Layer treatment monitoring system for roller compacted concrete gravity dam

Abstract

The invention belongs to the technical field of concrete engineering monitoring, and particularly discloses a roller compacted concrete gravity dam layer processing monitoring system, which comprises a data acquisition module used for acquiring surface image information and internal section image information; the data processing module is used for processing the surface image information and the internal section image information; the three-dimensional construction module is used for constructing a gravity dam virtual scene, a virtual roller compacted concrete layer and a 3D model; the coordinate conversion module is used for converting the three-dimensional coordinates into GPS coordinates; and the data analysis module is used for regulating and controlling the concrete unloading point location and the leveling strategy in real time based on the parameter information of the previous roller compacted concrete layer. Problems are foreseen in advance, forward guidance is provided for the construction technology, the quality problems in the construction process are found and solved in time, the reworking and repairing cost caused by unqualified quality is avoided, and the construction cost of the whole gravity dam is reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of concrete engineering monitoring, and in particular relates to a layer processing monitoring system for a roller compacted concrete gravity dam. Background Art

[0002] The construction of reservoir projects is an effective engineering measure to solve the production and living water needs of some main urban areas and surrounding towns. It provides a reliable guarantee for water use for social and economic development, effectively solves the contradiction of medium- and long-term water shortage in county towns, and promotes the development of local social economy. In the current construction of reservoir facilities, roller compacted concrete has been widely used due to its advantages such as fast construction speed, low cost and good mechanical properties. However, the quality of the roller compacted concrete layer is directly related to the safety, stability and durability of the entire engineering structure.

[0003] Traditional RCC layer detection methods, such as coring, can obtain actual samples from the interior of the concrete, but this method is not only time-consuming and labor-intensive, with low detection efficiency, but also causes irreparable damage to the concrete structure, affecting the integrity and safety of the structure; when using the buried method to detect the thickness of the RCC layer and the internal structure of the concrete layer, equipment such as rollers and vibrators may damage the buried sensor equipment. In recent years, with the rapid development of computer technology and information technology, detection methods based on image recognition and data analysis have gradually emerged. However, these methods still have many problems in practical applications, such as the detection accuracy is difficult to meet the requirements of complex projects, the real-time performance is poor, the quality status of the concrete layer cannot be intuitively fed back, and the construction cannot be provided with positive guidance before the RCC layer is formed. Summary of the invention

[0004] The purpose of the present invention is to overcome the defects in the prior art and provide a roller compacted concrete gravity dam surface treatment monitoring system.

[0005] The present invention provides a roller compacted concrete gravity dam surface treatment monitoring system, comprising: A data acquisition module, used to acquire surface image information and internal cross-sectional image information of the RCC layer; A data processing module, connected to the data acquisition module, for processing the surface image information and the internal cross-sectional image information so that the surface image information and the internal cross-sectional image information reflect the real RCC layer information; A three-dimensional construction module is used to construct a virtual scene of a gravity dam, and configure a virtual roller compacted concrete layer based on the design requirements based on the virtual scene of the gravity dam. The three-dimensional construction module is also connected to the data processing module, and the roller compacted concrete layer is modeled in layers based on the real roller compacted concrete layer information, and the 3D model of each roller compacted concrete layer is placed in the corresponding virtual scene of the gravity dam, and a dynamic stress belt is set based on the preset rolling trajectory of each group of rolling equipment, so as to simulate the roller compaction equipment in the virtual scene of the gravity dam to pressurize each roller compacted concrete layer; A coordinate conversion module is used to convert and match the three-dimensional coordinates in the gravity dam virtual scene with the measured GPS coordinates of the gravity dam; a data analysis module, connected to the three-dimensional construction module and the data processing module respectively, extracting parameters of the processed surface image information and the internal cross-sectional image information, obtaining parameter information of each RCC layer, regulating the 3D model of each RCC layer based on the parameter information, regulating the concrete unloading point and the liquidation strategy in real time based on the parameter information of the previous RCC layer, and simulating the next RCC layer in the gravity dam virtual scene based on the concrete unloading point and the liquidation strategy; The judgment module is used to compare the latter RCC layer generated by simulation with the virtual RCC layer configured based on design requirements, output a comparison value, and fine-tune the concrete unloading point and closing strategy based on the comparison value.

[0006] A further solution is that the data acquisition module includes a high-definition camera unit, an X-ray and CT scanning unit; The high-definition camera unit is equipped with a plurality of high-definition cameras for acquiring surface image information of the RCC layer from different angles and positions; The X-ray and CT scanning unit uses X-ray and CT scanning equipment to perform internal scanning on the RCC layer to obtain internal cross-sectional image information of the RCC layer.

[0007] A further solution is that the data processing module includes a geometric correction unit, a radiation correction unit, an image enhancement unit and a data enhancement unit; The geometric correction unit is used to perform geometric correction on the surface image information and the internal cross-sectional image information, eliminate the image deformation caused by the shooting angle and lens distortion, and restore the shape and position of the object in the image to the real state; The radiation correction unit is used to perform radiation correction on the image and adjust the brightness and contrast of the image by counting the brightness distribution of the surface image information; The image enhancement unit processes the image using histogram equalization to highlight edge and texture features in the image; The data enhancement unit scales, rotates, and flips the image according to a preset ratio to generate a large amount of new image data.

