Method for detecting defects of embedded part in prefabricated building

Abstract

The invention provides a method for detecting defects of an embedded part in a prefabricated building. The method comprises the following steps: acquiring three-dimensional point cloud data of the surface of the embedded part through a three-dimensional scanning device; extracting geometric feature parameters of the embedded part based on the point cloud data, wherein the geometric feature parameters comprise deformation quantity, position offset and surface roughness; according to the geometrical characteristic parameters, a defect judgment threshold value is calculated through a dynamic reference generation algorithm; logic comparison is carried out on the geometrical characteristic parameters and a threshold value, and the defect type and grade of the embedded part are judged; and outputting a defect detection result and generating a repair suggestion scheme. The prefabricated building construction quality and efficiency can be improved, and potential safety hazards and repair cost caused by defects of the embedded part are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of defect detection of embedded parts, and more specifically, the present invention relates to a method for detecting defects of embedded parts in prefabricated buildings. Background Art

[0002] In the construction of prefabricated buildings, the quality of embedded parts plays a crucial role in the safety and stability of building structures. Embedded parts are usually used to connect precast components to the main structure, and their installation accuracy and quality directly affect the overall performance of prefabricated buildings. Traditional methods for detecting embedded parts mainly rely on manual inspection and simple measuring tools. This method is not only inefficient but also easily affected by human factors, resulting in low detection accuracy. In addition, traditional methods are difficult to comprehensively detect subtle defects of embedded parts, especially in the detection of defects in complex shapes and hidden parts, there are significant limitations.

[0003] With the advancement of building industrialization and informatization, three-dimensional scanning technology has gradually been applied to the construction field. Three-dimensional scanning technology can quickly obtain three-dimensional point cloud data of the object surface, providing new possibilities for the accurate detection of embedded parts. However, in the prior art, the high-precision characteristics of three-dimensional scanning data have not been fully utilized, and there is a lack of effective defect judgment algorithms and dynamic threshold generation mechanisms, resulting in insufficient accuracy and reliability of detection results.

[0004] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the prior art: traditional detection methods are inefficient and inaccurate, and it is difficult to meet the requirements of rapid construction and high quality of prefabricated buildings; there is a lack of a dynamic detection and intelligent judgment mechanism for defects of embedded parts in the prior art, and it is impossible to effectively identify and classify defect types; in addition, the prior art fails to provide repair suggestions matching the detection results and cannot achieve the organic combination of detection and repair. Summary of the Invention

[0005] The present invention provides a method for detecting defects of embedded parts in prefabricated buildings, including:

[0006] S1. Obtain three-dimensional point cloud data of the surface of the embedded part through a three-dimensional scanning device;

[0007] S2. Based on the three-dimensional point cloud data, extract geometric feature parameters of the embedded part, including deformation amount, position offset amount, and surface roughness;

[0008] S3. According to the geometric feature parameters, calculate a defect judgment threshold through a dynamic reference generation algorithm;

[0009] S4. Logically compare the geometric feature parameters with the defect judgment threshold to determine the defect type and grade of the embedded part;

[0010] S5. Output the defect detection result and generate a repair suggestion plan.

[0011] Further, the step S1 includes:

[0012] S11. Use a laser scanner to scan the surface of the embedded part from multiple angles at a preset scanning density to generate an initial point cloud data set;

[0013] S12. Denoise and fuse the initial point cloud data set through an image stitching algorithm to obtain complete three-dimensional point cloud data, where the image stitching algorithm satisfies the following constraint conditions:

[0014]

[0015] where, ΔP is the mean stitching error, p i is the coordinate of the i-th stitching point, q i is the theoretical coordinate, n is the total number of stitching points, and δ is the preset error threshold.

[0016] Further, the step S2 includes:

[0017] S21. Convert the three-dimensional point cloud data into the local coordinate system of the embedded part, and calculate the deformation amount ΔD, the position offset amount ΔL, and the surface roughness R, where:

[0018]

[0019] ΔL = max(∥x max - x min ∥, ∥y max - y min ∥, ∥z max - z min ∥)

[0020]

[0021] where, (x c , y c , z c ) is the current point coordinate, (x r , y r , z r ) is the theoretical coordinate, (x max , x min ) is the coordinate extreme value, h i is the surface height measurement value, h is the average height, and N is the number of sampling points.

[0022] Further, the dynamic reference generation algorithm of the step S3 includes:

[0023] S31. Statistically analyze the geometric characteristic parameters of historical defect-free embedded parts, and calculate the mean value μ, variance σ 2 and range R hist ;

[0024] S32. Generate a dynamic threshold θ based on the difference between the parameters of the current embedded part and the historical data:

[0025] θ = α·μ + β·σ 2 + γ·R hist

[0026] where α, β, and γ are weight coefficients, and α + β + γ = 1.

[0027] Furthermore, the weight coefficients are determined in the following way:

[0028] S321. Construct a training set containing defect sample data under different working conditions;

[0029] S322. Use the backpropagation optimization algorithm to iteratively adjust α, β, and γ to maximize the defect recognition accuracy. The optimization objective function is:

[0030]

[0031] where T k is the correct recognition number of the kth type of defect, F k is the misjudgment number, and M is the total number of defect categories.

[0032] Furthermore, the defect determination in step S4 includes:

[0033] S41. If the deformation amount ΔD > θ1 and the position offset amount ΔL > θ2, it is determined as a structural deformation defect;

[0034] S42. If the surface roughness R > θ3 and the area of the continuous abnormal region exceeds A min , it is determined as a surface damage defect;

[0035] where θ1, θ2, and θ3 are dynamic thresholds, and A min is the preset minimum abnormal area.

