the relationship between the hydration reaction rate of cementitious materials in the adhesive and ambient temperature, humidity, and curing time. The solid heat transfer equation describes the heat conduction law in the insulation layer and the relationship between the temperature field distribution and the insulation layer thickness and internal porosity. These physical constraint equations are embedded in the loss function of the decoder. When the decoder output violates the physical laws, the loss function applies a penalty, forcing the decoder output to conform to the physical evolution law.

[0067] Step S403: Determine whether the deviation values ​​of each process in the initial construction behavior deviation vector are all less than the preset deviation threshold.

[0068] Specifically, the preset deviation threshold is a pre-set upper limit value used to judge whether the reverse solution result is reasonable and reliable. After obtaining the initial construction behavior deviation vector, it is checked one by one whether the deviation value corresponding to each process in the vector is less than the preset deviation threshold.

[0069] Step S404: If the deviation value of each process is less than the preset deviation threshold, the initial construction behavior deviation vector is output as the construction behavior deviation vector.

[0070] Specifically, when the deviation values ​​of all processes are less than the preset deviation threshold, the reverse solution result is deemed reasonable and reliable, and no correction is required. The initial construction behavior deviation vector is directly output as the final construction behavior deviation vector for subsequent compliance score calculation.

[0071] Step S405: If at least one process has a deviation value that is not less than a preset deviation threshold, then obtain the construction and maintenance environment record of the abnormal sub-region.

[0072] Specifically, when a deviation in a process step is not less than a preset deviation threshold, the reverse engineering result is deemed to be affected by environmental factors. Construction and maintenance environmental records refer to environmental monitoring data continuously collected during construction and maintenance of the target insulation area, including daily meteorological data such as ambient temperature, humidity, wind speed, and rainfall.

[0073] Step S406: Correct the initial construction behavior deviation vector based on the construction and maintenance environment records, and output the corrected construction behavior deviation vector as the construction behavior deviation vector.

[0074] Specifically, based on the construction and maintenance environment records, the degree to which the ambient temperature and humidity deviate from the standard maintenance conditions during the maintenance period is analyzed. If there is a significant deviation in environmental factors, the initial construction behavior deviation vector is corrected to eliminate the interference of environmental factors on the reverse solution results. The corrected construction behavior deviation vector is then output as the final construction behavior deviation vector.

[0075] In one embodiment of this example, correcting the initial construction behavior deviation vector based on construction and maintenance environment records includes steps S501 to S507: Step S501: Obtain the ambient temperature and humidity sequences of the abnormal sub-region during construction and maintenance.

[0076] Specifically, all environmental temperature and humidity measurements of the abnormal sub-areas during the construction and maintenance of the insulation layer are extracted from the construction and maintenance environmental records and arranged in chronological order of collection time to form environmental temperature and environmental humidity sequences.

[0077] Step S502: Obtain the maintenance environment deviation index based on the ambient temperature sequence and ambient humidity sequence.

[0078] Specifically, the maintenance environment deviation index is a quantitative indicator used to measure the degree to which actual maintenance environmental conditions deviate from standard maintenance conditions. It is obtained by first determining the standard maintenance temperature range and the standard maintenance humidity range, then calculating the cumulative temperature deviation value exceeding the standard temperature range in the environmental temperature series and the cumulative humidity deviation value exceeding the standard humidity range in the environmental humidity series. The cumulative temperature deviation value and the cumulative humidity deviation value are then weighted and synthesized to obtain the maintenance environment deviation index.

[0079] Step S503: Determine whether the maintenance environment deviation index exceeds the environment deviation threshold.

[0080] Specifically, the environmental deviation threshold is a pre-set upper limit value used to determine whether the maintenance environment conditions will have a significant impact on the quality of the insulation layer.

[0081] Step S504: If the maintenance environment deviation index does not exceed the environment deviation threshold, the initial construction behavior deviation vector is used as the corrected construction behavior deviation vector.

[0082] Specifically, when the maintenance environment deviation index does not exceed the environmental deviation threshold, it indicates that environmental factors are not the main cause of quality defects. The reverse solution does not need to be corrected for environmental factors, and the initial construction behavior deviation vector is directly used as the corrected construction behavior deviation vector.