[0008] A further solution is that the data analysis module includes a thickness analysis unit, a pore and crack analysis unit, an aggregate distribution analysis unit, and a color and texture analysis unit; The thickness analysis unit is used to obtain the thickness distribution information of the actual rolled concrete layer, and compare and analyze it with the designed thickness to determine whether the thickness of the concrete layer meets the design requirements; The pore and crack analysis unit is used to count the number, area, and length parameters of pores and cracks, and determine whether the severity of the pores and cracks exceeds the allowable range; The aggregate distribution analysis unit is used to identify the location, type and distribution density of aggregates and quantitatively evaluate the uniformity of aggregate distribution; The color and texture analysis unit is used to compare the color and texture with preset values, determine whether the color and texture of the concrete layer surface meet the design requirements and quality standards, and then infer the construction process and quality status of the concrete layer.

[0009] A further solution is that the thickness distribution information, the number, area, and length parameters of pores and cracks are determined by a first recognition model, and the construction process of the first recognition model is: The cross-sectional image information processed by the data processing module is annotated with thickness, pores and cracks by professionals. After annotation, the cross-sectional image information is input into the first neural network unit for iterative training, and the thickness distribution information, number, area and length parameters of pores and cracks of the current rolled concrete layer are output to obtain the first recognition model.

[0010] The cross-sectional image information includes images at different construction stages and in different quality conditions.

[0011] A further solution is that the aggregate distribution, color and texture are determined by a second recognition model, and the construction process of the second recognition model is: The surface image information processed by the data processing module is annotated by professionals for aggregate distribution, color and texture. After annotation, the surface image information is input into the second neural network unit for iterative training, and the aggregate distribution, color and texture information of the current roller compacted concrete layer is output to obtain the second recognition model.

[0012] A further solution is that the data analysis module also includes a point determination unit and a position closing strategy unit; The point determination unit is connected to the thickness analysis unit and the coordinate conversion module respectively, determines the three-dimensional coordinates of the unloading point on the 3D model based on the thickness difference of different positions of the previous rolled concrete layer, and converts the three-dimensional coordinates into GPS coordinates through the coordinate conversion module; The closing strategy unit is connected to the color and texture analysis unit respectively, and determines the closing strategy based on the color and texture of the surface of the concrete layer; The closing strategy is the working strategy of the rolling equipment.

[0013] A further solution is that the process of obtaining the three-dimensional coordinates of the unloading point is: The three-dimensional coordinates of the measurement point are set based on the thickness distribution information as (x i ,y i , z i ), corresponding to the thickness h i , i=1,2,…,n, n is the number of measurement points; For the measurement point (x j ,y j , z j ), other measurement points in its neighborhood (x k ,y k , z k ) and the measurement point (x j ,y j , z j ) is: ; Wherein, j and k are both integers less than or equal to n; Calculate the thickness change rate within the neighborhood: ; in, is the measuring point (x k ,y k , z k ) thickness, is the measuring point (x j ,y j , z j ) thickness; The thickness change rate represents the thickness change per unit distance. The average thickness change rate of all thickness change rates in the neighborhood is obtained to obtain the average thickness change rate of the point. ; Based on the design thickness H of each RCC layer, for areas where the current thickness is less than the target thickness, determine the three-dimensional coordinates of the unloading point: the measurement point (x j ,y j , z j ) The added thickness is: , then the three-dimensional space distance offset is ; By analyzing the thickness of each point in the neighborhood and the measured point (x j ,y j , z j) thickness difference, determine the offset direction vector (a, b, c) in three-dimensional space, and normalize the offset direction vector (a, b, c) to obtain a unit vector ( ), then the three-dimensional coordinates of the next unloading point of the 3D model are (x new ,y new , z new ),in: ; The coordinate conversion module converts the three-dimensional coordinates (x new ,y new , z new ) is converted into GPS coordinates to obtain the actual location of the next unloading point.

[0014] A further solution is that the process of determining the closing strategy is: Convert the surface image information processed by the data processing module into the CIELAB color space and set the color deviation threshold ,when When the vibration frequency of the rolling equipment is increased by 5Hz and the amplitude is increased by 0.1mm for the light-colored area, the vibration frequency is reduced by 3Hz and the amplitude is reduced by 0.05mm for the dark-colored area. : For light-colored areas, the vibration frequency is increased by 10Hz, the amplitude is increased by 0.2mm, and the number of rolling times is increased by 2 times; for dark-colored areas, the vibration frequency is reduced by 5Hz, the amplitude is reduced by 0.1mm, and the number of rolling times is reduced by 1 time; Calculating a texture clarity index and a direction consistency index based on the texture information output by the second recognition model; Set the texture definition threshold 15, direction standard deviation threshold is 10°; When the texture clarity : Increase the compaction power of the rolling equipment by 20% or replace it with a heavier rolling wheel to enhance the compaction effect and make the texture clearer; When the standard deviation of the grain direction : Re-plan the driving route of the rolling equipment to ensure that the deviation between the rolling direction and the design direction each time is less than 5°, and at the same time increase the overlapping width of the rolling wheels.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention discloses a comprehensive monitoring and control system. The data acquisition module collects surface and internal cross-sectional image information in multiple dimensions. The data processing module processes it to ensure that the information is true and accurate, providing a reliable basis for subsequent analysis. The three-dimensional construction module creates a virtual scene of the gravity dam and models it in layers, simulating the pressure of the rolling equipment, so that construction personnel can intuitively understand the structure and construction effect of the gravity dam and foresee problems in advance. The data analysis module adjusts the unloading point and liquidation strategy of the back layer in real time according to the information of the front layer, provides positive guidance for the construction process, avoids blind construction and repeated operations that may occur in traditional construction, and improves construction efficiency. By timely discovering and solving quality problems in the construction process, rework and repair costs caused by unqualified quality are avoided, and the cost of the entire gravity dam construction is reduced.

