Invisible damage batch detection system and analysis method for historic building repair

Abstract

The invention relates to the technical field of historic building repair, in particular to an invisible damage batch detection system for historic building repair and an analysis method. The system comprises a detection system used for carrying out invisible damage detection on building bodies before and after historic building repair in batches, the detection system comprises a driving assembly and a detection assembly, the system further comprises a server, and a sensor module transmits detected invisible damage data of the building bodies to the server through a data line. Through cooperation of the mechanical arm, the telescopic frame and the multiple sensors, the detection angle can be flexibly adjusted, layered scanning can be conducted on a building body, batch, comprehensive and accurate invisible damage detection is achieved, the device can adapt to a curved surface or a complex structure, the sampling frequency is high, and data integrity is guaranteed; according to the method, damage data are processed by utilizing a GLAU-Net model and a generalized dice loss function, a control group before and after repair is constructed, a repair result is evaluated by adopting multiple indexes such as an intensity attenuation rate and an elastic modulus, and the repair effect can be scientifically pre-judged.

Description

Technical Field

[0001] The present invention relates to the technical field of ancient building restoration, and particularly to a batch detection system and analysis method for invisible damage in ancient building restoration. Background Art

[0002] As historical and cultural heritages, ancient buildings have important cultural, artistic and scientific values. However, over time and under the influence of the natural environment, invisible damages often occur in ancient buildings, such as internal cracks, material deterioration, etc. Ancient building restoration is a relief measure taken when invisible damages appear in ancient buildings; and to determine how to repair, it is necessary to inspect the building structure, and accurate diagnosis requires taking pictures. For example, for the ancient building damage testing device and testing method with the Chinese patent publication number CN110568081A, the restorer manually three-dimensionally reconstructs the three-dimensional structures of various parts of the building structure and the relationship of defect positions to determine the repair type and repair plan. However, this detection method is inefficient and lacks precision, making it difficult to comprehensively and accurately obtain the invisible damage images before and after the restoration of ancient buildings, and lacking scientific means for evaluating the restoration effect, unable to effectively guarantee the quality of ancient building restoration. Therefore, there is an urgent need for an ancient building restoration invisible damage detection technology that can achieve batch, comprehensive and accurate detection and can scientifically evaluate the restoration effect. Summary of the Invention

[0003] The technical problem to be solved by the present invention is: to overcome the deficiencies of the prior art and provide a batch detection system and analysis method for invisible damage in ancient building restoration.

[0004] The technical solution adopted by the present invention is as follows: A batch detection system for invisible damage in ancient building restoration includes a detection device for batch detecting invisible damage to the building body before and after the restoration of ancient buildings. The detection device includes the following components: A driving component, including a manipulator, a connecting piece and a telescopic frame. The lifted manipulator is placed on one side of the building body. A vertical connecting piece is installed at the end of the manipulator, and a plurality of detection components arranged in a folded manner are connected to the connecting piece through the telescopic frame; the telescopic frame drives the detection components to perform telescopic movement relative to the connecting piece; The detection component, including a driving push rod and a sensor module. One end of the driving push rod is installed on the telescopic frame, and the other end of the driving push rod performs batch detection on different parts of the building body through the sensor module; the driving push rod drives the sensor module to move up and down relative to the telescopic frame; It further includes a server, and the sensor module transmits the detected invisible damage data of the building body to the server through a data line.

