JP7219910B2 - Optimal Compaction Judgment Construction System for Concrete - Google Patents

Description

The present invention relates to a concrete optimal compaction determination construction system for determining an optimal compaction state when compacting fresh concrete.

Normally, when compacting fresh concrete, it is required to remove the air entrained in the concrete from the concrete during transportation and placing, and to ensure that the reinforcing bars and buried objects are well adhered to the concrete to make the concrete uniform and dense. In general, compaction of concrete is performed by inserting a vibrator into the concrete that has been poured into the formwork and applying vibration to the concrete. Further, in this case, determination of sufficient compaction of concrete, that is, determination of compaction, is made by visual judgment of the concrete surface by an experienced worker or engineer. Sufficient vibration compaction allows concrete to be densely packed into every corner of the formwork and reinforcing bars, and voids such as air bubbles inside and on the surface of the concrete are eliminated to improve strength, watertightness, and durability. You can make good concrete.

Regarding sufficient compaction of concrete, as a general guideline, the Standard Specifications for Concrete states, ``One of the evidences that vibration compaction is sufficient is the appearance of a line of cement paste on the contact surface between the concrete and the dam. In addition, it can be seen from the fact that the decrease in the volume of the concrete is no longer recognized, the surface is glossy, and the entire concrete seems to have melted uniformly.”

Conventionally, Patent Document 1 proposes a concrete compaction method for the purpose of ensuring the required quality of cast concrete without being influenced by the individual skill of workers. Japanese Patent Application Laid-Open Nos. 2003-100001 and 2003-200010 propose techniques for determining the compaction of placed concrete using a measurement technique using a distance sensor.

In the concrete compaction method of Patent Document 1, a floating vibrator is installed at the bottom of the concrete formwork to be poured, and the concrete is placed in the formwork to a predetermined height, and the concrete near the vibrator is By levitating the inside of the concrete upward while vibrating the vibrator according to the pressure, the cast concrete is compacted upward from the bottom.

In this case, when the concrete placed in the formwork reaches a predetermined thickness, the vibrator is driven by a detection signal of the concrete pressure detected by the pressure sensor, thereby vibrating the vibrator at a predetermined frequency. It compacts the surrounding concrete and reduces the frictional resistance between the vibrator and the concrete. Further, since the bulk specific gravity of the vibrator is lighter than the specific gravity of the concrete, a well-known buoyant force is generated in the vibrator as the frictional resistance between the vibrator and the concrete decreases, and the vibrator vibrates to tighten the surrounding concrete. As it hardens, it rises upwards. Then, when the vibrator is positioned on the upper surface of the concrete, a detection signal indicating that the concrete pressure is 0 is transmitted from the pressure sensor to the prime mover, the vibrator stops, and compaction of the concrete ends.

As described above, in this concrete compaction method, by using a floating vibrator, the concrete compaction work can be automated and freed from conventional heavy work, and the degree of compaction of concrete can be determined by individual skills. It is said that the reliability of construction can be improved because it is not dependent.

In the technology for judging the compaction of concrete disclosed in Patent Document 2, compaction is performed by applying vibration with a vibrator inserted into the concrete. Here, the vibrator is temporarily stopped at predetermined time intervals set in advance, and the shape data of the surface S of the concrete is acquired by the shape data acquisition means (camera, etc.). By analyzing the shape data by the smoothness detection means, the uneven shape of the surface is grasped, and the acquired smoothness of the surface is detected. The compaction degree determination means quantitatively determines the compaction degree of concrete based on the detected smoothness. As a result, when a predetermined degree of compaction is reached, it is determined that compaction is completed.