[0016] The present invention uses a high-definition camera unit and an X-ray and CT scanning unit to obtain surface image information and internal cross-sectional image information of the RCC layer from different angles and positions. The high-definition camera can fully capture the surface of the concrete layer, while the X-ray and CT scanning can deeply understand the internal structure, realizing all-round monitoring of the RCC layer from the outside to the inside, providing a rich and accurate data basis for subsequent analysis.

[0017] The geometric correction unit in the data processing module of the present invention can eliminate image deformation caused by shooting angle and lens distortion, the radiation correction unit adjusts the brightness and contrast of the image, the image enhancement unit highlights the edge and texture features in the image, and the data enhancement unit generates a large amount of new image data. These processing steps enable the collected image information to more accurately reflect the real RCC layer information, thereby improving the quality and usability of the data.

[0018] The data analysis module of the present invention extracts parameters from the processed image information to obtain parameter information of each RCC layer, such as thickness, pores and cracks, aggregate distribution, color and texture, etc. Based on these parameter information, the 3D model of each RCC layer can be adjusted to make it more in line with the actual situation, and provide accurate data support for subsequent construction decisions.

[0019] The present invention can adjust the concrete unloading point and closing strategy in real time by analyzing the parameter information of the previous layer of rolled concrete. For example, the point determination unit determines the three-dimensional coordinates of the unloading point according to the thickness difference and converts it into GPS coordinates to ensure the accuracy of the unloading position; the closing strategy unit determines the working strategy of the rolling equipment according to the color and texture of the concrete layer surface, such as adjusting the vibration frequency, amplitude, rolling times, etc., so as to ensure the construction quality of each layer of concrete.

[0020] By building the first recognition model and the second recognition model, the thickness distribution, pore and crack parameters of the RCC layer, as well as aggregate distribution, color and texture information are determined respectively. These models are iteratively trained using a large number of annotated images, and can accurately and quickly identify relevant parameters, which improves the efficiency and accuracy of quality inspection and helps to detect potential quality problems in a timely manner.

[0021] In the construction of gravity dams, the accuracy of concrete unloading points has a significant impact on the quality of gravity dams. The present invention can accurately adjust the unloading points according to actual needs by analyzing thickness differences, calculating the neighborhood thickness change rate, and determining the offset and direction vector. For example, when the thickness of a certain area is less than the design value, the location where additional concrete needs to be added can be accurately found to avoid local over-thinness or over-thickness, ensure uniform thickness of various parts of the gravity dam, effectively prevent stress concentration problems caused by uneven thickness, and enhance the structural stability of the gravity dam. Moreover, after the coordinates are converted into GPS coordinates of the gravity dam, construction personnel can use the global positioning system to quickly locate the unloading point at the actual construction site, improve construction efficiency, reduce errors, and ensure that the construction is carried out efficiently and accurately according to design requirements.