[0005] This technical solution realizes the flexible arrangement and operation of the detection components through mechanical automation means, obtains damage data using the sensor module, and conducts centralized management and analysis through the server, thereby improving the detection efficiency and accuracy and realizing the effect prediction before and after repair. Specifically, through its own joint movement and posture adjustment, the manipulator can accurately lift to a specified position on one side of the building body, ensuring that the detection components can accurately approach the area of the ancient building that needs to be detected; the connecting piece plays a role in connecting the manipulator and the telescopic frame; the telescopic frame is electrically controlled and drives the detection components to independently extend and approach for detection respectively. For some sunken parts, the telescopic frame shortens to avoid collision between the detection components and the building body; the driving push rod precisely controls the position of the sensor module in the vertical direction through its own telescopic movement, thereby realizing batch detection of different height parts of the building body; the sensor module senses various physical properties of the building body, such as strength attenuation rate, elastic modulus, crack development rate, mortise and tenon looseness, etc., and judges whether there are hidden damages in the building body through the changes of these properties. The server compares and analyzes the detection data at different times and different positions, and the server identifies the changes in hidden damages of the building body before and after repair and generates a detailed damage report.

[0006] In addition, according to the above-mentioned batch detection system for hidden damages in ancient building restoration proposed by the present invention, it also has the following additional technical features: According to an embodiment of the present invention, the sensor module includes a connecting part and a detection head. The connecting part is connected to the telescopic frame through a thread, and the detection head is detachably installed on the connecting part. The detection head is connected to the manipulator through a data line passing through the connecting part and the telescopic frame.

[0007] In this technical solution, due to the different materials, structures and damage types of ancient buildings, different types of detection heads are needed to meet different detection requirements; therefore, the detection head is detachably installed on the connecting part, which is convenient and fast to replace, without replacing the entire sensor module, reducing the maintenance cost and time.

[0008] According to an embodiment of the present invention, the server is respectively connected to a number of manipulators, and the server includes the following components: The repaired image processing module constructs a repair model and a damage function to preprocess the hidden damage data; The repaired image evaluation module constructs a control group before and after repair to comprehensively evaluate and predict the repair result.

[0009] In this technical solution, in the scenario of detecting invisible damages in the restoration of ancient buildings, multiple manipulators are distributed in different detection areas or perform different detection tasks. As the central control unit, the server can uniformly coordinate the actions, task assignments, and data interactions of these manipulators; through centralized management, it avoids problems such as resource waste, task conflicts, and data inconsistencies caused by the independent operation of each manipulator, and improves the operation efficiency of the entire detection system.

[0010] According to an embodiment of the present invention, the restoration model is the GLAU-Net model. By collecting data containing images of invisible damages and using a convolutional neural network to extract the features of invisible damages, full convolutional kernels with convolution kernels of 3, 5, and 7 are obtained, as well as dilated convolutions with convolution kernels of 3 and 5. Through complementary full convolutions and dilated convolutions, invisible damage features covering different global and local areas are obtained; an attention mechanism is introduced to enable the GLAU-Net model to selectively focus on features during the training process and pay attention to effective invisible damage images.

[0011] In this technical solution, to solve the problems of difficult continuous segmentation and over-segmentation in the restoration of ancient buildings, convolutions of different scales are used to obtain features. Specifically, dilated convolutions are introduced and combined with convolution kernels of different sizes to capture features of different sizes and enrich the images required for restoration. At the same time, using the selective focusing characteristic of the GLAU-Net model, during training, through operations such as squeezing, excitation, and scaling, the model focuses on the main channel features, suppresses the secondary channel features, discards redundant images, and accurately obtains key features.

[0012] According to an embodiment of the present invention, the loss function is the generalized dice loss function, and its calculation formula is as follows:

[0013] Where, |X∩Y| represents the intersection between set X and set Y, and |X|+|Y| respectively represent the number of pixel points in set X and set Y. Multiplying the numerator by 2 ensures that the threshold range is between 0 and 1 because the overlapping interval will be calculated one more time when the denominators are added; set X represents the pixels of the real area, and Y represents the pixels of the predicted area; Introduce w c to adjust the weight of each pixel in L dice That is:

[0014] N represents the total number of pixels in set X, M represents the total number of pixels in set X, i represents the i-th type of pixel, j represents the j-th pixel, represents the real value of the j-th pixel of the i-th type; represents the predicted value of the j-th pixel of the i-th type; w c is the weight, and c represents the c-th class; Introduce α to adjust L GDThe weight of each pixel in is: L focal (X, Y) = α*L GD +(1-α*L GD )*L GD

[0015] Among them: α>0 is a parameter.