[0083] Step S505: If the maintenance environment deviation index exceeds the environment deviation threshold, extract the maintenance process deviation value and the non-maintenance process deviation value from the initial construction behavior deviation vector.

[0084] Specifically, when the environmental deviation index exceeds the environmental deviation threshold, environmental factors need to be separated. The deviation values ​​corresponding to the maintenance process and those corresponding to other non-maintenance processes are separated from the initial construction behavior deviation vector. The maintenance process refers to the finished product maintenance process, and its deviation value reflects the degree to which maintenance operations (such as watering frequency and maintenance start time) deviate from the standard. Non-maintenance processes include adhesive preparation, adhesive layer application, insulation board laying and extrusion, anchor installation, mesh fabric laying, and finishing mortar application.

[0085] Step S506: Based on the maintenance environment deviation index, the deviation values ​​of maintenance procedures are attenuated and the deviation values ​​of non-maintenance procedures are amplified.

[0086] Specifically, when the curing environment deviates, the final defects in the insulation layer may be the result of the combined effects of environmental factors and construction practices. For deviations in curing procedures, excessively high or low ambient temperatures can lead to abnormal adhesive hydration reactions, potentially causing defects even if the curing work itself meets standards. Therefore, it is necessary to attenuate the deviation values ​​of curing procedures based on the curing environment deviation index, i.e., reduce the contribution of environmental factors to the inversion results of curing procedures. For deviations in non-curing procedures, since curing environment deviations can mask or offset some of the effects of construction practice deviations, it is necessary to amplify the deviation values ​​of non-curing procedures to compensate for the suppression of inversion results by environmental factors.

[0087] Step S507: Combine the attenuated deviation value of the maintenance process and the amplified deviation value of the non-maintenance process into a corrected construction behavior deviation vector.

[0088] Specifically, the revised construction behavior deviation vector more accurately reflects the actual degree of deviation in construction behavior.

[0089] In one embodiment of this example, obtaining the compliance score of each construction process based on the construction behavior deviation vector includes steps S601 to S608: Step S601: Obtain the deviation values ​​of adhesive powder ratio and water addition ratio for the adhesive preparation process in the construction behavior deviation vector.

[0090] Specifically, the adhesive preparation process is the first step in the installation of insulation boards, involving mixing adhesive powder and water in a certain proportion. The adhesive powder ratio deviation refers to the deviation of the actual mass ratio of adhesive powder to cement or other base materials from the standard ratio. The water addition ratio deviation refers to the deviation of the actual amount of water added from the standard amount. These two deviations are extracted from the construction behavior deviation vector and are used to determine whether there are serious violations in the adhesive preparation.

[0091] Step S602: Determine whether the deviation of the adhesive powder ratio is greater than the first target threshold or whether the deviation of the water addition ratio is greater than the second target threshold.

[0092] Specifically, the first target threshold is the upper limit of the warning value for deviation of the adhesive powder ratio, and the second target threshold is the upper limit of the warning value for deviation of the water addition ratio.

[0093] Step S603: If the deviation of the adhesive powder ratio is greater than the first target threshold or the deviation of the water addition ratio is greater than the second target threshold, then the adhesive preparation process is marked as a target violation process.

[0094] Specifically, the target violation process refers to a construction process with fatal violations that could lead to serious consequences such as large-scale detachment of the insulation layer or failure of the fireproof barrier. When the deviation of the adhesive powder ratio exceeds the first target threshold or the deviation of the water addition ratio exceeds the second target threshold, the adhesive preparation process is marked as a target violation process.

[0095] Step S604: If the deviation of the adhesive powder ratio is not greater than the first target threshold and the deviation of the water addition ratio is not greater than the second target threshold, then obtain the anchor bolt torque deviation value of the anchoring construction process.

[0096] Specifically, if there are no fatal violations in the adhesive preparation process, then proceed to inspect the anchoring installation process. The anchoring installation process refers to the construction procedure of driving anchors into the base wall to enhance the fixing strength of the insulation board. Anchor bolt torque deviation refers to the deviation of the anchor bolt tightening torque applied by an electric torque wrench during actual construction from the standard torque.

[0097] Step S605: If the anchor bolt torque deviation value is greater than the third target threshold, then mark the anchoring construction process as a target violation process.