[0022] The present invention determines the closing strategy according to the color and texture of the concrete surface, and realizes the fine control of the working parameters of the rolling equipment. The processed surface image information is converted to the CIELAB color space of the gravity dam, and the vibration frequency, amplitude and rolling times are adjusted according to the color deviation threshold, which can improve the compaction effect and surface quality of the concrete. Different adjustment strategies are adopted for light-colored areas and dark-colored areas, respectively, so that the concrete can be compacted more evenly and quality problems caused by insufficient or excessive compaction can be avoided. The clarity and direction consistency indexes are calculated based on the texture information, and the rolling equipment is adjusted accordingly, which can further ensure the construction quality of the gravity dam. When the texture clarity is not enough, increasing the compaction power or replacing the rolling wheel can enhance the compaction effect and improve the density and uniformity of the concrete; when the standard deviation of the texture direction does not meet the standard, re-planning the driving route and increasing the overlapping width of the rolling wheel can make the rolling direction more standardized, ensure that the various parts of the gravity dam are evenly stressed, and effectively improve the overall quality and stability of the gravity dam. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The following drawings are only used to illustrate and explain the present invention, and are not used to limit the scope of the present invention, wherein: Figure 1 : Principle block diagram of the monitoring system of the present invention; Figure 2 : Schematic diagram of the virtual gravity dam of the present invention; In the figure: 1. Virtual scene of gravity dam; 2. Virtual rolled concrete layer; 3. 3D model; 4. Data acquisition module; 5. Data processing module; 6. Three-dimensional construction module; 7. Coordinate conversion module; 8. Data analysis module; 9. Judgment module; 10. High-definition camera unit; 11. X-ray and CT scanning unit; 12. Geometry correction unit; 13. Radiation correction unit; 14. Image enhancement unit; 15. Data enhancement unit; 16. First recognition model; 17. Second recognition model; 18. Thickness analysis unit; 19. Pores and cracks analysis unit; 20. Aggregate distribution analysis unit; 21. Color and texture analysis unit; 22. Point determination unit; 23. Closing strategy unit. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution, design method and advantages of the present invention clearer, the present invention is further described in detail by specific embodiments in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0025] like Figure 1 As shown, the present invention provides a roller compacted concrete gravity dam surface treatment monitoring system, comprising: Data acquisition module 4, used to acquire surface image information and internal cross-sectional image information of the RCC layer; A data processing module 5, connected to the data acquisition module 4, for processing the surface image information and the internal cross-sectional image information so that the surface image information and the internal cross-sectional image information reflect the real RCC layer information; The three-dimensional construction module 6 is used to construct a gravity dam virtual scene 1, such as Figure 2As shown, a virtual rolled concrete layer 2 is configured based on the design requirements based on the gravity dam virtual scene 1, and the three-dimensional construction module 6 is also connected to the data processing module 5. The rolled concrete layer is hierarchically modeled based on the real rolled concrete layer information, and the 3D model 3 of each rolled concrete layer is placed in the corresponding gravity dam virtual scene 1. The dynamic force belt is set based on the preset rolling track of each group of rolling equipment to simulate the rolling equipment in the gravity dam virtual scene 1. The rolling track is divided into several continuous force areas according to the preset rolling track and the wheel width of the rolling equipment. Each force area can be regarded as the range covered by the rolling equipment in a rolling process. The properties of the dynamic force belt are determined for each force area, including the force magnitude, direction and action time, etc. The force magnitude can be determined by mechanical calculation according to the parameters of the rolling equipment and the characteristics of the concrete. For example, according to the weight and vibration parameters of the rolling wheel, combined with the compressive strength of the concrete, the pressure value applied to the concrete layer at different rolling stages is calculated. The force direction is perpendicular to the surface of the RCC layer, and the action time is related to the driving speed of the rolling equipment and the length of the rolling track, which is determined by calculating the time the equipment stays in each force area. The force analysis is as follows: The rolling equipment's own gravity G=Mg, g is the gravitational acceleration, and the gravity acts perpendicularly on the surface of the RCC layer and generates uniform static pressure on the contact surface. , S is the contact area between the rolling wheel and the rolled concrete layer.

[0026] A coordinate conversion module 7 is used to convert and match the three-dimensional coordinates in the gravity dam virtual scene 1 with the measured GPS coordinates of the gravity dam; The data analysis module 8 is connected to the three-dimensional construction module 6 and the data processing module 5 respectively, performs parameter extraction on the processed surface image information and the internal cross-sectional image information, obtains parameter information of each RCC layer, adjusts the 3D model 3 of each RCC layer based on the parameter information, and adjusts the concrete unloading point and the liquidation strategy in real time based on the parameter information of the previous RCC layer, and simulates the next RCC layer in the gravity dam virtual scene 1 based on the concrete unloading point and the liquidation strategy; The judgment module 9 is used to compare the simulated and generated next layer of RCC with the virtual RCC layer 2 configured based on the design requirements, output a comparison value, and fine-tune the concrete unloading point and the closing strategy based on the comparison value. Specifically, if the comparison value shows that the thickness of the simulated and generated RCC layer is greater than the virtual RCC layer 2 in some areas, then during the subsequent unloading, the unloading point should be appropriately moved away from these areas and offset to the area with insufficient thickness. For example, if it is calculated that the thickness of a certain area is 5cm thicker than the design value, the unloading point can be offset by 2m in the direction of the thinner thickness around the area. For the closing strategy, in areas with larger thickness, the number of operations of the closing equipment can be appropriately reduced or its operation intensity can be reduced to avoid further increase in thickness due to excessive compaction; in areas with insufficient thickness, the number of operations and intensity of the closing equipment can be increased to ensure that the concrete is evenly distributed to reach the designed thickness.

[0027] In the above, the data acquisition module 4 includes a high-definition camera unit 10 and an X-ray and CT scanning unit 11; wherein the high-definition camera unit 10 is equipped with multiple high-definition cameras for obtaining surface image information of the rolled concrete layer from different angles and positions; the X-ray and CT scanning unit 11 uses X-ray and CT scanning equipment to perform internal scanning on the rolled concrete layer to obtain internal cross-sectional image information of the rolled concrete layer.

[0028] In order to improve the monitoring accuracy, in this embodiment, the data processing module 5 includes a geometric correction unit 12, a radiation correction unit 13, an image enhancement unit 14 and a data enhancement unit 15; the geometric correction unit 12 is used to perform geometric correction on the surface image information and the internal cross-sectional image information, eliminate the image deformation caused by the shooting angle and the lens distortion, and restore the shape and position of the object in the image to the real state; the radiation correction unit 13 is used to perform radiation correction on the image by counting the brightness distribution of the surface image information, and adjust the brightness and contrast of the image; the image enhancement unit 14 uses histogram equalization to process the image to highlight the edge and texture features in the image; the data enhancement unit 15 scales, rotates, flips the image according to a preset ratio, and generates a large amount of new image data to form a large amount of data sets for iterative training of the neural network below.