[0016] According to an embodiment of the present invention, the evaluation indexes of the control group include strength attenuation rate, elastic modulus, crack development rate, and mortise and tenon looseness, wherein: Strength decay rate: the original flexural strength is 50MPa, and the decay rate is not less than 30%; Elastic modulus, using stress wave detection unit to obtain elastic modulus data; Crack development rate: intervention is required when the crack width of a wood component is greater than 0.2 mm and the depth reaches more than 80% of the component thickness; Looseness of mortise and tenon joints and joint gap of wooden components > 3mm require intervention.

[0017] In this technical solution, if the strength decays too much, the component cannot bear the normal load, resulting in structural damage; for example, when the beam bears the weight of the roof and external loads, if the bending strength decays by more than 30%, bending deformation or even fracture occurs, endangering the safety of the entire building. The elastic modulus reflects the ability of the material to resist elastic deformation. For example, a component with too low elastic modulus is prone to excessive deformation when subjected to force, affecting the overall stability of the building. When the width of the crack exceeds a certain limit, the effective bearing cross-section of the component will be reduced. At the same time, when the depth of the crack reaches a certain level, it will destroy the integrity of the component and reduce its bending and compression resistance. For example, a large crack on the column will make the column prone to splitting and damage when under pressure, resulting in building instability. When the node gap is too large, the friction and bite force between the tenon and the mortise are reduced, and the node is prone to relative displacement when subjected to force, resulting in structural deformation or even local damage. For example, the loose mortise and tenon in the bracket will make the bracket's force transmission function ineffective, affecting the seismic performance of the entire building.

[0018] In order to achieve the above-mentioned purpose, the present invention also provides a method for batch detection and analysis of invisible damage in ancient building restoration.

[0019] A batch detection and analysis method for invisible damage in ancient building restoration includes the following steps: S1. Use the lifting mechanism to lift the manipulator to one side of the building body; S2. By controlling the driving component, the telescopic frame is extended to a preset length along the axis of the connecting piece, driving the folded detection component to unfold to the surface of the building body; using the joint freedom of the manipulator, the detection angle of the sensor module is adjusted to ensure that it fits the building surface and adapts to curved surfaces or complex structures; S3. Drive the push rod to push the sensor module to move vertically along the surface of the building body at a constant speed to achieve layer-by-layer scanning; S4. Data synchronous transmission: The sensor module transmits the invisible damage data of the building body to the server in real time through the data line, and the sampling frequency ≥ 1 kHz.

[0020] This technical solution realizes the batch detection and processing of invisible damage in ancient building restoration: First, in step S1, the lifting mechanism is used to lift the manipulator to one side of the building body, so that the detection component can approach the target area; then, in step S2, by controlling the driving component, the telescopic frame is extended to a preset length (2 - 5 m) to deploy the detection component to the surface of the building body, and the detection angle of the sensor module (±30° pitch angle) is adjusted by means of the joint degrees of freedom of the manipulator to adapt to different structures, ensuring that the detection component can accurately contact the building body; then, in step S3, the push rod is driven to push the sensor module to move vertically along the surface of the building body at a constant speed to achieve layer-by-layer scanning, comprehensively obtaining the invisible damage images of different layers of the building body; finally, in step S4, the invisible damage data collected by the sensor module is transmitted to the server in real time through the data line at a sampling frequency ≥ 1 kHz for subsequent timely processing and analysis of the data, thereby completing the batch detection and processing of invisible damage in ancient buildings.

[0021] According to an embodiment of the present invention, in the above S1, the preset length is 2 - 5 m, and the detection angle of the sensor module is ±30° pitch angle.