[0098] Specifically, the third target threshold is a pre-set upper limit for the anchor bolt torque deviation. Excessive anchor bolt torque deviation will cause the anchor to exert excessive pressure on the insulation board, resulting in surface depression or cracking around the anchor point; insufficient anchor bolt torque deviation (i.e., excessive negative deviation) will lead to insufficient anchoring force, failing to effectively secure the insulation board. When the anchor bolt torque deviation exceeds the third target threshold, the anchoring construction process will be marked as a target violation.

[0099] Step S606: Obtain the deviation value of each process in the construction behavior deviation vector, and calculate the initial compliance score of each process based on the deviation value of each process and the preset deviation weight.

[0100] Specifically, for each construction procedure, its corresponding deviation value in the construction behavior deviation vector is obtained. The preset deviation weights are weight coefficients pre-assigned to each construction procedure. Different procedures have different degrees of influence on the final quality of the insulation layer, therefore their weight coefficients differ. For example, the adhesive preparation and anchoring procedures have higher weight coefficients, while the mesh fabric laying procedure has a relatively lower weight coefficient. The initial compliance score is obtained by converting the deviation values ​​of each procedure into a percentage score according to a preset scoring mapping function, multiplying it by the corresponding deviation weight of the procedure, and then summing the results.

[0101] Step S607: If there is a target non-compliant process, the initial compliance score of the target non-compliant process is penalized to obtain the construction behavior compliance score of each process.

[0102] Specifically, if a target non-compliant process exists, an additional penalty will be applied to the initial compliance score of that process. This penalty will be deducted from the initial compliance score to obtain the penalized compliance score. For non-target non-compliant processes, the initial compliance score will be used directly.

[0103] Step S608: If there is no target non-compliant process, the initial compliance score will be used as the compliance score for the construction behavior of each process.

[0104] Specifically, if there is no target violation process, the initial compliance score of all processes does not need to be penalized and will be directly used as the compliance score of the construction behavior of each process.

[0105] In one embodiment of this example, the initial compliance score of the target non-compliant process is penalized to obtain the compliance score of the construction behavior of each process, including steps S701 to S706: Step S701: Obtain the deviation value of the target non-compliant process, and determine the penalty coefficient based on the deviation value. The larger the deviation value, the larger the penalty coefficient.

[0106] Specifically, for a construction process marked as a target violation, the deviation value of that process in the construction behavior deviation vector is first obtained. Then, based on the magnitude of the deviation value, a penalty coefficient is determined using a preset penalty coefficient mapping function. The principle for setting the penalty coefficient mapping function is that the larger the deviation value, the larger the penalty coefficient, that is, the more serious the violation, the heavier the penalty.

[0107] Step S702: Obtain the defect type and defect area corresponding to the target violation process, and obtain the penalty intensity base based on the defect type and defect area.

[0108] Specifically, defect type refers to the type of physical defect caused by the non-compliant process, such as hollow areas, cracks, and delamination. Defect area refers to the distribution area of ​​the defect on the surface of the insulation layer. Different defect types have different degrees of severity; for example, large-area hollow areas are far more harmful than minor cracks, and therefore have different corresponding penalty intensity bases. The corresponding penalty intensity base is found in a preset penalty intensity table by looking up the defect type and defect area. The larger the defect area and the more severe the defect type, the higher the penalty intensity base.

[0109] Step S703: Calculate the score deduction value for the target non-compliant process based on the penalty coefficient and the penalty intensity base.

[0110] Specifically, the initial compliance score of the target violation process is obtained and used as the base score; the penalty coefficient is multiplied by the penalty intensity base to obtain the comprehensive penalty coefficient; the preset maximum deduction value for a single violation is obtained, and the comprehensive penalty coefficient is multiplied by the maximum deduction value for a single violation to obtain the score deduction value for the target violation process; where, when the comprehensive penalty coefficient is greater than 1, the maximum deduction value for a single violation is taken as the score deduction value.

[0111] Step S704: Subtract the deduction value from the initial compliance score of the target non-compliant process to obtain the compliance score of the construction behavior of the target non-compliant process.

[0112] Specifically, the final compliance score of the construction behavior of the target non-compliant process is obtained by deducting the score deduction value calculated in step S703 from the initial compliance score of the target non-compliant process.