In addition, the technology for determining the compaction of concrete in Patent Document 3 includes a three-dimensional distance sensor that projects a light ray onto an object and acquires three-dimensional distance data that is a set of point cloud data, and the acquired A coordinate conversion unit that converts three-dimensional distance data into point cloud data based on three-dimensional orthogonal coordinates, a plane extraction unit that extracts a point group recognized as a plane from the converted point cloud data, and the extracted plane A determination system comprising a point cloud number calculation unit that calculates the number of point clouds of the point cloud that constitutes the , it is determined that the compaction of concrete is completed when the number of point groups is equal to or greater than the threshold.

JP-A-4-357274 JP 2017-025609 A JP 2017-053084 A

However, conventional determination of concrete compaction is based on the visual judgment of the concrete surface by experienced workers and engineers, and the judgment criteria depend on the experience and skill of the workers and engineers. The number of people who can make appropriate judgments is limited, and on the other hand, there may be individual differences in judgment results between workers and engineers. There is a problem that the finish quality of concrete may vary.

In addition, although the concrete compaction method of Patent Document 1 has an advantage in that it does not depend on the individual skill of the worker, since the vibrator itself floats, determination of compaction of concrete is made by the vibrator in the concrete. In this case as well, since it is not possible to grasp the state of propagation of vibration inside the concrete, there is a possibility that the finished quality of the concrete may vary.

In addition, in the concrete compaction determination method of Patent Document 2, in order to acquire the determination image of the placed concrete, the vibrator is temporarily stopped at predetermined time intervals, and the shape data acquisition means (camera, etc.) ) to obtain the shape data of the surface S of the concrete, there is a problem that the work is interrupted.

Further, in the concrete compaction determination method of Patent Document 3, a photographing method such as a camera is not used, and the three-dimensional distance data is converted into point cloud data based on three-dimensional orthogonal coordinates, and the point cloud data There is a problem that preliminary work by a human (designer), such as designing in advance a threshold value representing the relationship between the degree of concrete compaction and the degree of compaction of concrete, is troublesome.

The present invention solves such conventional problems, and its object is to determine compaction of concrete by an operation similar to the visual judgment of the concrete surface by experienced workers and engineers. To provide a construction system for determining optimal compaction of concrete, eliminating individual differences such as the experience and skill of workers and engineers in compaction determination of concrete by computerizing determination of compaction.

A second object of the present invention is to provide a concrete optimal compaction determination construction system that optimizes concrete compaction determination processing by a computer and enables high-quality compaction determination.

In order to achieve the above object, the present invention provides a vibrator that is inserted into placed concrete to vibrate the concrete to compact the concrete, and a photographing means that acquires an image of the compacted concrete. Then, supervised learning is performed using images or sets of image feature values and known numerical sequences related to the degree of compaction of concrete as learning data, and compaction of the concrete based on the images acquired from the photographing means. an image feature amount calculation unit for calculating a numerical value string that is a feature amount related to the condition; and supervised learning based on the numerical value sequence calculated by the image feature amount calculation unit, and a numerical value representing the compaction condition of the concrete. and a compaction determination unit that calculates the predicted value from the learning unit and determines compaction. In the compaction determination system, the learning unit may derive a numerical predicted value related to the surface state of the concrete as the predicted value.

In order to achieve the above object, the present invention also acquires an image of the surface of cast concrete using a photographing means, and an image feature amount calculation unit calculates the image or the image feature amount and the compaction condition of the concrete. Supervised learning is performed using a set of known numerical value sequences for learning as learning data, and based on the image acquired from the imaging means, a numerical value sequence that is a feature amount related to the compaction condition of the concrete is calculated and learned The unit performs supervised learning based on the numerical value sequence calculated by the image feature amount calculation unit, derives a predicted value of the numerical value representing the degree of compaction of the concrete , and the compaction determination unit determines from the learning unit The gist of this paper is the optimum compaction determination method for concrete, which calculates the predicted value of .