[0022] According to an embodiment of the present invention, in the above S2, the layer spacing of the layer-by-layer scanning is 10 - 20 mm.

[0023] According to an embodiment of the present invention, in the above S3, the sensor module is a multi-sensor collaborative detection, including the following sensors: A stress wave detection unit that emits low-frequency stress waves of 1 - 10 kHz and calculates the depth and direction of internal cracks through the time difference of reflected waves; an infrared thermal imaging unit that detects the surface temperature distribution and identifies local temperature abnormalities ΔT > 0.5°C caused by internal damage; a three-dimensional laser scanning unit that acquires the surface topography at a resolution of 0.1 mm and generates point cloud data for calculating crack width and deformation.

[0024] Compared with the prior art, the present invention has the following beneficial effects: (1) Through the cooperation of the manipulator, the telescopic frame and multiple sensors, the detection angle can be flexibly adjusted, the building body can be scanned layer by layer, and batch, comprehensive and accurate detection of invisible damage can be realized, which can adapt to curved surfaces or complex structures, has a high sampling frequency, and ensures data integrity; (2) Use the GLAU-Net model and the generalized dice loss function to process the damage data, construct the control groups before and after repair, and evaluate the repair results using multiple indicators such as the strength attenuation rate and elastic modulus, which can scientifically predict the repair effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic structural diagram of the system of the present invention.

[0026] Figure 2 It is a schematic structural diagram of the detection system.

[0027] Figure 3 It is a schematic structural diagram of the sensor module.

[0028] Figure 4 It is a flow principle block diagram of the repair model.

[0029] Figure 5 It is a diagram of the control group of the invisible damage batch.

[0030] Figure 6 It is a diagram of the experimental group of the invisible damage batch.

[0031] In the figure: 1, building body; 2, manipulator; 3, connecting frame; 4, telescopic frame; 5, detection component; 6, driving push rod; 7, sensor module; 71, connecting part; 72, detection head. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0032] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0033] Embodiment 1 As Figures 1 to 4 shown, this embodiment provides a batch detection system for invisible damage in ancient building repair, including a detection device 5 for batch detecting invisible damage to the building body 1 before and after ancient building repair. The detection device 5 includes the following components: The driving component includes a manipulator 2, a connecting member, and a telescopic frame 4. The lifted manipulator 2 is placed on one side of the building body 1. A vertical connecting member is installed at the end of the manipulator 2, and a plurality of detection components arranged in a folded manner are connected to the connecting member through the telescopic frame 4; the telescopic frame 4 drives the detection components to perform telescopic movement relative to the connecting member; The detection component includes a driving push rod 6 and a sensor module 7. One end of the driving push rod 6 is installed on the telescopic frame 4, and the other end of the driving push rod 6 is batch-detected with different parts of the building body 1 through the sensor module 7; the driving push rod 6 drives the sensor module 7 to move up and down relative to the telescopic frame 4; It further includes a server. The sensor module 7 transmits the detected invisible damage data of the building body 1 to the server through a data cable.

[0034] As Figures 1 to 3 shown, this technical solution realizes the flexible arrangement and operation of the detection component through mechanical automation means, uses the sensor module 7 to obtain damage data, and conducts centralized management and analysis through the server, thereby improving the detection efficiency and accuracy and realizing the effect prediction before and after repair. Specifically, the manipulator 2 can be accurately lifted to a specified position on one side of the building body 1 through its own joint movement and posture adjustment, ensuring that the detection component can accurately approach the area of the ancient building to be detected; the connecting piece plays a role in connecting the manipulator 2 and the telescopic frame 4; the telescopic frame 4 is electrically controlled and drives the detection component to independently extend and approach for detection respectively. For some sunken parts, the telescopic frame 4 shortens to avoid collision between the detection component and the building body 1; the driving push rod 6 precisely controls the position of the sensor module 7 in the vertical direction through its own telescopic movement, thereby realizing batch detection of different height parts of the building body 1; the sensor module 7 senses various physical properties of the building body 1, such as strength attenuation rate, elastic modulus, crack development rate, mortise and tenon looseness, etc., and judges whether there is invisible damage to the building body 1 through the changes of these properties. The server compares and analyzes the detection data at different times and different positions, and the server identifies the changes in the invisible damage of the building body 1 before and after repair and generates a detailed damage report.