[0036] Furthermore, the determination method for the area of the continuous abnormal region is:

[0037] S421. Perform grid processing on the surface roughness data, and divide it into square units with a side length of d;

[0038] S422. Mark the units with roughness exceeding θ3 as abnormal units;

[0039] S423. Use the region growing algorithm to merge adjacent abnormal units, and calculate the total area S of the merged regiontotal :

[0040] S total = n·d 2

[0041] Where n is the total number of merged abnormal units.

[0042] Furthermore, the generation of the repair suggestion plan in step S5 includes:

[0043] S51. Match the preset repair strategy library according to the defect type, and extract the material strength correction coefficient K m and the structural compensation amount ΔS;

[0044] S52. Generate a three-dimensional repair model, and its compensation amount calculation formula is:

[0045] ΔS = K m ·(ΔD + 0.5ΔL)

[0046] And map ΔS to the three-dimensional point cloud data of the embedded parts.

[0047] Furthermore, the material strength correction coefficient K m is determined by the following method:

[0048] S511. Detect the material type of the embedded parts and query its standard tensile strength σ b ;

[0049] S512. Adjust K according to the defect level m :

[0050]

[0051] Furthermore, the method also includes real-time feedback control:

[0052] S6. During the construction process, input the defect detection results into the control system of the assembly robot, and dynamically adjust the installation parameters of the embedded parts. The calculation formula for the adjustment amount ΔP is:

[0053] ΔP = K p ·ΔD + K i ·∫ΔLdt

[0054] Where K p is the proportionality coefficient, K i is the integral coefficient, and t is the construction time variable.

[0055] The above embodiments of the present invention have at least the following beneficial effects: The detection method of the present invention can effectively improve the efficiency and accuracy of defect detection of embedded parts in prefabricated buildings. By obtaining high-precision three-dimensional point cloud data of the surface of the embedded parts through a three-dimensional scanning device and combining a dynamic reference generation algorithm to calculate the defect judgment threshold, it is possible to quickly and accurately identify defect types and grades such as deformation, position offset, and surface roughness of the embedded parts. This method avoids the subjectivity and limitations of traditional manual detection, significantly improves the reliability and repeatability of the detection results, and is applicable to the quality control of embedded parts in large-scale prefabricated building construction.

[0056] In addition, the present invention can also generate a targeted repair suggestion plan according to the detection results. By matching a preset repair strategy library, combining a material strength correction coefficient and a structural compensation amount, a three-dimensional repair model suitable for the defect type and grade is generated and mapped to the three-dimensional point cloud data of the embedded parts. This process not only provides clear repair guidance for construction personnel, but also can be fed back to the assembly robot control system in real time to dynamically adjust the installation parameters of the embedded parts, thereby optimizing the construction process, reducing the rework cost and safety hazards caused by defects in the embedded parts, and improving the overall construction quality of prefabricated buildings. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] By referring to the following detailed description with reference to the accompanying drawings, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become readily understood. In the drawings, several embodiments of the present invention are shown in an exemplary rather than restrictive manner, wherein:

[0058] Figure 1 It is a flowchart showing the method for detecting defects of embedded parts in a prefabricated building provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to be able to convey the scope of the present invention fully to those skilled in the art.

[0060] Those skilled in the art know that the embodiments of the present invention can be implemented as a system, device, equipment, method, or computer program product. Therefore, the present invention can be specifically implemented in the following forms: completely hardware, completely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0061] It should be noted that the quantity of any element in the drawings is for illustration rather than limitation, and any naming is only for distinction without any restrictive meaning.

[0062] The following refers to Figure 1 , Figure 1 which is a schematic flow chart of a method for detecting defects in embedded parts in prefabricated buildings provided by an embodiment of the present invention. As Figure 1 shown, a method 100 for detecting defects in embedded parts in prefabricated buildings includes:

[0063] S1. Obtain the three-dimensional point cloud data of the surface of the embedded part through a three-dimensional scanning device;

[0064] S2. Based on the three-dimensional point cloud data, extract the geometric feature parameters of the embedded part, including the deformation amount, the position offset amount, and the surface roughness;

[0065] S3. According to the geometric feature parameters, calculate the defect judgment threshold through a dynamic reference generation algorithm;

[0066] S4. Logically compare the geometric feature parameters with the defect judgment threshold to determine the defect type and grade of the embedded part;

[0067] S5. Output the defect detection result and generate a repair suggestion plan.

[0068] It should be noted that when implementing the method for detecting defects in embedded parts, first, it is necessary to obtain the three-dimensional point cloud data of the surface of the embedded part through a three-dimensional scanning device. The three-dimensional point cloud data refers to the three-dimensional coordinate set of a large number of points on the surface of an object obtained by a scanning device, which can accurately reflect the shape and size of the object. The embedded part is an important component used to connect precast components and the main structure in prefabricated buildings, and its quality directly affects the safety and stability of the building. The point cloud data obtained through three-dimensional scanning technology provides the basic data support for subsequent defect detection. In actual operation, high-precision devices such as laser scanners can be used for scanning to ensure the integrity and accuracy of the data.

[0069] Specifically, a 3D scanning device can use a laser scanner, and the scanning density can be set according to the size and complexity of the embedded part. For example, under high-precision requirements, the scanning density can be set to more than 100 points per square centimeter. During the scanning process, the laser scanner scans the surface of the embedded part from multiple angles to generate an initial point cloud data set. These data sets need to be denoised and fused through an image stitching algorithm to eliminate possible errors and noises during the scanning process. The average stitching error should be less than a preset error threshold, such as 0.1 mm, to ensure the accuracy of the finally generated 3D point cloud data. In addition, the extraction of geometric feature parameters includes deformation amount, position offset, and surface roughness, and these parameters are extracted from the point cloud data through specific calculation formulas to provide a basis for subsequent defect judgment.