[0029] In the above, the data analysis module 8 includes a thickness analysis unit 18, a pore and crack analysis unit 19, an aggregate distribution analysis unit 20, and a color and texture analysis unit 21; the thickness analysis unit 18 is used to obtain the thickness distribution information of the actual rolled concrete layer, and compare and analyze it with the design thickness to determine whether the thickness of the concrete layer meets the design requirements; the pore and crack analysis unit is used to count the number, area, and length parameters of pores and cracks, and determine whether the severity of pores and cracks exceeds the allowable range; if the pores and cracks seriously exceed the allowable range, structural reinforcement is required. Methods such as pasting carbon fiber cloth and pasting steel plates can be used to increase the force-bearing area and strength of the structure to improve the bearing capacity and stability of the structure. Before pasting carbon fiber cloth or steel plates, the concrete surface should be treated to ensure that the pasting surface is flat, dry, and oil-free, and then the pasting construction should be carried out in accordance with the specifications, and the carbon fiber cloth or steel plate should be firmly pasted on the concrete surface using a special adhesive. The aggregate distribution analysis unit 20 is used to identify the location, type and distribution density of aggregates, and quantitatively evaluate the uniformity of aggregate distribution. When the aggregate distribution is uneven, for the concrete that has been poured but not initially set, the vibration work is strengthened, and an inserted vibrating rod is used to vibrate at a certain interval and vibration time, so that the aggregates are redistributed evenly under the action of gravity and vibration force. The color and texture analysis unit 21 is used to compare the color and texture with the preset value, judge whether the color and texture of the surface of the concrete layer meet the design requirements and quality standards, and then infer the construction process and quality status of the concrete layer. For areas with lighter colors and less cement paste, it means that the cement paste in this area fails to fully fill the gap between aggregates. At this time, the vibration frequency and amplitude of the rolling equipment should be increased. Higher vibration frequency and amplitude can generate greater vibration force, prompting cement paste to flow and fill more fully between aggregates, thereby improving the uniformity of concrete. If the texture clarity is not enough, it means that the compaction effect of the concrete surface is poor. In the rolling stage, the compaction effect can be improved by increasing the compaction power of the rolling equipment to make the texture clearer.

[0030] Among them, the thickness distribution information, the number, area, and length parameters of pores and cracks are determined by the first recognition model 16. The construction process of the first recognition model 16 is: the cross-sectional image information processed by the data processing module 5 is annotated by professionals for thickness, pores, and cracks, and after annotation, the cross-sectional image information is input into the first neural network unit for iterative training, and the thickness distribution information, the number, area, and length parameters of the pores and cracks of the current roller compacted concrete layer are output to obtain the first recognition model 16. The cross-sectional image information includes images of different construction stages and different quality conditions. In this embodiment, the first neural network unit uses a U-Net network structure, which has a symmetrical encoder-decoder architecture. It includes: an input layer for receiving a data set (internal cross-sectional image information) preprocessed by the data processing module 5, the format is an RGB image, and the size is 256×256×3. The function of this layer is to input the original image data into the network to prepare for subsequent feature extraction. Encoder part: It is composed of multiple convolutional layers and pooling layers alternating. Specifically, the convolutional layer uses a 3×3 convolution kernel, a step size of 1, and a padding of 1 to keep the image size unchanged after the convolution operation. Each convolutional layer uses the ReLU activation function to increase the nonlinear expression ability of the network. For example, the first convolutional layer performs a convolution operation on the input image, converting a 3-channel image into a 64-channel feature map with a size of 256×256. As the number of network layers increases, the number of output channels of the convolutional layer gradually doubles, such as the subsequent convolutional layers output feature maps of 128, 256, and 512 channels, etc., to extract richer image features. The pooling layer follows the convolutional layer and uses a 2×2 maximum pooling method with a step size of 2. The role of the pooling layer is to reduce the resolution of the feature map, reduce the amount of data, and expand the receptive field, so that the network can extract more abstract features. After the pooling layer, the size of the feature map is halved, such as from 256×256 to 128×128, and then to 64×64, while the number of channels remains unchanged. The intermediate connection layer is located between the encoder and the decoder and consists of multiple convolutional layers. The convolutional layer in this part also uses a 3×3 convolution kernel, a step size of 1, a padding of 1, and a ReLU activation function. The intermediate connection layer further extracts the deep features of the image, deepens the network's understanding of the image, and prepares for the decoder's feature recovery and segmentation. The decoder part is symmetrical with the encoder and consists of an upsampling layer and a convolutional layer. Among them, the upsampling layer uses deconvolution or interpolation to restore the low-resolution feature map to the original image size. The deconvolution kernel size is 2×2, and the step size is 2. Through the deconvolution operation, the size of the feature map is doubled, from 32×32 to 64×64. At the same time, the number of channels is halved, from 512 channels to 256 channels. After each upsampling, one or more convolutional layers are connected to further fuse and refine the feature maps after the restored size. The convolutional layer also uses a 3×3 convolution kernel, a step size of 1, a padding of 1, and a ReLU activation function.Through these convolution operations, the accuracy of the segmentation results is ensured, so that the output segmentation mask can accurately reflect the position and shape of pores and cracks in the image. Output layer: Output the corresponding results according to different tasks. For thickness recognition, a scalar representing the thickness value is output, and the thickness distribution information within a certain range is formed on the premise of determining the position of the detection point; for pores and cracks, the number, area, and length parameters of pores and cracks are output.