In order to achieve the above object, the present invention further provides a vibrator that is inserted into placed concrete to vibrate the concrete to compact the concrete, and a camera that acquires an image of the compacted concrete. means, a vibrator driving means for moving the vibrator upward and downward and changing its position with respect to the cast concrete, and a photographing means for moving upward and downward and changing its position with respect to the concrete to be compacted. Supervised learning is performed using as learning data an imaging means driving means and a set of images or image feature quantities and known numerical sequences related to the degree of compaction of concrete, and based on the images acquired from the imaging means, the above An image feature quantity calculation unit that calculates a numerical value string that is a feature quantity related to the degree of compaction of concrete; and supervised learning is performed based on the numerical value sequence calculated by the image feature quantity calculation unit, and compaction of the concrete. A learning unit that derives a predicted value of a numerical value representing the condition , a compaction determination unit that performs compaction determination by calculating the predicted value from the learning unit, and an operation that drives and controls the vibrator driving means and the photographing means driving means. and a control unit.

According to the optimal compaction determination construction system for concrete of the present invention, when inserting a vibrator into the concrete cast into the formwork and applying vibration to compact the concrete, the machine learning function of AI (Artificial Intelligence) is used. Therefore, it is possible to determine the optimum compaction of concrete by eliminating individual differences such as the experience and skill of workers and engineers in determining the compaction of concrete. construction can be carried out. In addition, by optimizing the compaction judgment processing of concrete by a computer, it is possible to judge compaction with high quality.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a configuration diagram showing the overall configuration of a concrete optimum compaction determination construction system according to an embodiment of the present invention; It is a block diagram which shows an example of the hardware constitutions of the computer with which the control apparatus in the said embodiment was equipped. 3 is a functional block diagram showing the functional configuration of the control device shown in FIG. 2; FIG. 4 is a flow chart showing a processing procedure of compaction determination and construction control by the control device of the compaction determination construction system according to the embodiment; It is a processing flow chart showing an example of the operation procedure by CNN adopted in the learning operation of the above-mentioned embodiment. FIG. 10 is a diagram showing how a labeled frame image acquired by the image acquiring unit changes over time due to a compaction operation in the above embodiment. FIG. 4 is a diagram showing the quality of compaction determination results when a compaction test is carried out on three types of concrete having different freshness in the above-described embodiment.

Next, a mode for carrying out the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the overall configuration of a concrete optimal compaction determination construction system (hereinafter simply referred to as "compaction determination construction system") according to an embodiment of the present invention. As shown in FIG. 1, this compaction determination construction system 1 includes a formwork 3 containing concrete 2, a vibrator 4 inserted into the concrete 2 in the formwork 3, and a vibrator 4 during compaction work. It comprises a video camera 5 as a photographing means for photographing the surface of the concrete 2 near the insertion position, and a control device 6 connected to the video camera 5 for acquiring photographing signals from the video camera 5 and performing various data processing. It consists

The vibrator 4 is attached to the tip (lower end) of a support rod 7 that suspends and supports the vibrator 4 . The video camera 5 is attached to the tip (lower end) of a camera support rod 8 that suspends and supports the video camera 5 . The support bar 7 and the camera support bar 8 are each movably connected to separate rails (not shown) and adapted to be moved within the planar area of the formwork 3 by drives. An elevator (not shown) is attached to the support rod 7, and the elevator moves the support rod 7 vertically, thereby moving the vibrator 4 vertically with respect to the concrete 2. . Another camera elevating member (not shown) is attached to the camera support rod 8 , and the camera elevating member vertically moves the camera support rod 8 to move the video camera 5 to the surface of the concrete 2 . The height from the surface of the concrete 2 can be adjusted by moving it in the vertical direction. Also, the video camera 5 is set so as to photograph the surface of the concrete 2 within a range of about 30 cm (centimeters) from the vibrator 4 by adjusting its position and height.