[0113] Step S705: Obtain the initial compliance score of the non-target non-compliance process as the compliance score of the construction behavior of the non-target non-compliance process.

[0114] Specifically, for all other construction processes that are not marked as the target violation process, no penalties will be imposed, and their initial compliance scores will be used directly as their respective construction behavior compliance scores.

[0115] Step S706: Combine the compliance scores of the construction behavior of the target non-compliance process with the compliance scores of the construction behavior of the non-target non-compliance process to form the compliance score of the construction behavior of each process.

[0116] Specifically, obtain a list of process categories corresponding to the target insulation area. The process category list includes adhesive preparation, adhesive coating, insulation board laying and extrusion, anchor installation, mesh fabric laying, finishing mortar application, and finished product curing. Following the order of processes in the process category list, determine whether each process belongs to the target violation process or a non-target violation process. If the process belongs to the target violation process, fill in the corresponding compliance score of the construction behavior in the process category list. If the process belongs to a non-target violation process, fill in the corresponding compliance score of the construction behavior in the process category list. After traversing all processes, obtain the compliance score of the construction behavior for each process.

[0117] In one embodiment of this example, the overall quality level assessment of the target insulation area based on the construction behavior compliance score and preset evaluation rules includes steps S801 to S807: Step S801: Obtain the parameter deviation values ​​of each detection sub-region in the target insulation area, and obtain the initial quality level of the entity based on the parameter deviation values ​​of each detection sub-region and the preset evaluation rules.

[0118] Specifically, the parameter deviation values ​​of each detection sub-region within the target insulation area are obtained; according to the parameter deviation value range corresponding to each quality level set in the preset evaluation rules, the parameter deviation value of each detection sub-region is mapped to the corresponding single sub-region quality level; the number of detection sub-regions corresponding to each quality level is counted; the quality level with the highest proportion is taken as the first candidate level; it is determined whether the proportion of the number of detection sub-regions corresponding to the first candidate level to the total number of detection sub-regions exceeds a set proportion threshold; if it exceeds the set proportion threshold, the first candidate level is taken as the initial quality level of the entity; if it does not exceed the set proportion threshold, the lowest level among the single sub-region quality levels of all detection sub-regions is taken as the initial quality level of the entity.

[0119] Step S802: Obtain the compliance score of the construction behavior of each construction process corresponding to each detection sub-region, and obtain the behavior quality correction coefficient based on the compliance score of the construction behavior of each construction process.

[0120] Specifically, the compliance scores of construction behaviors for each construction procedure corresponding to all sub-regions in the target insulation area are obtained; for each sub-region, the average of the compliance scores of all construction procedures corresponding to that sub-region is taken to obtain the sub-region behavior score of that sub-region; the average of the sub-region behavior scores of all sub-regions is taken to obtain the average behavior score of the target insulation area; the average behavior score is divided by the preset standard full score to obtain the behavior quality correction coefficient.

[0121] Step S803: Determine whether there is any non-compliant process in the target insulation area.

[0122] Step S804: If there is a target non-compliant process, the initial quality level of the entity is lowered by at least one level based on the behavior quality correction coefficient to obtain the overall quality level of the target insulation area.

[0123] Specifically, if a target non-compliance process exists, it indicates a fatal construction violation that could lead to a serious quality and safety accident. Therefore, regardless of the physical quality inspection data, the initial physical quality level must be forcibly downgraded by at least one quality level. For example, if the initial physical quality level is B, it should be downgraded by one level to C, or downgraded by two levels to D.

[0124] Step S805: If there is no target violation process, determine whether the behavior quality correction coefficient is less than the correction threshold.

[0125] Specifically, the correction threshold is a pre-set critical value used to determine whether the initial quality level of an entity needs to be downgraded.

[0126] Step S806: If the behavior quality correction coefficient is less than the correction threshold, the initial quality level of the entity is downgraded based on the behavior quality correction coefficient to obtain the overall quality level of the target insulation area.