[0070] Preferably, to further improve the accuracy and reliability of detection, the scanning process can be optimized. For example, during multi-angle scanning, multiple scanning paths can be set to ensure that every part of the embedded part can be fully scanned. For embedded parts with complex shapes, higher-precision scanning equipment can be used, or the number of scans can be increased to improve the integrity and accuracy of the data. In the image stitching algorithm, more advanced denoising techniques, such as deep learning-based denoising algorithms, can be introduced to further reduce the impact of noise on the detection results.

[0071] Furthermore, the dynamic reference generation algorithm can be optimized according to the data of historical defect-free embedded parts. By adjusting the calculation method of the weight coefficient, the adaptability and accuracy of the defect judgment threshold can be improved. These optimization measures can effectively enhance the overall performance of the defect detection of embedded parts.

[0072] In some embodiments, step S1 includes:

[0073] S11. Use a laser scanner to perform multi-angle scanning on the surface of the embedded part at a preset scanning density to generate an initial point cloud data set;

[0074] S12. Denoise and fuse the initial point cloud data set through an image stitching algorithm to obtain complete 3D point cloud data, where the image stitching algorithm satisfies the following constraint conditions:

[0075]

[0076] where ΔP is the average stitching error, p i is the coordinate of the i-th stitching point, q i is the theoretical coordinate, n is the total number of stitching points, and δ is the preset error threshold.

[0077] It should be noted that when implementing the method for detecting defects in embedded parts, it is necessary to use a laser scanner to scan the surface of the embedded part from multiple angles at a preset scanning density to generate an initial point cloud data set. A laser scanner is a device that uses a laser beam to scan the surface of an object and obtain its three-dimensional coordinates, capable of quickly and accurately generating the point cloud data of the object's surface. The preset scanning density refers to the number of scanning points per unit area, which directly affects the resolution and accuracy of the point cloud data. Scanning from multiple angles is to ensure that all surfaces of the embedded part can be fully covered, avoiding data loss caused by occlusion or dead angles. The generated initial point cloud data set contains a large amount of point coordinate information on the surface of the embedded part, but there may be noise and discontinuities in these data, which need further processing. Subsequently, the initial point cloud data set is denoised and fused through an image stitching algorithm to obtain complete three-dimensional point cloud data. The image stitching algorithm is a technology used to merge multiple local point cloud data into a whole, and its constraint conditions ensure the accuracy and reliability of the stitching.

[0078] Specifically, the scanning density of the laser scanner can be adjusted according to the size and complexity of the embedded part. For example, for small or simple-shaped embedded parts, the scanning density can be set to 50 points per square centimeter; while for large or complex-shaped embedded parts, the scanning density can be increased to 100 points per square centimeter or higher. Scanning from multiple angles usually includes a 360° scan in the horizontal direction and multiple-angle scans in the vertical direction to ensure that all surfaces of the embedded part can be fully scanned. In the image stitching algorithm, the mean stitching error is an important indicator to measure the stitching accuracy, and its calculation formula is the mean of the difference between the stitching point coordinates and the theoretical coordinates. The preset error threshold can be set according to actual needs, such as 0.1 mm, to ensure the accuracy of the three-dimensional point cloud data after stitching. In addition, the total number of stitching points refers to the number of points participating in the stitching, and theoretically, it should cover the entire surface of the embedded part.

[0079] Preferably, in order to further improve the accuracy of scanning and stitching, dynamic scanning path planning can be introduced during the scanning process. For example, according to the shape and size of the embedded part, the scanning angle and position of the laser scanner are adjusted in real time to reduce the scanning blind area. In the image stitching algorithm, a stitching method based on feature point matching can be adopted, and accurate stitching is carried out by extracting feature points (such as edge points, corner points, etc.) in the point cloud data.

[0080] Furthermore, the calculation of the mean stitching error can combine multiple error evaluation indicators, such as the root mean square error (RMSE), etc., to more comprehensively evaluate the stitching quality. For embedded parts with complex shapes, a block stitching method can also be adopted. First, the surface of the embedded part is divided into multiple regions, and the point cloud data of each region is stitched separately, and finally, the point cloud data of each region is merged to improve the stitching efficiency and accuracy.

[0081] In some embodiments, step S2 includes:

[0082] S21. Convert the three-dimensional point cloud data into the local coordinate system of the embedded part, and calculate the deformation amount ΔD, the position offset amount ΔL, and the surface roughness R, where:

[0083]

[0084] ΔL = max(∥x max - x min ∥, ∥y max - y min ∥, ∥z max - z min ∥)

[0085]

[0086] where, (x c , y c , z c ) is the current point coordinate, (x r , y r , z r ) is the theoretical coordinate, (x max , x min ) are the coordinate extreme values, h i is the surface height measurement value, h is the average height, and N is the number of sampling points.

[0087] It should be noted that when implementing the defect detection method for the embedded part, converting the three-dimensional point cloud data into the local coordinate system of the embedded part is one of the key steps. The local coordinate system is a coordinate system with the embedded part itself as the reference, used to more accurately describe the geometric feature parameters of the embedded part, including the deformation amount, the position offset amount, and the surface roughness. The deformation amount refers to the deviation between the actual shape and the designed shape of the embedded part; the position offset amount refers to the difference between the actual installation position and the theoretical position of the embedded part; the surface roughness refers to the degree of microscopic unevenness of the surface of the embedded part. By calculating these parameters, the quality status of the embedded part can be quantitatively evaluated, providing data support for subsequent defect judgment.