As described above, the video camera 5 and the control device 6 are in a connection relationship. Alternatively, the video camera 5 and the control device 6 may be connected via a public communication network (the Internet, etc.) or a short-distance simple communication system (Bluetooth, etc.). Further, the video camera 5 is equipped with a distance measuring means such as a laser rangefinder, and measures the distance between the concrete placing surface 2 and the video camera 5 to cope with the difference in the scale of the captured image. there is Scale correspondence can also be achieved by including a scale in the acquired image. In this compaction determination construction system 1, the degree of compaction of concrete 2 as an object is predicted by machine learning, and the completion of compaction is determined. In addition to the video camera 5, the photographing means may be a digital camera, a 3D distance sensor, or any other means capable of capturing a two-dimensional array image.

The machine learning used in this embodiment is a pattern conversion that converts an image or image features into a numeric string by learning learning data, which is a set of images or image features and known numeric strings related to the degree of compaction of concrete. It is a process of obtaining a function and predicting an unknown value using the pattern conversion function. In this embodiment, teacher data, which is a sequence of numerical values, is used, and a pattern conversion function obtained from the teacher data is used to predict a value at a future point in time. Note that the numerical value sequence is a sequence of numerical values obtained from various observed values of a phenomenon related to the concrete 2, and is a sequence of numerical values obtained by observing the phenomenon based on a certain rule. In this embodiment, an image to which information indicating that compaction is not completed (before) or compaction is completed (just) is used as teaching data. As an example of machine learning, there is a convolution neural network (hereinafter referred to as "CNN"), and in the present embodiment, an example using deep learning in this neural network will be described.

Other examples of machine learning include a recurrent neural network (RNN Recurrent Neural Network), a multi-layer perceptron (MLP Multilayer Perceptron), a support vector machine (SVM Support Vector Machine), or a support vector regression corresponding to the SVM. (SVR Support Vector Regression), decision tree learning, association rule learning, Bayesian network, etc., and the compaction determination construction system 1 according to the present embodiment uses one of these algorithms depending on its usefulness. may

An object predicted by the compaction determination construction system 1 according to the present embodiment is the condition of the surface of the concrete 2 as the degree of compaction of the placed concrete 2 .

The compaction determination construction system 1 is configured so that an image signal acquired by the video camera 5 can be acquired by the control device 6 via the signal line 9 or a communication network. Further, the movement and lifting operation of the vibrator 4 can be controlled by control signals sent from the control device 6 . Furthermore, the movement and elevation of the video camera 5 can be controlled by control signals sent from the control device 6 .

Controller 6 comprises one or more computers. When the control device 6 includes a plurality of computers, each functional element of the control device 6, which will be described later, is implemented by distributed processing. The type of individual computer is not limited. For example, a stationary or portable personal computer (PC) may be used, a workstation may be used, and a mobile terminal such as a high-performance mobile phone (smartphone), mobile phone, or personal digital assistant (PDA) may be used. may be used. Alternatively, the control device 6 may be constructed by combining various types of computers. When using multiple computers, these computers are connected via a communication network such as the Internet or an intranet.

FIG. 2 is a block diagram showing an example of the hardware configuration of each computer 50 provided in the control device 6. As shown in FIG. The computer 50 includes a CPU (processor) 51, which is an arithmetic unit that executes an operating system and an application program for compaction determination, a ROM (read-only memory) that stores the application program, Alternatively, a main storage unit 52 composed of RAM (random access memory: memory rewritable at any time) storing various data processed at any time in data processing operation, and an auxiliary storage unit composed of hard disk, flash memory, etc. 53, a communication control unit 54 composed of a network card or a wireless communication module, an input device 55 such as a keyboard and a mouse, and an output device 56 such as a display and a printer. Note that the hardware modules to be installed may differ depending on the type of computer 50 . For example, stationary PCs and workstations often have keyboards and mice as input and output devices, and monitors such as displays as output devices, but smartphones have touch panels that function as input and output devices. often do.