[0035] In addition, the invisible damage batch detection system for ancient building repair proposed above according to the present invention further has the following additional technical features: As Figure 3 shown, the sensor module 7 includes a connecting part 71 and a detection head 72. The connecting part 71 is connected to the telescopic frame 4 through a thread, and the detection head 72 is detachably installed on the connecting part 71. The detection head 72 is connected to the manipulator 2 through a data cable passing through the connecting part 71 and the telescopic frame 4.

[0036] In this technical solution, due to the different materials, structures and damage types of ancient buildings, different types of detection heads 72 are required to meet different detection needs; therefore, the detection head 72 is detachably installed on the connecting part 71, which is convenient and quick to replace, and there is no need to replace the entire sensor module 7, reducing the maintenance cost and time.

[0037] According to an embodiment of the present invention, the server is respectively connected to a number of manipulators 2, and the server includes the following components: A repaired image processing module that constructs a repair model and a damage function to preprocess the invisible damage data; A repaired image evaluation module that constructs a control group before and after repair to comprehensively evaluate and predict the repair results.

[0038] In this technical solution, in the scenario of detecting invisible damage in ancient building repair, multiple manipulators 2 are distributed in different detection areas or perform different detection tasks. As the central control unit, the server can uniformly coordinate the actions, task assignments, and data interactions of these manipulators 2; through centralized management, problems such as resource waste, task conflicts, and data inconsistencies caused by the independent operation of each manipulator 2 are avoided, and the operation efficiency of the entire detection system is improved.

[0039] As Figure 4 shown, the repair model is the GLAU-Net model. By collecting data containing invisible damage images and using a convolutional neural network to extract the features of invisible damage, fully convolutional kernels with convolution kernels of 3, 5, and 7 are obtained, as well as dilated convolutions with convolution kernels of 3 and 5. The complementary fully convolutional and dilated convolutions are used to obtain invisible damage features covering different global and local areas; an attention mechanism is introduced to enable the GLAU-Net model to selectively focus on features during training and pay attention to effective invisible damage images.

[0040] In this technical solution, to solve the problems of difficult continuous segmentation and over-segmentation in some parts of ancient building repair, convolutions of different scales are used to obtain features. Specifically, dilated convolutions are introduced and combined with convolution kernels of different sizes to capture features of different sizes and enrich the images required for repair. At the same time, using the selective focusing characteristic of the GLAU-Net model, during training, through operations such as squeezing, excitation, and scaling, the model focuses on the main channel features, suppresses the secondary channel features, discards redundant images, and accurately obtains key features.

[0041] According to an embodiment of the present invention, the damage function is a generalized dice loss function, and its calculation formula is as follows:

[0042] where, |X∩Y| represents the intersection between set X and set Y, and |X|+|Y| respectively represent the number of pixel points in set X and set Y. The numerator is multiplied by 2 to ensure that the threshold range is between 0 and 1 because the overlapping interval will be calculated one more time when the denominators are added; set X represents the pixels of the real area, and Y represents the pixels of the predicted area; Introduce w c to adjust the weight of each pixel in L dice , that is:

[0043] N represents the total number of pixels in set X, M represents the total number of pixels in set X, i represents the pixel of the i-th class, j represents the j-th pixel, Represents the true value of the jth pixel of the i-th class; represents the predicted value of the jth pixel of the i-th class; w c is the weight, c represents the cth class; Introduce α to adjust L GD The weight of each pixel in is: L focal (X, Y) = α*L GD +(1-α*L GD )*L GD

[0044] Among them: α>0 is a parameter.