[0088] Specifically, the calculation formula for the deformation amount is

[0089]

[0090] where, (x 实际 ,, y 实际 , z 实际 ) is the actual coordinate of the current point, (x 理论 , y 理论 , z 理论 ) is the theoretical coordinate. The calculation formula for the position offset amount is

[0091] ΔP = max(|x 实际 - x 理论 |, |y 实际 - y 理论 |, |z 实际 - z 理论 |)

[0092] For evaluating the maximum deviation degree of the embedded parts in three directions. The calculation formula of surface roughness is

[0093]

[0094] where h i is the measured surface height value of the i-th sampling point, is the average height of all sampling points, and N is the number of sampling points. These formulas provide a clear mathematical basis for the calculation of deformation amount, position offset amount and surface roughness, ensuring the scientificity and accuracy of the detection results.

[0095] Preferably, in order to further improve the detection accuracy, a weighting factor can be introduced in the calculation of the deformation amount to weight the deviation according to the importance of different directions. For example, for some embedded parts, the deviation in the vertical direction may be more important than that in the horizontal direction, and this difference can be reflected by setting the weighting factor. In the calculation of the position offset amount, in addition to evaluating the maximum offset amount, the average offset amount can also be calculated to more comprehensively evaluate the installation accuracy of the embedded parts. For the calculation of surface roughness, more advanced statistical methods, such as quantile analysis, can be used to better reflect the distribution characteristics of surface roughness. In addition, in order to improve the calculation efficiency, an optimization algorithm can be used to accelerate the calculation of the deformation amount and the position offset amount, for example, by reducing unnecessary calculation steps or using parallel computing technology.

[0096] In some embodiments, the dynamic reference generation algorithm in step S3 includes:

[0097] S31. Statistically analyze the geometric feature parameters of historical defect-free embedded parts, and calculate the mean value μ, variance σ2 and range R hist ;

[0098] S32. Generate a dynamic threshold θ according to the difference between the parameters of the current embedded part and the historical data:

[0099] θ = α·μ + β·σ 2 + γ·R hist

[0100] where α, β, γ are weight coefficients and satisfy α + β + γ = 1.

[0101] It should be noted that when implementing the defect detection method for embedded parts, the dynamic reference generation algorithm performs statistical analysis based on the geometric feature parameters of historical defect-free embedded parts, so as to generate a dynamic threshold for judging whether there are defects in the current embedded parts. The generation of the dynamic threshold takes into account the mean, variance, and range of the geometric feature parameters of the embedded parts. These statistics can reflect the fluctuation range of the embedded parts in the normal state. By calculating the dynamic threshold, it is possible to more flexibly adapt to the detection requirements of embedded parts under different working conditions, and avoid misjudgment or missed judgment caused by a fixed threshold. The core of this method lies in using the statistical characteristics of historical data and combining weight coefficients to evaluate the parameters of the current embedded parts, so as to achieve accurate defect judgment.

[0102] Specifically, the dynamic reference generation algorithm includes two main steps. First, it is necessary to perform statistical analysis on the geometric feature parameters of historical defect-free embedded parts and calculate their mean (μ), variance (σ 2 ) and range (R). The mean represents the average level of historical data, the variance reflects the degree of dispersion of the data, and the range is the difference between the maximum and minimum values of the data, which is used to evaluate the fluctuation range of the data. Secondly, according to the difference between the parameters of the current embedded part and the historical data, a dynamic threshold (T) is generated. The calculation formula for the dynamic threshold is T = a·μ + b·σ 2 + c·R, where a, b, and c are weight coefficients and satisfy a + b + c = 1. The setting of the weight coefficients can be adjusted according to the actual working conditions and detection requirements of the embedded parts to optimize the accuracy of defect identification.

[0103] Preferably, in order to further improve the adaptability and accuracy of the dynamic threshold, more optimization mechanisms can be introduced in the process of determining the weight coefficients. For example, by constructing a training set containing defect sample data under different working conditions, the backpropagation optimization algorithm can be used to iteratively adjust the weight coefficients to maximize the accuracy of defect identification.

[0104] Furthermore, the weight coefficients can also be set in layers according to the material, shape, and use environment of the embedded parts to better reflect the influence of different characteristic parameters on defect judgment. For example, for embedded parts with a relatively high surface roughness, the weight of the range can be appropriately increased; while for embedded parts with high requirements for position accuracy, the weight of the mean can be increased. In addition, the calculation of the dynamic threshold can also be combined with real-time data feedback to dynamically adjust the weight coefficients according to the changes in the current detection environment, thereby further enhancing the flexibility and reliability of the detection.

[0105] In some embodiments, the weight coefficients are determined in the following manner:

[0106] S321. Construct a training set containing defect sample data under different working conditions;

[0107] S322. Use the backpropagation optimization algorithm to iteratively adjust α, β, and γ to maximize the defect recognition accuracy. The optimization objective function is:

[0108]

[0109] where T k is the number of correctly recognized defects of the k-th type, F k is the number of misjudgments, and M is the total number of defect categories.