Each functional element of the control device 6, which will be described later, causes the CPU 51 or the main storage unit 52 to read predetermined software, operates the communication control unit 54, the input device 55, the output device 56, etc. under the control of the CPU 51, It is realized by reading and writing data in the main storage unit 52 or the auxiliary storage unit 53 . Data and databases required for data processing are stored in the main storage unit 52 or the auxiliary storage unit 53 .

FIG. 3 is a functional block diagram showing the functional configuration of the control device 6 shown in FIG. 2 above. As functional elements, the control device 6 is connected to the video camera 5, and has an image acquisition section 21 for inputting images continuously acquired by the video camera 5, and the main storage section 52 or the auxiliary storage section 53. A database 22 provided to store teacher data and data necessary for various processes, a teacher data acquisition unit 23 connected to the database 22 and receiving teacher data from the database 22, an image acquisition unit 21, and a teacher data acquisition unit 23, and calculates the feature amount of the input image based on the data from the operation units 21 and 23; 5 using machine learning to learn the relationship between the image obtained from 5 and the numerical value sequence calculated by the image feature amount calculation unit, and learning to derive a predicted numerical value representing the degree of compaction of the concrete. and a section (predicted value derivation section) 25 . The control device 6 further includes a calculation unit 26 that performs calculations for compaction determination based on the processing results of the image feature amount calculation unit 24 and the learning unit 25, and is connected to the calculation unit 26, and performs tightening based on the calculation result of the calculation part. A compaction determination unit 27 that determines whether the compaction condition is incomplete (before) or completed (just), and is connected to the compaction determination unit 27. In the compaction determination unit 27 It is connected to an output unit 28 for outputting the judgment result to an output device 56 such as a monitor and to a compaction judging unit 27. Based on the judgment result in the compaction judging unit 27, the installation position of the vibrator 4 can be changed, and the video camera 5 and an operation control unit 29 that performs a control operation such as changing the installation position of the. The image feature amount calculation unit 24 passes the received image to the learning unit 25 as it is or calculates the image feature amount and then passes it to the learning unit 25 by machine learning used in the learning unit 25 .

FIG. 4 is a flow chart showing a processing procedure of compaction determination and construction control by the control device 6 of the compaction determination construction system according to the present embodiment.

First, when the process of compaction determination construction control by the control device 6 is activated by an instruction input by the user, etc., the vibrator 4 is lowered at a predetermined position (insertion point A) in the formwork 3, that is, , the lowering operation is started, and the vibrator 4 is inserted into the concrete 2 (step S1). The vibrator 4 is equipped with a mechanism for raising and lowering the vibrator 4 with respect to the surface of the concrete 2 and is controlled by the control device 6 .

Next, the vibrator 4 is started (step S2), and the operation of the video camera 5 is started, and the surface of the concrete 2 being vibrated by the vibrator is photographed (step S3). A mechanism for maintaining the distance between the video camera 5 and the surface of the concrete 2 and a distance measuring device (such as a laser rangefinder) are mounted on the video camera 5 or the camera support rod 8 and controlled by the control device 6 . Symbol P1 in FIG. 1 represents the imaging range of the video camera 5 in this embodiment. This photographing range P1 is, for example, a range of approximately 30 cm (centimeters) from the vibrator 4 .

Next, image data is acquired by the image acquisition unit 21 and teacher data is acquired by the teacher data acquisition unit 22, and a two-dimensional array image is captured (step S4). In this process of capturing two-dimensional array images, the image acquiring unit 21 receives images successively acquired by the video camera 5 as image signals of labeled frame images. The image acquisition unit 21 analog/digital converts the received image to generate image data, and sends the generated image data to the image feature quantity calculation unit 24 as a two-dimensional array image. The teacher data acquisition unit 23 acquires teacher data from the database 22 and sends the teacher data as a two-dimensional array image to the image feature quantity calculator 24 .