[0031] In the above, the aggregate distribution, color and texture are determined by the second recognition model 17. The construction process of the second recognition model 17 is as follows: the surface image information processed by the data processing module 5 is annotated by professionals for aggregate distribution, color and texture, and after annotation, the surface image information is input into the second neural network unit for iterative training, and the aggregate distribution, color and texture information of the current roller compacted concrete layer is output to obtain the second recognition model 17. In this embodiment, the second neural network unit uses the ResNet residual network structure, including: an input layer for receiving a data set (surface image information) preprocessed by the data processing module 5, with a size of 256×256×3, which passes the original image data to the subsequent network layer for processing. Convolution layer and pooling layer: At the beginning of the network, a 7×7 convolution layer is set with a step size of 2 and a padding of 3. The input image of 256×256×3 is preliminarily extracted and downsampled, and a 127×127×64 feature map is output. Next is a 3×3 max pooling layer with a stride of 2, which further reduces the resolution of the feature map and outputs a 63×63×64 feature map. Residual block part: It is composed of multiple residual blocks stacked together. Each residual block contains two or more convolutional layers, and jump connections are used between the convolutional layers. Convolutional layer: A 3×3 convolution kernel is used. Before the first convolutional layer, there is a 1×1 convolutional layer to adjust the number of channels to match the dimension of the jump connection. Each convolutional layer is connected to a batch normalization layer and a ReLU activation function. The batch normalization layer is used to accelerate network convergence, and the ReLU activation function increases the nonlinearity of the network. Specifically, in a residual block, the first 3×3 convolutional layer converts the 64-channel feature map into 128 channels. After batch normalization and ReLU activation, the 128-channel feature map is converted back to 64 channels through the second 3×3 convolutional layer, and then added to the input feature map. After ReLU activation, a 64-channel feature map is output. Jump connection: Add the input features directly to the features after the convolution operation. This connection method effectively solves the gradient vanishing and gradient exploding problems of deep neural networks during training, allowing the network to be trained to a deeper layer and improving the network's ability to extract complex features. After the residual blocks are stacked, a global average pooling layer is connected to compress the spatial size of the feature map to 1×1, retaining only the information of the channel dimension. For aggregate category recognition, the number of output neurons of the fully connected layer is equal to the number of aggregate categories; for color feature vector output, the number of neurons is determined according to the color space dimension, such as 3 for RGB color space; for texture feature vector output, the number of neurons is determined according to the number of extracted texture features. The fully connected layer maps the feature vector after global average pooling to the corresponding output dimension, and the output includes the results of the aggregate distribution location, category, and color and texture feature vectors.

[0032] In order to achieve real-time control of the concrete unloading point and the closing strategy, in the present embodiment, the data analysis module 8 also includes a point determination unit 22 and a closing strategy unit 23; wherein the point determination unit 22 is respectively connected to the thickness analysis unit 18 and the coordinate conversion module 7, and determines the three-dimensional coordinates of the unloading point on the 3D model 3 based on the thickness difference at different positions of the previous rolled concrete layer, and converts the three-dimensional coordinates into GPS coordinates through the coordinate conversion module 7; the closing strategy unit 23 is respectively connected to the color and texture analysis unit 21, and determines the closing strategy based on the color and texture of the concrete layer surface; the closing strategy is the working strategy of the rolling equipment.

[0033] Specifically, the process of obtaining the three-dimensional coordinates of the unloading point is as follows: The three-dimensional coordinates of the measurement point are set based on the thickness distribution information as (x i ,y i , z i ), corresponding to the thickness h i , i=1,2,…,n, n is the number of measurement points; For the measurement point (x j ,y j , z j ), other measurement points in its neighborhood (x k ,y k , z k ) and the measurement point (x j ,y j , z j ) is: ; Wherein, j and k are both integers less than or equal to n; in this implementation, the neighborhood refers to a specific area around a certain measuring point. This area contains multiple other measuring points for analyzing the variation of the thickness of the RCC in the area. j ,y j , z j ) as an example, its neighborhood is the area within a certain spatial range around it. The point (x k ,y k , z k ) forms a neighborhood relationship with it. In actual operation, the size of the neighborhood is set according to the actual situation and accuracy requirements of the construction site. For example, it is set to measure the point (x j ,y j , z j) as the center and a circular area with a radius of 3 meters, then all measurement points in this circular area belong to the neighborhood of this point. By analyzing the thickness difference and distance between each point in the neighborhood and this point, the thickness change rate is calculated, and then the average thickness change rate is obtained, which provides a basis for the subsequent determination of the offset and direction of the unloading coordinates.