[0127] Specifically, when the behavior quality correction coefficient is less than the correction threshold, it indicates that although there are no fatal violations, the overall standardization of construction behavior is poor, with numerous violations, which may lead to the deterioration of the insulation layer's performance over long-term use. Therefore, it is necessary to downgrade the initial quality level of the entity based on the behavior quality correction coefficient, with the downgrade magnitude proportional to the degree to which the behavior quality correction coefficient deviates from the correction threshold. Specifically: calculate the difference between the correction threshold and the behavior quality correction coefficient to obtain the deviation difference; divide the deviation difference by the correction threshold to obtain the deviation ratio; obtain the correspondence between the preset deviation ratio range and the number of downgrade levels, with a larger deviation ratio corresponding to more downgrade levels; specifically, a deviation ratio less than 0.1 results in a downgrade of level 0, a deviation ratio greater than or equal to 0.1 and less than 0.25 results in a downgrade of level 1, a deviation ratio greater than or equal to 0.25 and less than 0.4 results in a downgrade of level 2, and a deviation ratio greater than or equal to 0.4 results in a downgrade of level 3; determine the number of downgrade levels based on the range of the deviation ratio; and downgrade the initial quality level of the entity by the specified number of downgrade levels to obtain the overall quality level of the target insulation area.

[0128] Step S807: If the behavior quality correction coefficient is not less than the correction threshold, then the initial quality level of the entity is taken as the overall quality level of the target insulation area.

[0129] Specifically, when the behavior quality correction coefficient is not less than the correction threshold, it indicates that the overall standardization of the construction behavior is good, and there is no need to lower the initial quality level of the entity. The initial quality level of the entity can be directly used as the overall quality level of the target insulation area.

[0130] In one embodiment of this example, steps S901 to S906 are further included: Step S901: Based on the construction behavior deviation vector, obtain the target process with the largest deviation value and the corresponding construction period of the target process.

[0131] Specifically, the process with the largest deviation value is identified from the construction behavior deviation vector as the target process, which is the most significant construction error leading to the defect. Then, based on the construction log or construction progress record, the specific construction period of this target process in the target insulation area is obtained, including the construction date and start and end times.

[0132] Step S902: Based on the construction period, retrieve the video clips corresponding to the construction site monitoring video for that period.

[0133] Specifically, the construction site monitoring video is a continuous recording made by cameras fixed at the construction site throughout the entire insulation layer construction process. Based on the target process construction time period obtained in step S901, video clips of the corresponding construction area within that time period are retrieved from the monitoring video storage system.

[0134] Step S903: Perform behavior recognition on the video clips to obtain actual construction operation parameters.

[0135] Specifically, behavior recognition refers to using computer vision algorithms to analyze the operational actions of construction workers in video clips, and to identify and measure the actual construction operation parameters. For example, by analyzing video clips of anchoring construction, the number of rotations and operation time of the electric torque wrench can be identified to estimate the actual applied anchor bolt torque value; by analyzing video clips of adhesive application, the application method and application area can be identified to estimate the actual adhesive application area ratio.

[0136] Step S904: Determine whether the deviation values ​​of the corresponding procedures in the deviation vector between the actual construction operation parameters and the construction behavior point to the same violation direction.

[0137] Specifically, the direction of violation refers to whether the deviation value is positive or negative, and the specific type of violation it corresponds to. For example, the direction of violation corresponding to the anchor bolt torque deviation value is either too much torque or too little torque. If the construction behavior deviation vector shows that the anchor bolt torque deviation value is positive (the actual torque is greater than the standard value), and the video behavior recognition also identifies the operation feature of excessive torque, then both point to the same direction of violation.

[0138] Step S905: If they point to the same direction of violation, then the defect feature vector, construction behavior deviation vector and video clip are associated and stored as evidence chain data.

[0139] Specifically, when the video recognition result and the reverse engineering result point to the same direction of violation, it indicates that the reverse engineering result is corroborated by on-site construction video evidence. These three elements constitute a complete and traceable chain of evidence. The defect feature vector of the abnormal sub-region, the construction behavior deviation vector, and the corresponding video clip are linked using timestamps and spatial coordinates and stored in a database. This chain of evidence can be used as a basis for subsequent tracing of responsibility for quality accidents and settlement of quality assurance deposits.

[0140] Step S906: If the same violation direction is not pointed to, the multimodal defect detection data of the abnormal sub-region is reacquired and the reverse solution is performed again.