[0045] According to one embodiment of the present invention, the evaluation indexes of the control group include strength attenuation rate, elastic modulus, crack development rate, and mortise and tenon looseness, wherein: Strength decay rate: the original flexural strength is 50MPa, and the decay rate is not less than 30%; Elastic modulus, using stress wave detection unit to obtain elastic modulus data; Crack development rate: intervention is required when the crack width of a wood component is greater than 0.2 mm and the depth reaches more than 80% of the component thickness; Looseness of mortise and tenon joints and joint gap of wooden components > 3mm require intervention.

[0046] In this technical solution, if the strength decays too much, the component cannot bear the normal load, resulting in structural damage; for example, when the beam bears the weight of the roof and external loads, if the bending strength decays by more than 30%, bending deformation or even fracture occurs, endangering the safety of the entire building. The elastic modulus reflects the ability of the material to resist elastic deformation. For example, a component with too low elastic modulus is prone to excessive deformation when subjected to force, affecting the overall stability of the building. When the width of the crack exceeds a certain limit, the effective bearing cross-section of the component will be reduced. At the same time, when the depth of the crack reaches a certain level, it will destroy the integrity of the component and reduce its bending and compression resistance. For example, a large crack on the column will make the column prone to splitting and damage when under pressure, resulting in building instability. When the node gap is too large, the friction and bite force between the tenon and the mortise are reduced, and the node is prone to relative displacement when subjected to force, resulting in structural deformation or even local damage. For example, the loose mortise and tenon in the bracket will make the bracket's force transmission function ineffective, affecting the seismic performance of the entire building.

[0047] Example 2 Based on Example 1, Figures 1 to 6 As shown, this embodiment provides a method for batch detection and analysis of invisible damage in ancient building restoration, comprising the following steps: S1. Use the lifting mechanism to lift the manipulator 2 to one side of the building body 1; S2. By controlling the driving component, extend the telescopic frame 4 along the axial direction of the connecting piece to a preset length, driving the folded and arranged detection component to unfold onto the surface of the building body 1; utilize the joint degrees of freedom of the manipulator 2 to adjust the detection angle of the sensor module 7 to ensure it fits the building surface and adapts to curved surfaces or complex structures; S3. Drive the push rod 6 to push the sensor module 7 to move in the vertical direction of the surface of the building body 1 at a constant speed to achieve layered scanning; S4. Data synchronous transmission: The sensor module 7 transmits the invisible damage data of the building body 1 to the server in real time through the data line, and the sampling frequency ≥ 1 kHz.

[0048] This technical solution realizes the batch detection and processing of invisible damage in ancient building restoration: First, in step S1, the lifting mechanism is used to lift the manipulator 2 to one side of the building body 1, enabling the detection component to approach the target area; then, in step S2, by controlling the driving component, the telescopic frame 4 is extended to a preset length (2 - 5 m) to unfold the detection component onto the surface of the building body 1, and the detection angle of the sensor module 7 (±30° pitch angle) is adjusted with the help of the joint degrees of freedom of the manipulator 2 to adapt to different structures, ensuring that the detection component can accurately contact the building body 1; then, in step S3, the push rod 6 is driven to push the sensor module 7 to move in the vertical direction of the surface of the building body 1 at a constant speed to achieve layered scanning, comprehensively obtaining the invisible damage images of different layers of the building body 1; finally, in step S4, the invisible damage data collected by the sensor module 7 is transmitted to the server in real time through the data line at a sampling frequency ≥ 1 kHz for subsequent timely data processing and analysis, thus completing the batch detection and processing of invisible damage in ancient buildings.

[0049] According to an embodiment of the present invention, in S1, the preset length is 2 - 5 m, and the detection angle of the sensor module 7 is ±30° pitch angle.

[0050] According to an embodiment of the present invention, in S2, the layer spacing of the layered scanning is 10 - 20 mm.