[0110] It should be noted that when implementing the embedded part defect detection method, the determination of the weight coefficients is achieved by constructing a training set and using the backpropagation optimization algorithm, with the aim of maximizing the defect recognition accuracy. The training set contains defect sample data under different working conditions, and these sample data reflect various defect types and degrees that may occur in the embedded parts during actual use. The backpropagation optimization algorithm is an optimization method based on neural networks. By iteratively adjusting the weight coefficients, the defect recognition results output by the model are made as close as possible to the actual defect situation. The optimization objective function is used to measure the performance of the model. By minimizing the misjudgment rate and maximizing the correct recognition rate, the reasonable distribution of the weight coefficients is ensured, thereby improving the accuracy and reliability of the entire detection system.

[0111] Specifically, the construction of the training set requires collecting a large number of embedded part defect samples under different working conditions. These samples should cover various common defect types such as deformation, position offset, and surface damage. Each sample data includes the geometric feature parameters of the embedded part (such as the amount of deformation, position offset, and surface roughness) and the corresponding defect category and grade. The backpropagation optimization algorithm adjusts the weight coefficients by calculating the objective function, and the form of the objective function is

[0112]

[0113] where C k is the number of correctly recognized defects of the k-th type, E k is the number of misjudgments, and M is the total number of defect categories. The algorithm iteratively updates the weight coefficients a, b, and c to maximize the value of the objective function L, thereby optimizing the defect recognition accuracy. The initial values of the weight coefficients can be set randomly or estimated based on experience, but the final values need to be determined by the optimization algorithm.

[0114] Preferably, in order to further improve the efficiency and accuracy of the weight coefficient optimization, more optimization strategies can be introduced during the training process. For example, a momentum term can be used to accelerate the convergence speed of the backpropagation algorithm and reduce the training time. In addition, the training set can be divided by the method of cross-validation, dividing the data into a training set and a validation set to prevent the model from overfitting and ensure the stability and generalization ability of the weight coefficients on different data sets.

[0115] Furthermore, in practical applications, the objective function can also be adjusted according to the specific working conditions and inspection requirements of the embedded parts. For example, the weights for certain key defect types can be increased to improve the recognition accuracy of important defects. In addition, to improve the robustness of training, data augmentation techniques can be introduced. By randomly perturbing or transforming the training samples, the diversity of the samples can be increased, thereby enhancing the adaptability of the model to different working conditions.

[0116] In some embodiments, the defect determination in step S4 includes:

[0117] S41. If the deformation amount ΔD > θ1 and the position offset amount ΔL > θ2, it is determined as a structural deformation defect;

[0118] S42. If the surface roughness R > θ3 and there is a continuous abnormal area exceeding A min , it is determined as a surface damage defect;

[0119] where θ1, θ2, and θ3 are dynamic thresholds, and A min is the preset minimum abnormal area.

[0120] It should be noted that when implementing the defect detection method for embedded parts, the defect determination is completed based on the logical comparison of geometric feature parameters such as deformation amount, position offset amount, and surface roughness with dynamic thresholds. This process aims to compare the actual state of the embedded parts with the normal range by setting clear determination conditions, so as to accurately identify whether there are defects and the types of defects. For example, when the deformation amount and the position offset amount both exceed the set dynamic thresholds, it can be determined as a structural deformation defect; while when the surface roughness exceeds the threshold and there is a continuous abnormal area exceeding the preset value, it is determined as a surface damage defect. This logical comparison method can effectively distinguish different types of defects and provide a basis for subsequent repair.

[0121] Specifically, the determination condition for a structural deformation defect is that the deformation amount ΔL is greater than the dynamic threshold T1 and the position offset amount ΔP is greater than the dynamic threshold T2. The dynamic thresholds T1 and T2 here are obtained based on the statistical normal state of the embedded parts and are used to determine whether the deformation and position offset exceed the allowable range. The determination of surface damage defects involves the surface roughness R and the area A 异常 of the continuous abnormal area. When the surface roughness R exceeds the dynamic threshold T3, if there is a continuous abnormal area with an area exceeding the preset minimum abnormal area A min at the same time, it is determined as surface damage. The preset minimum abnormal area A min is set according to the actual use requirements and surface quality standards of the embedded parts. For example, it can be set to 10 square millimeters to ensure that only significant surface damage will be recognized.

[0122] Preferably, in order to further improve the accuracy and reliability of defect determination, more refined conditions can be introduced during the determination process. For example, for structural deformation defects, in addition to the threshold judgment of the amount of deformation and the position offset, the stress direction of the embedded parts and the actual working conditions can be combined, and different weights can be set for deformations and offsets in different directions, so as to more accurately evaluate the impact degree of the defects. For surface damage defects, in addition to area judgment, the shape of the abnormal area can also be analyzed, for example, by identifying the trend and branching of cracks to distinguish different types of surface damage.

[0123] Furthermore, the dynamic threshold can be adjusted according to the material and use environment of the embedded parts. For example, for embedded parts made of high-strength steel, the thresholds of deformation and roughness can be appropriately increased to adapt to their higher strength and wear resistance.

[0124] In some embodiments, the determination method for the area of the continuous abnormal region is as follows:

[0125] S421. Perform grid processing on the surface roughness data, and divide it into square units with a side length of d;

[0126] S422. Mark the units with roughness exceeding θ3 as abnormal units;

[0127] S423. Use the region growing algorithm to merge adjacent abnormal units, and calculate the total area S of the merged region total :

[0128] S total = n·d 2

[0129] where n is the total number of merged abnormal units.

[0130] It should be noted that when implementing the defect detection method for embedded parts, the determination of the area of the continuous abnormal region is achieved by performing grid processing on the surface roughness data and using the region growing algorithm. Grid processing is to divide the surface of the embedded part into multiple small units, and the side length of each unit is a preset value for evaluating the local surface roughness. The region growing algorithm is an image processing technology used to identify and merge adjacent abnormal units, so as to calculate the total area of the continuous abnormal region. This method can effectively identify the distribution range of surface damage and provide a quantitative basis for the determination of surface damage defects.