[0141] Specifically, if the video recognition result and the reverse engineering result point to different directions of violation, it indicates that the reverse engineering result may not match the actual situation. This could be due to errors in the defect detection data acquisition or insufficient inversion accuracy of the reverse engineering neural network under specific conditions. Therefore, a re-detection and re-engineering process is triggered, returning to step S105 to reacquire multimodal defect detection data of the abnormal sub-region and re-execute subsequent steps until a reliable evaluation result is obtained.

[0142] Secondly, this application also discloses an intelligent evaluation system for the construction quality of thermal insulation layers.

[0143] Reference Figure 5 A smart evaluation system for the construction quality of thermal insulation layers, comprising: The parameter acquisition module is used to acquire the measured thermal parameters and design thermal parameters corresponding to different detection sub-regions in the target insulation area; The deviation calculation module is used to obtain the parameter deviation value based on the measured thermal parameters and the design thermal parameters; The deviation judgment module is used to determine whether the parameter deviation value exceeds the preset deviation threshold; The first evaluation module is used to evaluate the overall quality level of the target insulation area based on the preset evaluation rules if the parameter deviation value does not exceed the preset deviation threshold. The feature extraction module is used to acquire multimodal defect detection data of abnormal sub-regions if the parameter deviation value exceeds the preset deviation threshold, and to extract features from the multimodal defect detection data to obtain defect feature vectors. The reverse solving module is used to input the defect feature vector into the reverse solving neural network to obtain the construction behavior deviation vector of each construction process corresponding to the abnormal sub-region; The scoring acquisition module is used to obtain the compliance score of each construction process based on the construction behavior deviation vector; The second evaluation module is used to evaluate the overall quality level of the target insulation area based on the compliance score of construction behavior and preset evaluation rules.

[0144] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for intelligent evaluation of the construction quality of thermal insulation layers, characterized in that, include: Obtain the measured thermal parameters and design thermal parameters corresponding to different detection sub-regions within the target insulation area; Based on the measured thermal parameters and the design thermal parameters, the parameter deviation value is obtained; Determine whether the parameter deviation value exceeds a preset deviation threshold; If the parameter deviation value does not exceed the preset deviation threshold, the overall quality level of the target insulation area is evaluated based on the preset evaluation rules. If the parameter deviation value exceeds the preset deviation threshold, then the multimodal defect detection data of the abnormal sub-region is obtained, and feature extraction is performed on the multimodal defect detection data to obtain the defect feature vector. The defect feature vector is input into the inverse solving neural network to obtain the construction behavior deviation vector of each construction process corresponding to the abnormal sub-region; Based on the construction behavior deviation vector, the compliance score of construction behavior for each construction procedure is obtained; Based on the construction behavior compliance score and the preset evaluation rules, the overall quality level of the target insulation area is evaluated.

2. The intelligent evaluation method for the construction quality of thermal insulation layer according to claim 1, characterized in that, If the parameter deviation value does not exceed the preset deviation threshold, then based on the preset evaluation rules, the overall quality level assessment of the target insulation area includes: If the parameter deviation value does not exceed the preset deviation threshold, the thermal parameter difference between adjacent detection sub-regions in the target insulation area is obtained; Determine whether the difference in the thermal parameters exceeds the local difference threshold; If the difference in thermal parameters does not exceed the local difference threshold, the overall quality level of the target insulation area is evaluated based on the preset evaluation rules. If the difference in thermal parameters exceeds the local difference threshold, the boundary position of the adjacent detection sub-region is obtained; Determine whether the boundary position coincides with the board seam position or the inside / outside corner position; If the boundary position coincides with the board seam position or the internal and external corner position, the difference in thermal parameters is determined to be a structural temperature difference, and the overall quality level of the target insulation area is evaluated based on the preset evaluation rules. If the boundary position does not coincide with the board seam position and the inside and outside corner position, then the quality level of the adjacent detection sub-regions is evaluated separately, and the lowest evaluated quality level is taken as the overall quality level of the target insulation area.