[0051] According to an embodiment of the present invention, in S3, the sensor module 7 is for multi-sensor collaborative detection, including the following sensors: A stress wave detection unit that emits low-frequency stress waves of 1 - 10 kHz and calculates the depth and direction of internal cracks through the time difference of reflected waves; an infrared thermal imaging unit that detects the surface temperature distribution and identifies local temperature differences ΔT > 0.5°C caused by internal damage; a three-dimensional laser scanning unit that acquires the surface topography at a resolution of 0.1 mm and generates point cloud data for calculating crack width and deformation.

[0052] Combined with Table 1 below, and asFigure 5 and Figure 6 As shown, compared with the control group model, the model proposed in the present invention has significantly improved in most evaluation indicators. This indicates that the relevant modules in the improved model, relying on a more refined feature weighting mechanism, further improve the segmentation accuracy of the model. Compared with the control group model, the advantages of this model are obvious: First, its performance is more excellent and it can perform better in tasks such as the segmentation of invisible damages in ancient building restoration; Second, it effectively alleviates the problems of non-segmentation, over-segmentation or under-segmentation existing in the control group model, making the segmentation results more accurate; Third, the segmentation process has better stability and robustness and can adapt to different scenarios. In addition, the segmentation performance of this model has been significantly improved, which alleviates the over-segmentation and discontinuous segmentation problems of the control group to a certain extent. However, there is still room for improvement in accuracy. For example, when α = 0.1, almost all indicators of the model reach or approach the optimal values, indicating the direction for subsequent optimization.

[0053] Table 1

[0054] Although the present invention has been described in detail by referring to the accompanying drawings and in combination with the preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, those of ordinary skill in the art can make various equivalent modifications or substitutions to the embodiments of the present invention, and these modifications or substitutions should all be within the scope of the present invention. / Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. An invisible damage batch detection system for ancient building restoration, characterized in that It includes a detection device for batch detecting invisible damages of the building body (1) before and after the restoration of ancient buildings. The detection device includes the following components: A driving component, including a manipulator (2), a connecting piece and a telescopic frame (4). The lifted manipulator (2) is moved to one side of the building body (1). A vertical connecting piece is installed at the end of the manipulator (2). A number of detection components (5) arranged in a folded manner are connected to the connecting piece through the telescopic frame (4); the telescopic frame (4) drives the detection components (5) to perform telescopic movement relative to the connecting piece; The detection component (5), including a driving push rod (6) and a sensor module (7). One end of the driving push rod (6) is installed on the telescopic frame (4), and the other end of the driving push rod (6) performs batch detection on different parts of the building body (1) through the sensor module (7); the driving push rod (6) drives the sensor module (7) to move up and down relative to the telescopic frame (4); It also includes a server. The sensor module (7) transmits the detected invisible damage data of the building body (1) to the server through a data line.

2. The invisible damage batch detection system for ancient building restoration according to claim 1, characterized in that, The sensor module (7) includes a connecting part (71) and a detection head (72). The connecting part (71) is connected to the telescopic frame (4) by a thread. The detection head (72) is detachably installed on the connecting part (71). The detection head (72) is connected to the manipulator (2) through a data line passing through the connecting part (71) and the telescopic frame (4).

3. The invisible damage batch detection system for ancient building restoration according to claim 1, characterized in that, The server is respectively connected to a number of manipulators (2). The server includes the following components: A restoration image processing module, which constructs a restoration model and a damage function to preprocess the invisible damage data; A restoration image evaluation module, which constructs a control group before and after restoration to comprehensively evaluate and predict the restoration result.

4. The invisible damage batch detection system for ancient building restoration according to claim 3, wherein The restoration model is the GLAU-Net model. By collecting data containing invisible damage images and using a convolutional neural network to extract the features of invisible damages, full convolutional kernels with convolution kernels of 3, 5, and 7 are obtained, as well as dilated convolutions with convolution kernels of 3 and 5. Through complementary full convolution and dilated convolution, invisible damage features covering different global and local areas are obtained; An attention mechanism is introduced to enable the GLΔU-Net model to selectively focus on features during the training process and pay attention to effective invisible damage images.