The operation of the compaction judgment construction system according to the present embodiment has a learning phase and a prediction (judgment execution) phase. Therefore, first, the operation of the learning phase will be described.
Learning Phase In this learning phase, a set (learning data) of an image or an image feature amount and a known numerical value string representing the degree of compaction is stored in the database 22. After that, the stored learning data is read out and machine learning is performed. It is a mode of operation that allows the CNN to obtain a pattern transformation function using When CNN is used for machine learning as in the present embodiment, the two-dimensional array image passes through the image feature amount calculation unit 24 and is input to AI, that is, CNN ( step S5). The CNN has previously learned the correspondence relationship between the frame images of the video input from the video camera 5 and the compaction state. Therefore, the input frame image is classified into a class with high likelihood (likelihood). The image feature amount calculator 24 calculates a numerical vector (numerical sequence) as a feature amount relating to the degree of compaction of the concrete 2, which is the object, using machine learning based on the image data. For example, the image feature amount calculation unit 24 uses the image data and the teacher data to calculate the difference in compaction amount of the concrete 2 caused by the time difference (for example, after a predetermined time from the acquisition of each image data) in the future ( Hereinafter, supervised machine learning is performed with a numerical value (objective variable) to be predicted by machine learning (referred to as “differential compaction amount)”. The above CNN is an example of machine learning, and is a known deep learning method A convolutional neural network.

FIG. 5 is a processing flow chart showing an example of the operation procedure of the learning phase by CNN. As shown in FIG. 5, the CNN performs convolution processing in the first convolution layer on the input image data 40 (step S21), and then performs pooling processing in the first pooling layer (step S22 ), and then execute convolution processing in the second convolution layer (step S23), then execute pooling processing in the second pooling layer (step S24), and then execute convolution processing in the third convolution layer 3. (step S25), then the processing of the fully connected layer A is executed (step S26), and the processing of the fully connected layer B is executed (step S27), thus repeating the convolution processing operation and the pooling processing operation. This is a method that can mechanically extract feature values from image data. As an example of using a CNN, the image feature amount calculation unit 24 repeats the processing in the convolution layer and the pooling layer several times (for example, 5 times), and then passes through the processing in the output layer to obtain the target Calculate the predicted value of the differential compaction amount, which is a variable. This predicted value represents the future compaction condition of the concrete 2 . At that time, the image feature amount calculation unit 24 learns the relationship between the image and the numerical value sequence related to the degree of compaction according to the CNN learning algorithm, and updates the CNN connection weight, thereby updating the pattern conversion function. can. Here, once the image feature amount calculation unit 24 constructs the pattern conversion function, updating of the pattern conversion function using the measured value data may be stopped, and the pattern conversion function is constructed from the beginning. In some cases, the function of updating the pattern conversion function may not be included. Furthermore, the image feature amount calculation unit 24 outputs the output of the output layer targeting the image data in the constructed state to the learning unit 25 as a numerical vector (for example, a 256-column numerical vector) that is a feature amount. do.

Next, the prediction (determination execution) phase will be described.
Prediction (Judgment Execution) Phase This prediction (judgment execution) phase is an operation mode in which the pattern conversion function obtained in the learning phase described above is used to predict a numerical sequence using the CNN obtained by the machine learning. FIG. 6 shows an operation example of this prediction (decision execution) phase. FIG. 6 is a diagram showing how the labeled frame image acquired by the image acquisition unit 21 changes over time due to the compaction operation. As shown in FIG. 6, compaction has not progressed immediately after placing the concrete 2 (after 0 seconds), and the compaction determination result is also the result of compaction incomplete (before). Four seconds after the concrete 2 is placed, the compaction has not yet progressed sufficiently, and the compaction determination result is still incomplete compaction (before). Eight seconds after the concrete 2 is placed, the compaction has progressed sufficiently, and the compaction determination result indicates that compaction is completed (just). In the state 16 seconds after the concrete 2 is placed, the compaction has progressed sufficiently, and the compaction determination result is the compaction completion (just) determination.