[0131] Specifically, during the meshing process, the surface of the embedded part is divided into square grid cells with side length d. The roughness value of each cell is determined by calculating the average surface roughness of all sampling points within the cell. The preset grid side length d can be adjusted according to the size and surface characteristics of the embedded part. For example, for small embedded parts, the grid side length can be set to 1 mm; for large embedded parts, the grid side length can be set to 5 mm. In the region growing algorithm, first, the cells with roughness exceeding the threshold T3 are marked as abnormal cells, and then the adjacent abnormal cells are searched and merged. The total area A of the merged abnormal region 异常 Through the formula A 异常 = n·d 2 is calculated, where n is the total number of merged abnormal cells. The preset minimum abnormal area A min is a key parameter for distinguishing normal surface fluctuations and actual damages, and is usually set according to the usage requirements and surface quality standards of the embedded part. For example, it can be set to 20 square millimeters.

[0132] Preferably, in order to further improve the accuracy and efficiency of the determination of the area of continuous abnormal regions, an adaptive grid division strategy can be introduced in the meshing process. For example, for regions with large roughness variations, smaller grid side lengths can be used to improve the resolution; while for regions with small roughness variations, larger grid side lengths can be used to reduce the calculation amount. In the region growing algorithm, a directional constraint can be introduced to ensure that the merged abnormal cells are physically continuous, thus avoiding misjudgment caused by accidental roughness fluctuations.

[0133] Furthermore, machine learning algorithms can also be combined to analyze the shape and distribution patterns of abnormal regions, further optimizing the identification and determination of abnormal regions. For example, by training a classifier to identify the characteristic patterns of different types of damages such as cracks and scratches, the accuracy and reliability of the determination of surface damage defects can be improved.

[0134] In some embodiments, the generation of the repair suggestion plan in step S5 includes:

[0135] S51. Matching the preset repair strategy library according to the defect type, and extracting the material strength correction coefficient K m and the structural compensation amount ΔS;

[0136] S52. Generating a three-dimensional repair model, and its compensation amount calculation formula is:

[0137] ΔS = K m ·(ΔD + 0.5ΔL)

[0138] and mapping ΔS to the three-dimensional point cloud data of the embedded part.

[0139] It should be noted that when implementing the pre-embedded part defect detection method, the generation of the repair suggestion plan is based on matching the defect type with a preset repair strategy library and is completed in combination with the material strength correction coefficient and the structural compensation amount. The repair strategy library is a database containing various defect types and their corresponding repair methods, and is used to provide targeted repair plans for different types of defects. The material strength correction coefficient is a coefficient adjusted according to the material and defect level of the pre-embedded part, and is used to ensure that the repaired pre-embedded part meets the structural strength requirements. The structural compensation amount is the repair size calculated according to the specific situation of the defect, and is used to compensate for the dimensional deviation caused by the defect. Through these steps, a three-dimensional repair model suitable for the defect type and level can be generated, providing clear guidance for the actual repair work.

[0140] Specifically, the construction of the repair strategy library needs to cover various defect types that may occur in the pre-embedded part, such as deformation, position offset, and surface damage, etc., and provide detailed repair methods and steps for each defect type. The determination of the material strength correction coefficient is based on the standard tensile strength of the material of the pre-embedded part. For example, for first-level defects, the correction coefficient is 1.2 times the standard tensile strength; for second-level defects, the correction coefficient is 1.0 times; for third-level defects, the correction coefficient is 0.8 times. The calculation formula for the structural compensation amount is ΔC = K s ·(C d + 0.5ΔC d ), where K s is the material strength correction coefficient, C d is the preset repair compensation amount, and ΔC d is the compensation amount increment adjusted according to the defect level. The setting of these parameters needs to be optimized in combination with the actual use environment and safety requirements of the pre-embedded part to ensure that the repaired pre-embedded part can meet the requirements of structural strength and use function.

[0141] Preferably, in order to further improve the practicability and accuracy of the repair suggestion plan, more detailed repair methods can be introduced into the repair strategy library. For example, for surface damage defects, multiple options such as local grinding, coating repair, or component replacement can be provided for selection. When calculating the structural compensation amount, the stress situation and installation position of the pre-embedded part can be considered, and the compensation amount for different parts can be adjusted differentially. In addition, the repair suggestion plan can be combined with three-dimensional visualization technology to generate an intuitive three-dimensional repair model and map it to the three-dimensional point cloud data of the pre-embedded part, so that construction personnel can more intuitively understand and execute the repair operation. This visualization solution not only improves the repair efficiency but also reduces repair errors caused by understanding errors.

[0142] In some embodiments, the material strength correction coefficient K m is determined by the following method:

[0143] S511. Detect the material type of the embedded part and query its standard tensile strength σ b ;

[0144] S512. Adjust K according to the defect level m :

[0145]

[0146] It should be noted that when implementing the defect detection method for embedded parts, the determination of the material strength correction factor is adjusted based on the material type of the embedded part and its standard tensile strength. The standard tensile strength refers to the maximum stress that the material can withstand in a tensile test and is an important indicator to measure the material strength. According to different defect levels, the material strength correction factor will adjust the standard tensile strength to ensure that the repaired embedded part can meet the structural safety requirements. This method can flexibly adjust the repair strategy according to different materials and defect conditions, thereby improving the adaptability and reliability of the repair plan.