[0034] Calculate the thickness change rate within the neighborhood: ; in, is the measuring point (x k ,y k , z k ) thickness, is the measuring point (x j ,y j , z j ) thickness; The thickness change rate represents the thickness change per unit distance. The average thickness change rate of all thickness change rates in the neighborhood is obtained to obtain the average thickness change rate of the point. ; Based on the design thickness H of each RCC layer, for areas where the current thickness is less than the target thickness, determine the three-dimensional coordinates of the unloading point: the measurement point (x j ,y j , z j ) The added thickness is: , then the three-dimensional space distance offset is ; By analyzing the thickness of each point in the neighborhood and the measured point (x j ,y j , z j ) thickness difference, determine the offset direction vector (a, b, c) in three-dimensional space, and normalize the offset direction vector (a, b, c) to obtain a unit vector ( ), then the three-dimensional coordinates of the next unloading point of 3D model 3 are (x new ,y new , z new ),in: ; The coordinate conversion module 7 converts the three-dimensional coordinates (x new ,y new , z new ) is converted into GPS coordinates to obtain the actual location of the next unloading point.

[0035] The process of determining the closing strategy is as follows: Convert the surface image information processed by data processing module 5 into CIELAB color space and set the color deviation threshold ,when When the vibration frequency of the rolling equipment is increased by 5Hz and the amplitude is increased by 0.1mm for the light-colored area, the vibration frequency is reduced by 3Hz and the amplitude is reduced by 0.05mm for the dark-colored area. : For light-colored areas, the vibration frequency is increased by 10Hz, the amplitude is increased by 0.2mm, and the number of rolling times is increased by 2 times; for dark-colored areas, the vibration frequency is reduced by 5Hz, the amplitude is reduced by 0.1mm, and the number of rolling times is reduced by 1 time; Calculating a texture clarity index and a direction consistency index based on the texture information output by the second recognition model; Set the texture definition threshold 15, direction standard deviation threshold is 10°; When the texture clarity : Increase the compaction power of the rolling equipment by 20% or replace it with a heavier rolling wheel to enhance the compaction effect and make the texture clearer; When the standard deviation of the grain direction : Re-plan the driving route of the rolling equipment to ensure that the deviation between the rolling direction and the design direction each time is less than 5°, and at the same time increase the overlapping width of the rolling wheels.

[0036] The embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. Roller compacted concrete gravity dam layer treatment monitoring system, characterized by: include: A data acquisition module, used to acquire surface image information and internal cross-sectional image information of the RCC layer; A data processing module, connected to the data acquisition module, for processing the surface image information and the internal cross-sectional image information so that the surface image information and the internal cross-sectional image information reflect the real RCC layer information; A three-dimensional construction module is used to construct a virtual scene of a gravity dam, and configure a virtual roller compacted concrete layer based on the design requirements based on the virtual scene of the gravity dam. The three-dimensional construction module is also connected to the data processing module, and the roller compacted concrete layer is modeled in layers based on the real roller compacted concrete layer information, and the 3D model of each roller compacted concrete layer is placed in the corresponding virtual scene of the gravity dam, and a dynamic stress belt is set based on the preset rolling trajectory of each group of rolling equipment, so as to simulate the roller compaction equipment in the virtual scene of the gravity dam to pressurize each roller compacted concrete layer; A coordinate conversion module is used to convert and match the three-dimensional coordinates in the gravity dam virtual scene with the measured GPS coordinates of the gravity dam; a data analysis module, connected to the three-dimensional construction module and the data processing module respectively, extracting parameters of the processed surface image information and the internal cross-sectional image information, obtaining parameter information of each RCC layer, regulating the 3D model of each RCC layer based on the parameter information, regulating the concrete unloading point and the liquidation strategy in real time based on the parameter information of the previous RCC layer, and simulating the next RCC layer in the gravity dam virtual scene based on the concrete unloading point and the liquidation strategy; The judgment module is used to compare the latter RCC layer generated by simulation with the virtual RCC layer configured based on design requirements, output a comparison value, and fine-tune the concrete unloading point and closing strategy based on the comparison value.

2. The roller compacted concrete gravity dam surface treatment monitoring system according to claim 1 is characterized in that: The data acquisition module includes a high-definition camera unit, an X-ray and CT scanning unit; The high-definition camera unit is equipped with a plurality of high-definition cameras for acquiring surface image information of the RCC layer from different angles and positions; The X-ray and CT scanning unit uses X-ray and CT scanning equipment to perform internal scanning on the RCC layer to obtain internal cross-sectional image information of the RCC layer.

3. The roller compacted concrete gravity dam surface treatment monitoring system according to claim 2 is characterized in that: The data processing module includes a geometric correction unit, a radiation correction unit, an image enhancement unit and a data enhancement unit; The geometric correction unit is used to perform geometric correction on the surface image information and the internal cross-sectional image information, eliminate the image deformation caused by the shooting angle and lens distortion, and restore the shape and position of the object in the image to the real state; The radiation correction unit is used to perform radiation correction on the image and adjust the brightness and contrast of the image by counting the brightness distribution of the surface image information; The image enhancement unit processes the image using histogram equalization to highlight edge and texture features in the image; The data enhancement unit scales, rotates, and flips the image according to a preset ratio to generate a large amount of new image data.