Based on the numerical vector calculated by the image feature amount calculation unit 24, the learning unit 25 derives a predicted value of the differential compaction amount as a numerical value representing the degree of compaction of the concrete 2. FIG. To explain an example of this predicted value derivation processing operation, the learning unit 25 performs machine learning based on a numerical vector obtained by combining a numerical vector and time-series environmental measurement data, thereby obtaining a difference, which is an objective variable. Calculate the transition of the predicted value of the compaction amount. At that time, the learning unit 25 updates pattern conversion functions such as filter parameters by executing machine learning based on the teacher data. Once the learning unit 25 constructs the pattern conversion function, it may stop updating the pattern conversion function using the measured value data. Function update functionality need not be included. Then, the learning unit 25 outputs to the compaction determination unit 27 data on the transition of the predicted value of the difference compaction amount obtained by machine learning in the state in which the pattern conversion function is constructed.

The compaction determination unit 27 determines whether the compaction condition of the concrete 2 is incomplete or completed based on the transition data of the predicted value of the differential compaction amount (step S6). If the compaction is not completed, a "before" signal is output, and the process returns to step S3 for photographing by the video camera 5 . On the other hand, if compaction is complete, a "just" signal is output. Furthermore, the compaction determination unit 27 sends the “before” signal or “just” signal to the operation control unit 29 .

The operation control unit 29 generates control signals for controlling operations of the vibrator 4 and the video camera 5 based on the compaction determination result of the concrete 2 output by the compaction determination unit 27 . Then, when the compaction determination unit 27 outputs the "just" signal, the operation control unit 29 starts the operation of winding up the vibrator (step S7), and after performing this operation, stops the vibrator 4 (step S7). S8).

Next, the CPU 50 of the control device checks whether or not the compaction of the entire area of the concrete 2 is completed (step S9). The motion control unit 29 is instructed to move (step S10). The motion control unit 29 moves the support rod 7 and the camera support rod 8 along the corresponding rails, and changes the compaction position by the vibrator 4 from the insertion point A to the insertion point B as shown in FIG. Also, the photographing area of the video camera 5 is changed from the photographing range P1 to the photographing range P2. Then, the operations of steps S1 to S9 are performed by the processing operation of the control device 6. FIG.

Example A compaction test was performed on three types of concrete with different freshness. In this compaction test, the video camera 5 captured 50 compaction test images at 30 frames per second.

Then, from the compaction test video of 30 frames per second, the average value of the appropriate time (specified by the frame number) for completion of compaction (just) judged by a plurality of skilled operators is taken, This average value was used as training data for the borderline between uncompacted compaction (before) and compaction completed (just) in the compaction test of the concrete. Further, all the frame images acquired by the image acquiring unit 21 were given correct labels of "before" or "just", and subjected to "compaction determination" and "evaluation of compaction determination" tests. In addition, considering the difference in the compaction state of concrete due to the position of the vibrator during the compaction test, the frame image was divided into two images, left and right, and the appropriate time was determined for each.

FIG. 7 is a diagram showing the quality of the compaction determination result when the compaction test is carried out on the three types of concrete 2 with different freshness as described above. In FIG. 7, Test 01, Test 02, and Test 03 represent compaction tests for three types of concrete 2 with different freshness. In each test, the upper row represents the boundary of compaction determination "before" or "just" in the teacher data, and the lower row represents the pass/fail of the determination result by the compaction determination construction system according to the present embodiment. Regarding the quality of the determination result in the lower part of FIG. 7, the determination result "correct" indicates that the determination result was good both in the compaction determination "before" and "just". On the other hand, the determination result "incorrect" indicates that both the compaction determinations "before" and "just" were erroneous. As shown in FIG. 7, in test 01, the judgment result (lower row) by the judgment construction system for the teacher data (upper row) is the boundary part of "before" or "just" and it is not so far from this boundary part Judgment results that match the outline teaching data are obtained because the judgment errors (black parts) appear in the range.