[0147] Specifically, the determination of the material strength correction factor is divided into two steps. First, detect the material type of the embedded part and query its corresponding standard tensile strength. Embedded parts of different materials (such as steel, aluminum alloy, etc.) have different standard tensile strength values, which can be obtained through material standard manuals or experimental data. Second, adjust the standard tensile strength according to the defect level. For example, for a first-level defect, the correction factor is 1.2 times the standard tensile strength, indicating that the repaired embedded part requires a higher strength reserve; for a second-level defect, the correction factor is 1.0 times, indicating that the strength remains unchanged; for a third-level defect, the correction factor is 0.8 times, indicating that a certain reduction in strength is allowed. This hierarchical adjustment method can reasonably allocate repair resources according to the severity of the defect, ensuring that the repaired embedded part not only meets the safety requirements but also avoids over-repair.

[0148] Preferably, in order to further improve the accuracy and applicability of the material strength correction factor, more precise non-destructive testing techniques, such as ultrasonic testing or spectral analysis, can be introduced during the material detection process to ensure the accurate identification of the material type. In addition, for embedded parts made of some special materials or under complex working conditions, the correction factor can be fine-tuned in combination with the actual use environment (such as a corrosive environment or a high-stress area). For example, in a highly corrosive environment, even if the defect level is low, the correction factor can be appropriately increased to enhance the durability of the embedded part. In addition, a dynamic material strength database can be established, and the calculation rules of the correction factor can be continuously updated according to actual detection data and repair feedback, thereby further optimizing the scientificity and practicality of the repair plan.

[0149] In some embodiments, the method further includes real-time feedback control:

[0150] S6. During the construction process, input the defect detection results into the control system of the assembly robot, and dynamically adjust the installation parameters of the embedded parts. The calculation formula for the adjustment amount ΔP is:

[0151] ΔP = K p ·ΔD + K i ·∫ΔLdt

[0152] Wherein, K p is the proportionality coefficient, K i is the integral coefficient, and t is the construction time variable.

[0153] It should be noted that when implementing the defect detection method for embedded parts, real-time feedback control is the key link to dynamically input the defect detection results into the control system of the assembly robot to adjust the installation parameters of the embedded parts. The control system of the assembly robot is an automated system used to control the robot to install the embedded parts during the construction of prefabricated buildings, and it can adjust the installation actions and parameters in real time according to the input detection results. In this way, it can ensure that the defects of the embedded parts are corrected in a timely manner during the installation process, thereby improving the construction quality and efficiency. The core of real-time feedback control lies in quickly calculating the adjustment amount based on the defect detection results and applying it to the motion control of the assembly robot to achieve the intelligentization and dynamic optimization of the construction process.

[0154] Specifically, the calculation formula for the adjustment amount of real-time feedback control is ΔP = K p ·K i + K i ·∫edt, where K p is the proportionality coefficient, K i is the integral coefficient, e is the detected defect deviation, and t is the construction time variable. The proportionality coefficient K p is used to directly adjust the installation parameters according to the current deviation, while the integral coefficient K i is used to eliminate the cumulative error and ensure the long-term installation accuracy. The settings of these two parameters need to be optimized according to the installation accuracy requirements of the embedded parts and the response characteristics of the robot system. For example, for the embedded parts with high-precision installation requirements, the proportionality coefficient can be appropriately increased to quickly respond to the deviation, and the integral coefficient can be reasonably set to avoid over-adjustment. In addition, the introduction of the construction time variable t enables the adjustment amount to change dynamically according to the construction progress, further improving the flexibility and adaptability of the installation.

[0155] Preferably, in order to further improve the accuracy and stability of real-time feedback control, a differential term can be introduced into the adjustment amount calculation to form a proportional-integral-differential (PID) controller. The differential term can adjust the installation parameters in advance according to the change rate of the deviation, thereby reducing the hysteresis and overshoot phenomena of the system.

[0156] Furthermore, the proportionality coefficient and the integral coefficient can be set in segments according to the installation position and the stress conditions of the embedded parts. For example, at the key stress parts of the embedded parts, the sensitivity of the adjustment amount can be appropriately increased to ensure the installation accuracy; while at the non-key parts, the adjustment amount can be appropriately reduced to improve the construction efficiency. At the same time, machine learning algorithms can be combined to analyze the historical installation data, and the values of the proportionality coefficient and the integral coefficient can be dynamically optimized, so as to achieve more accurate real-time feedback control.

[0157] The above-mentioned embodiments of the present invention have the following beneficial effects: By obtaining the three-dimensional point cloud data on the surface of the embedded parts through three-dimensional scanning technology and extracting geometric feature parameters based on this, high-precision defect detection can be achieved. Combining the dynamic reference generation algorithm to calculate the defect judgment threshold can quickly determine the defect type and grade of the embedded parts, and at the same time generate a repair suggestion plan, providing clear guidance for construction. In addition, using a laser scanner and an image stitching algorithm to obtain complete three-dimensional point cloud data, and determining the weight coefficient and defect judgment conditions through an optimization algorithm can further improve the accuracy and reliability of the detection. This method not only improves the detection efficiency, but also ensures the installation quality of the embedded parts, meeting the high-precision requirements of prefabricated building construction.

[0158] At the same time, this method also has a real-time feedback control function, which can dynamically input the defect detection results into the control system of the assembly robot to adjust the installation parameters of the embedded parts in real time. In this way, the construction process can be effectively optimized, the rework cost and safety hazards caused by the defects of the embedded parts can be reduced, and the overall construction quality of the prefabricated building can be improved. In addition, matching the repair strategy library according to the defect type and generating a three-dimensional repair model can provide targeted repair plans for different grades of defects, further enhancing the practicability and adaptability of this detection method.