3. The intelligent evaluation method for the construction quality of thermal insulation layer according to claim 1, characterized in that, The multimodal defect detection data includes visible light images, infrared thermal imaging spectra, array ultrasonic data, and laser three-dimensional point cloud data; the feature extraction of the multimodal defect detection data to obtain the defect feature vector includes: Cracks are identified in the visible light image to obtain crack width, crack length, and crack distribution direction; Temperature field analysis is performed on the infrared thermal imaging spectrum to obtain the infrared temperature difference amplitude and the boundary of the hollow area; Impedance analysis was performed on the array ultrasonic data to obtain the debonding depth and internal void distribution; Deformation analysis is performed on the laser three-dimensional point cloud data to obtain the deformation amplitude and thickness deviation of the plate surface; Obtain the design layout information of the target insulation area, which includes the position of the board joint, the position of the anchor bolt, and the position of the inside and outside corners; Based on the design layout information, the relative spatial distances between the abnormal sub-region and the board seam position, the anchor bolt position, and the inside and outside corner position are obtained; The crack width, crack length, crack distribution direction, infrared temperature difference amplitude, hollow area boundary, debonding depth, internal void distribution, plate surface deformation amplitude, thickness deviation value, and relative spatial distance are combined to form the defect feature vector.

4. The intelligent evaluation method for the construction quality of thermal insulation layer according to claim 1, characterized in that, The step of inputting the defect feature vector into the inverse solving neural network to obtain the construction behavior deviation vector of each construction procedure corresponding to the abnormal sub-region includes: Obtain the defect decoupling features output by the encoder of the inverse solving neural network after performing dimensionality reduction analysis on the defect feature vector; The initial construction behavior deviation vectors of each construction procedure are obtained by the decoder of the reverse solving neural network after it reverse solves the defect decoupling features. The reverse solving process of the decoder is constrained by physical constraint equations. Determine whether the deviation value of each process in the initial construction behavior deviation vector is less than a preset deviation threshold; If the deviation value of each process is less than the preset deviation threshold, then the initial construction behavior deviation vector is output as the construction behavior deviation vector. If at least one process has a deviation value not less than the preset deviation threshold, then the construction and maintenance environment record of the abnormal sub-region is obtained. The initial construction behavior deviation vector is corrected based on the construction and maintenance environment records, and the corrected construction behavior deviation vector is output as the construction behavior deviation vector.

5. The intelligent evaluation method for the construction quality of thermal insulation layer according to claim 4, characterized in that, The correction of the initial construction behavior deviation vector based on the construction and maintenance environment records includes: Obtain the ambient temperature and humidity sequences of the abnormal sub-region during construction and maintenance; Based on the ambient temperature sequence and the ambient humidity sequence, the maintenance environment deviation index is obtained; Determine whether the maintenance environment deviation index exceeds the environment deviation threshold; If the maintenance environment deviation index does not exceed the environment deviation threshold, then the initial construction behavior deviation vector is used as the corrected construction behavior deviation vector. If the maintenance environment deviation index exceeds the environment deviation threshold, then the maintenance process deviation value and the non-maintenance process deviation value are extracted from the initial construction behavior deviation vector. Based on the maintenance environment deviation index, the deviation values ​​of the maintenance process are attenuated and the deviation values ​​of the non-maintenance process are amplified. The deviation values ​​of the maintenance procedures after attenuation treatment and the deviation values ​​of the non-maintenance procedures after amplification treatment are combined to form the corrected construction behavior deviation vector.

6. The intelligent evaluation method for the construction quality of thermal insulation layer according to claim 1, characterized in that, The process of obtaining the compliance score for each construction procedure based on the construction behavior deviation vector includes: Obtain the deviation values ​​of adhesive powder ratio and water addition ratio in the adhesive preparation process from the construction behavior deviation vector; Determine whether the deviation of the adhesive powder ratio is greater than a first target threshold or whether the deviation of the water addition ratio is greater than a second target threshold; If the deviation of the adhesive powder ratio is greater than the first target threshold or the deviation of the water addition ratio is greater than the second target threshold, then the adhesive preparation process will be marked as a target violation process. If the deviation of the adhesive powder ratio is not greater than the first target threshold and the deviation of the water addition ratio is not greater than the second target threshold, then the anchor bolt torque deviation value of the anchoring construction process is obtained. If the deviation value of the anchor bolt torque is greater than the third target threshold, the anchoring construction process will be marked as a target violation process; Obtain the deviation value of each process in the construction behavior deviation vector, and calculate the initial compliance score of each process based on the deviation value of each process and the preset deviation weight; If the target non-compliant process exists, the initial compliance score of the target non-compliant process will be penalized to obtain the construction behavior compliance score of each process; If no such violation process exists, the initial compliance score will be used as the compliance score for the construction behavior of each process.