In test 02, the judgment results (lower) by the judgment construction system for the teacher data (upper) are not only the boundary part of the above "before" or "just", but also in the "before" direction from the boundary (about 550 frame positions) , and also in the "just" direction (positions ahead of about 1600 frames), a judgment error (black portion) appears, so that a considerably inaccurate judgment result is obtained.

In test 03, the judgment result (lower row) by the judgment construction system for the teacher data (upper row) shows a judgment error (blackened part) only in the part close to the boundary of "before" or "just". As a result, determination results that are very consistent with the training data are obtained.

In this compaction test, only the image obtained by resizing the right image of the frame image to 64×64 pixels was used.

According to the present invention, in the compaction determination of concrete, the machine learning function of AI is used to predict the compaction condition of concrete and use it for compaction determination. It is useful because it is possible to determine the optimal compaction of concrete and to perform construction without individual differences such as experience and skill.

1 Compaction judgment construction system 2 Concrete 3 Formwork 4 Vibrator 5 Video camera (photographing means)
6 control device 7 support rod 8 camera support rod 9 signal line 21 image acquisition unit 22 database 23 teacher data acquisition unit 24 image feature quantity calculation unit 25 learning unit (prediction value derivation unit)
26 calculation unit 27 compaction determination unit 28 output unit 29 operation control unit

Claims (4)

a vibrator that is inserted into the placed concrete to vibrate the concrete to compact the concrete;
a photographing means for acquiring an image of the concrete to be compacted;
Supervised learning is performed using an image or a set of image feature values and known numerical sequences related to the degree of compaction of concrete as learning data, and based on the images acquired from the photographing means, the degree of compaction of concrete is determined. an image feature amount calculation unit that calculates a numerical sequence that is a feature amount;
a learning unit that performs supervised learning based on the numerical sequence calculated by the image feature amount calculation unit and derives a predicted value of a numerical value representing the degree of compaction of the concrete ;
A compaction determination unit that calculates the predicted value from the learning unit and performs compaction determination;
Optimal compaction determination system for concrete. The learning unit derives a predicted numerical value related to the surface state of the concrete as the predicted value.
The optimal compaction determination system for concrete according to claim 1. Acquiring an image of the surface of the placed concrete using a photographing means,
The image feature amount calculation unit performs supervised learning using an image or a set of image feature amounts and known numerical sequences related to the degree of compaction of concrete as learning data, and based on the image acquired from the photographing means, calculating a numerical value sequence that is a feature quantity related to the degree of compaction of the concrete;
The learning unit performs supervised learning based on the numerical value sequence calculated by the image feature amount calculation unit, and derives a predicted numerical value representing the degree of compaction of the concrete ,
The compaction determination unit calculates the predicted value from the learning unit and performs compaction determination.
A method for determining optimum compaction of concrete, characterized by: a vibrator that is inserted into the placed concrete to vibrate the concrete to compact the concrete;
a photographing means for acquiring an image of the concrete to be compacted;
vibrator drive means for causing the vibrator to move up and down and change its position with respect to the placed concrete;
a photographing means driving means for causing the photographing means to move up and down and change its position with respect to the concrete to be compacted;
Supervised learning is performed using an image or a set of image feature values and known numerical sequences related to the degree of compaction of concrete as learning data, and based on the images acquired from the photographing means, the degree of compaction of concrete is determined. an image feature amount calculation unit that calculates a numerical sequence that is a feature amount;
a learning unit that performs supervised learning based on the numerical sequence calculated by the image feature amount calculation unit and derives a predicted value of a numerical value representing the degree of compaction of the concrete ;
A compaction determination unit that calculates the predicted value from the learning unit and performs compaction determination;
an operation control unit that drives and controls the vibrator driving means and the photographing means driving means;
Optimal compaction determination construction system for concrete.

Images