[0159] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement all the above methods. Among them, the program instructions can be stored in the above storage medium in the form of a software product, including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: various media that can store program codes such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs, or terminal devices such as computers, servers, mobile phones, and tablets.

[0160] The above description is only some preferred embodiments of the present invention and an explanation of the technical principles applied. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) having similar functions disclosed in the embodiments of the present invention.

Claims

1. A method for detecting defects of embedded parts in prefabricated buildings, characterized in that, It includes the following steps: S1. Obtain the three-dimensional point cloud data of the surface of the embedded part through a three-dimensional scanning device; the three-dimensional scanning device is a laser scanner; S2. Based on the three-dimensional point cloud data, extract the geometric feature parameters of the embedded part, and the geometric feature parameters include deformation amount, position offset amount, and surface roughness; based on the three-dimensional point cloud data, extracting the geometric feature parameters of the embedded part includes converting the three-dimensional point cloud data into the local coordinate system of the embedded part, and calculating the deformation amount, position offset amount, and surface roughness respectively based on the local coordinate system of the embedded part; S3. Calculate the defect judgment threshold through the dynamic reference generation algorithm according to the geometric feature parameters, including statistically analyzing the geometric feature parameters of historical defect-free embedded parts, calculating the mean value μ, variance σ 2 and range R hist ; generate a dynamic threshold θ according to the difference between the parameters of the current embedded part and the historical data; S4. Logically compare the geometric feature parameters with the defect judgment threshold to determine the defect type and grade of the embedded part; the defect type includes structural deformation defect and surface damage defect; S5. Output the defect detection result and generate a repair suggestion plan.

2. The detection method according to claim 1, wherein The step S1 includes: S11. Use a laser scanner to scan the surface of the embedded part from multiple angles with a preset scanning density to generate an initial point cloud data set; S12. Denoise and fuse the initial point cloud data set through an image stitching algorithm to obtain complete three-dimensional point cloud data, where the image stitching algorithm satisfies the following constraint conditions: Among them, ΔP is the mean value of the splicing error, p i is the coordinate of the i-th splicing point, q i is the theoretical coordinate, n is the total number of splicing points, and δ is the preset error threshold.

3. The detection method according to claim 1, characterized in that, The step S2 converts the three-dimensional point cloud data into the local coordinate system of the embedded part, and calculates the deformation amount, position offset amount, and surface roughness respectively, as shown in the following formula: ΔL = max(||x max - x min ||, ||y max - y min ||, ||z max - z min ||) Among them, (x c , y c , z c ) is the current point coordinate, (x r , y r , z r ) is the theoretical coordinate, (x max , x min ) are the coordinate extrema, h i is the surface height measurement value, h is the average height, N is the number of sampling points, ΔD is the deformation amount, ΔL is the position offset, and R is the surface roughness.

4. The detection method according to claim 1, wherein In the step S3, according to the difference between the parameters of the current embedded part and the historical data, a dynamic threshold θ is generated as shown in the following formula: θ = α·μ + β·σ 2 + γ·R hist Among them, α, β, and γ are weight coefficients, and α + β + γ = 1.

5. The detection method according to claim 4, wherein The weight coefficients are determined by the following method: S321. Construct a training set containing defect sample data under different working conditions; S322. Use the backpropagation optimization algorithm to iteratively adjust α, β, and γ to maximize the defect recognition accuracy rate, and the optimization objective function is: Among them, T k is the correct recognition number of the k-th type of defect, F k is the misjudgment number, and M is the total number of defect categories.

6. The detection method according to claim 1, wherein The step S4 logically compares the geometric feature parameters with the defect judgment threshold to determine the defect type of the embedded part, including: S41. If the deformation amount ΔD > θ1 and the position offset amount ΔL > θ2, it is determined as a structural deformation defect; S42. If the surface roughness R > θ3 and there is a continuous abnormal area with an area exceeding A min , it is determined as a surface damage defect; Among them, θ1, θ2, and θ3 are dynamic thresholds, and A min is the preset minimum abnormal area.

7. The detection method according to claim 6, characterized in that The determination method of the continuous abnormal area is: S421. Perform grid processing on the surface roughness data, and divide square units with a side length of d; S422. Mark the units with roughness exceeding θ3 as abnormal units; S423. Adopt the region growing algorithm to merge adjacent abnormal cells, and calculate the total area S of the merged region total : S total = n·d 2 Among them, n is the total number of merged abnormal units.

8. The detection method according to claim 1, wherein The generation of the repair suggestion plan in the step S5 includes: S51. Match the preset repair strategy library according to the defect type, and extract the material strength correction coefficient K m and the structural compensation amount ΔS; S52. Generate a three-dimensional repair model, obtain a compensation amount ΔS based on the three-dimensional repair model, and map the compensation amount ΔS into the three-dimensional point cloud data of the embedded part; where The calculation formula of the three-dimensional repair model is: ΔS = K m ·(ΔD + 0.5ΔL).

9. The detection method according to claim 8, wherein The material strength correction coefficient K m is determined as follows: S511. Detect the material type of the embedded part and query its standard tensile strength σ b ; S512. Adjust the material strength correction coefficient K according to the defect level m :

10. The detection method according to claim 1, wherein The method also includes real-time feedback control: S6. During the construction process, input the defect detection result into the control system of the assembly robot to dynamically adjust the installation parameters of the embedded part, and the calculation formula of the adjustment amount ΔP is: ΔP = K p ·ΔD + K i ·∫ΔLdt Among them, K p is the proportionality coefficient, and K i is the integral coefficient, and t is the construction time variable.

Images