7. The intelligent evaluation method for the construction quality of thermal insulation layer according to claim 6, characterized in that, The initial compliance score of the target non-compliant process is penalized to obtain the compliance score of the construction behavior for each process, including: Obtain the deviation value of the target non-compliant process, and determine the penalty coefficient based on the deviation value. The larger the deviation value, the larger the penalty coefficient. Obtain the defect type and defect area corresponding to the target violation process, and obtain the penalty intensity base based on the defect type and defect area; Based on the penalty coefficient and the penalty intensity base, calculate the score deduction value for the target violation process; The initial compliance score of the target non-compliant process is subtracted from the score deduction value to obtain the construction behavior compliance score of the target non-compliant process; Obtain the initial compliance score of non-target non-compliance procedures as the compliance score of construction behavior for non-target non-compliance procedures; The compliance scores of the construction behavior of the target non-compliant process and the compliance scores of the construction behavior of the non-target non-compliant process are combined to form the compliance score of the construction behavior of each process.

8. The intelligent evaluation method for the construction quality of thermal insulation layer according to claim 1, characterized in that, The overall quality rating of the target insulation area based on the construction behavior compliance score and the preset evaluation rules includes: Obtain the parameter deviation values ​​of each detection sub-region in the target insulation area, and obtain the initial quality level of the entity based on the parameter deviation values ​​of each detection sub-region and the preset evaluation rules; Obtain the compliance score of the construction behavior of each construction process corresponding to each detection sub-region, and obtain the behavior quality correction coefficient based on the compliance score of the construction behavior of each construction process; Determine whether there is a target non-compliant process in the target insulation area; If the target non-compliant process exists, the initial quality level of the entity will be lowered by at least one level based on the behavior quality correction coefficient to obtain the overall quality level of the target insulation area; If the target violation process does not exist, then determine whether the behavior quality correction coefficient is less than the correction threshold; If the behavior quality correction coefficient is less than the correction threshold, the initial quality level of the entity is downgraded based on the behavior quality correction coefficient to obtain the overall quality level of the target insulation area. If the behavior quality correction coefficient is not less than the correction threshold, then the initial quality level of the entity is taken as the overall quality level of the target insulation area.

9. The intelligent evaluation method for the construction quality of thermal insulation layer according to claim 1, characterized in that, Also includes: Based on the construction behavior deviation vector, the target process with the largest deviation value and the corresponding construction period of the target process are obtained; Based on the construction period, retrieve the corresponding video clips from the construction site monitoring video. Behavior recognition is performed on the video clips to obtain actual construction operation parameters; Determine whether the deviation values ​​of the actual construction operation parameters and the corresponding procedures in the construction behavior deviation vector point to the same violation direction; If they all point to the same direction of violation, the defect feature vector, the construction behavior deviation vector, and the video clip will be associated and stored as evidence chain data. If the error does not point to the same direction, the multimodal defect detection data of the abnormal sub-region is reacquired and the reverse solution is performed again.

10. An intelligent evaluation system for the construction quality of thermal insulation layers, characterized in that, include: The parameter acquisition module is used to acquire the measured thermal parameters and design thermal parameters corresponding to different detection sub-regions in the target insulation area; The deviation calculation module is used to obtain the parameter deviation value based on the measured thermal parameters and the design thermal parameters; The deviation judgment module is used to determine whether the parameter deviation value exceeds a preset deviation threshold; The first evaluation module is used to evaluate the overall quality level of the target insulation area based on preset evaluation rules if the parameter deviation value does not exceed the preset deviation threshold. The feature extraction module is used to acquire multimodal defect detection data of the abnormal sub-region if the parameter deviation value exceeds the preset deviation threshold, and to extract features from the multimodal defect detection data to obtain a defect feature vector. The reverse solving module is used to input the defect feature vector into the reverse solving neural network to obtain the construction behavior deviation vector of each construction procedure corresponding to the abnormal sub-region; The scoring acquisition module is used to obtain the compliance score of each construction process based on the construction behavior deviation vector. The second evaluation module is used to evaluate the overall quality level of the target insulation area based on the compliance score of the construction behavior and the preset evaluation rules.

Images