5. The invisible damage batch detection system for ancient building restoration according to claim 3, wherein The damage function is the generalized dice loss function, and its calculation formula is as follows: Where, |X∩Y| represents the intersection between set X and set Y, and |X|+|Y| respectively represent the number of pixel points in set X and set Y. The numerator is multiplied by 2 to ensure that the threshold range is between 0 and 1, because the overlapping interval will be calculated one more time when the denominator is added; set X represents the pixels of the real area, and Y represents the pixels of the predicted area; Introduce w c to adjust the weights of dice each pixel in L, i.e.: N represents the total number of pixels in set X, M represents the total number of pixels in set X, i represents the pixels of the i-th class, and j represents the j-th pixel. represents the true value of the j-th pixel of the i-th class; represents the predicted value of the j-th pixel of the i-th class; w c is the weight, and c represents the c-th class; Introduce α to adjust the weight of each pixel in L, i.e.: GD ​ L focal (X, Y) = α * L GD + (1 - α * L GD ) * L GD Where: α>0 is a parameter.

6. The invisible damage batch detection system for ancient building restoration according to claim 5, wherein, The evaluation indexes of the control group include the strength attenuation rate, elastic modulus, crack development rate, and mortise and tenon looseness. Among them: Strength attenuation rate, the original flexural strength is 50MPa, and the attenuation rate is not less than 30%; Elastic modulus, the elastic modulus data is obtained by using a stress wave detection unit; Crack development rate, for wooden components, when the crack width > 0.2mm and the depth reaches more than 80% of the component thickness, intervention is required; Mortise and tenon looseness, for wooden components, when the node gap > 3mm, intervention is required.

7. A method for batch detection and analysis of invisible damages in ancient building restoration, which uses the batch detection system for invisible damages in ancient building restoration as described in any one of claims 1-6, characterized in that It includes the following steps: S1. Use the lifting mechanism to lift the manipulator (2) to one side of the building body (1); S2. By controlling the driving component, extend the telescopic frame (4) along the axial direction of the connecting piece to a preset length, and drive the folded and arranged detection component (5) to unfold to the surface of the building body (1); utilize the joint degrees of freedom of the manipulator (2) to adjust the detection angle of the sensor module (7) to ensure it fits the building surface and adapts to curved surfaces or complex structures; S3. Drive the push rod (6) to push the sensor module (7) to move vertically along the surface of the building body (1) at a constant speed to achieve layered scanning; S4. Data synchronous transmission: The sensor module (7) transmits the invisible damage data of the building body (1) to the server in real time through the data line, and the sampling frequency ≥ 1 kHz.

8. The method for batch detection and analysis of invisible damages in ancient building restoration according to claim 7, characterized in that In the above S1, the preset length is 2 - 5 m, and the detection angle of the sensor module (7) is ±30° pitch angle.

9. The method for batch detection and analysis of invisible damages in the restoration of ancient buildings according to claim 7, wherein, In the above S2, the layer spacing of the layered scanning is 10 - 20 mm.

10. The method for batch detection and analysis of invisible damages in ancient building restoration according to claim 7, characterized in that, In the above S3, the sensor module (7) is for multi-sensor collaborative detection, including the following sensors: The stress wave detection unit emits low-frequency stress waves of 1 - 10 kHz, and calculates the depth and trend of internal cracks through the time difference of reflected waves; The infrared thermal imaging unit detects the surface temperature distribution and identifies local temperature anomalies ΔT > 0.5°C caused by internal damage; the three-dimensional laser scanning unit acquires the surface topography at a resolution of 0.1 mm and generates point cloud data for calculating crack widths and deformation amounts.

Images