4. The roller compacted concrete gravity dam surface treatment monitoring system according to claim 3 is characterized in that: The data analysis module includes a thickness analysis unit, a pore and crack analysis unit, an aggregate distribution analysis unit, and a color and texture analysis unit; The thickness analysis unit is used to obtain the thickness distribution information of the actual rolled concrete layer, and compare and analyze it with the designed thickness to determine whether the thickness of the concrete layer meets the design requirements; The pore and crack analysis unit is used to count the number, area, and length parameters of pores and cracks, and determine whether the severity of the pores and cracks exceeds the allowable range; The aggregate distribution analysis unit is used to identify the location, type and distribution density of aggregates and quantitatively evaluate the uniformity of aggregate distribution; The color and texture analysis unit is used to compare the color and texture with preset values, determine whether the color and texture of the concrete layer surface meet the design requirements and quality standards, and then infer the construction process and quality status of the concrete layer.

5. The roller compacted concrete gravity dam surface treatment monitoring system according to claim 4, characterized in that: The thickness distribution information, the number, area, and length parameters of pores and cracks are determined by a first recognition model, and the construction process of the first recognition model is as follows: The cross-sectional image information processed by the data processing module is annotated by professionals for thickness, pores, and cracks. After annotating, the cross-sectional image information is input into the first neural network unit for iterative training, and the thickness distribution information, the number, area, and length parameters of the pores and cracks of the current roller compacted concrete layer are output to obtain a first recognition model; The cross-sectional image information includes images at different construction stages and in different quality conditions.

6. The roller compacted concrete gravity dam surface treatment monitoring system according to claim 5, characterized in that: The aggregate distribution, color and texture are determined by a second recognition model, and the construction process of the second recognition model is: The surface image information processed by the data processing module is annotated by professionals for aggregate distribution, color and texture. After annotation, the surface image information is input into the second neural network unit for iterative training, and the aggregate distribution, color and texture information of the current roller compacted concrete layer is output to obtain the second recognition model.

7. The roller compacted concrete gravity dam surface treatment monitoring system according to claim 6, characterized in that: The data analysis module also includes a point determination unit and a position closing strategy unit; The point determination unit is connected to the thickness analysis unit and the coordinate conversion module respectively, determines the three-dimensional coordinates of the unloading point on the 3D model based on the thickness difference of different positions of the previous rolled concrete layer, and converts the three-dimensional coordinates into GPS coordinates through the coordinate conversion module; The closing strategy unit is connected to the color and texture analysis unit respectively, and determines the closing strategy based on the color and texture of the surface of the concrete layer; The closing strategy is the working strategy of the rolling equipment.

8. The roller compacted concrete gravity dam surface treatment monitoring system according to claim 7, characterized in that: The process of obtaining the three-dimensional coordinates of the unloading point is as follows: The three-dimensional coordinates of the measurement point are set based on the thickness distribution information as (x i ,y i , z i ), corresponding to the thickness h i , i=1,2,…,n, n is the number of measurement points; For the measurement point (x j ,y j , z j ), other measurement points in its neighborhood (x k ,y k , z k ) and the measuring point (x j ,y j , z j ) is: ； Wherein, j and k are both integers less than or equal to n; Calculate the thickness change rate within the neighborhood: ; in, is the measuring point (x k ,y k , z k ) thickness, is the measuring point (x j ,y j , z j ) thickness; The thickness change rate represents the thickness change per unit distance. The average thickness change rate of all thickness change rates in the neighborhood is obtained to obtain the average thickness change rate of the point. ; Based on the design thickness H of each RCC layer, for areas where the current thickness is less than the target thickness, determine the three-dimensional coordinates of the unloading point: the measurement point (x j ,y j , z j ) The added thickness is: ; Then the three-dimensional space distance offset is ; By analyzing the thickness of each point in the neighborhood and the measured point (x j ,y j , z j ) thickness difference, determine the offset direction vector (a, b, c) in three-dimensional space, and normalize the offset direction vector (a, b, c) to obtain a unit vector ( ), then the three-dimensional coordinates of the next unloading point of the 3D model are (x new ,y new , z new ),in: ； The coordinate conversion module converts the three-dimensional coordinates (x new ,y new , z new ) is converted into GPS coordinates to obtain the actual location of the next unloading point.

9. The roller compacted concrete gravity dam surface treatment monitoring system according to claim 8, characterized in that: The process of determining the closing strategy is as follows: Convert the surface image information processed by the data processing module into the CIELAB color space and set the color deviation threshold ,when When the vibration frequency of the rolling equipment is increased by 5Hz and the amplitude is increased by 0.1mm for the light-colored area, the vibration frequency is reduced by 3Hz and the amplitude is reduced by 0.05mm for the dark-colored area. : For light-colored areas, the vibration frequency is increased by 10Hz, the amplitude is increased by 0.2mm, and the number of rolling times is increased by 2 times; for dark-colored areas, the vibration frequency is reduced by 5Hz, the amplitude is reduced by 0.1mm, and the number of rolling times is reduced by 1 time; Calculating a texture clarity index and a direction consistency index based on the texture information output by the second recognition model; Set the texture definition threshold 15, direction standard deviation threshold is 10°; When the texture clarity : Increase the compaction power of the rolling equipment by 20% or replace it with a heavier rolling wheel to enhance the compaction effect and make the texture clearer; When the standard deviation of the grain direction : Re-plan the driving route of the rolling equipment to ensure that the deviation between the rolling direction and the design direction each time is less than 5°, and at the same time increase the overlapping width of the rolling wheels.

Images