Intelligent evaluation system for vibration quality of newly poured concrete

Abstract

The invention discloses an intelligent evaluation system for vibration quality of newly poured concrete. The intelligent evaluation system comprises a vibration energy distribution calculation module, a surface image classification and recognition module and a data fusion module, the vibration energy distribution calculation module collects and transmits multi-sensor data, and performs theoretical calculation of vibration energy distribution according to the sensing data; the surface image classification and recognition module collects a surface image of vibrated concrete and inputs the surface image into a convolutional neural network, and classification and judgment of vibration quality are carried out according to the image; the data fusion module comprises a grid division step, a vibration energy classification interval calibration step and a correction step based on relative entropy, vibration energy data and image judgment data are combined through formula calculation, and a final concrete vibration quality judgment result is obtained. Through fusion of multi-sensor data and image data, the real-time performance and accuracy of vibration quality evaluation are remarkably improved, and the method is suitable for the fields of constructional engineering, bridge construction and the like.

Description

Technical Field

[0001] The invention relates to the technical field of concrete vibration, and in particular to an intelligent evaluation system for the vibration quality of newly poured concrete. Background Art

[0002] Concrete is one of the most commonly used materials in infrastructure construction. With the rapid development of the construction industry, concrete usage has not only increased significantly, but also placed higher demands on its quality in new high-rise buildings, bridges, dams, and other large-scale projects. Vibration is a critical construction step in concrete pouring. Good vibration quality is crucial for ensuring concrete strength and component performance. Insufficient vibration can lead to honeycombs and voids in the hardened concrete, while excessive vibration can cause segregation and uneven aggregate distribution. Repairing these defects is time-consuming and labor-intensive, and improperly handled can easily lead to safety accidents. However, there is currently no reliable theory or system for evaluating the vibration quality of newly poured concrete. In actual construction, determining the appropriate vibration duration still largely depends on the experience of the construction workers. Construction workers subjectively select the insertion position of the vibrator based on its effective range, ensuring staggered coverage of the vibration range as much as possible. This manual visual inspection method is inherently subjective and arbitrary. During long construction periods, it is difficult for workers to accurately control the vibration position, spacing, and duration. At the same time, the high reliance on manual labor for vibration operations also makes it difficult to quantitatively evaluate vibration quality, and quality issues such as under-vibration, over-vibration, or missed vibration cannot be detected in real time. Therefore, proposing a rational monitoring method for the vibration quality of freshly poured concrete has considerable practical engineering application value.

[0003] Currently, several intelligent technologies and methods are available for assessing the vibration quality of freshly poured concrete, including infrared thermal imaging, satellite positioning, and computer vision. However, these methods lack quantitative analysis of the concrete itself, and the information density and real-time nature of the information used to infer vibration quality are insufficient. This results in limited usability and accuracy for assessing the vibration quality of freshly poured concrete.

[0004] To this end, the present invention proposes an intelligent assessment system for the vibration quality of newly poured concrete. Summary of the Invention

[0005] The purpose of the present invention is to address the shortcomings of the existing technology and provide an intelligent assessment system for the vibration quality of newly poured concrete. This method achieves real-time and accurate assessment of concrete vibration quality through the collaborative work of multiple modules, including a vibration energy distribution calculation module, a surface image classification and recognition module, and a data fusion module. First, the vibration energy distribution calculation module collects and transmits multi-sensor data, and performs theoretical calculations of vibration energy distribution based on the sensor data; second, the surface image classification and recognition module collects surface images of the vibrated concrete and inputs them into a convolutional neural network, classifying and determining the vibration quality based on the images; finally, the data fusion module combines the vibration energy data and image determination data through formula calculation to obtain the final concrete vibration quality determination result.

[0006] To achieve the above-mentioned object, the present invention provides the following technical solution: an intelligent assessment system for the vibration quality of newly poured concrete, comprising:

[0007] The vibration energy distribution calculation module is used to collect and transmit multi-sensor data and perform theoretical distribution calculations of vibration energy distribution based on the sensor data. It includes: a two-dimensional plane positioning sensor submodule system, an acceleration sensor submodule system, and an energy calculation model. The two-dimensional plane positioning sensor submodule system is used to measure the two-dimensional plane position of the vibrator in real time during the vibration process; the acceleration sensor submodule system is used to measure the vibration acceleration of the vibrator itself and the interior of the concrete in real time during the vibration process. The energy calculation model is used to process the obtained sensor data according to a theoretical formula to calculate the vibration energy distribution of the vibrated concrete, and includes: a data conversion part and a formula calculation part.

[0008] A surface image classification and recognition module is configured to capture surface images of vibrated concrete and input them into a convolutional neural network, whereby vibration quality is classified and determined based on the images. The module comprises a camera and a convolutional neural network model. The camera is attached to the vibrator and is configured to capture horizontal images of the surface of the vibrated concrete. The convolutional neural network model is configured to classify the surface images of the vibrated concrete captured by the camera according to vibration quality.

[0009] The data fusion module is used to combine the vibration energy data and the image judgment data through formula calculation to obtain the final concrete vibration quality judgment result; it includes: a grid division step submodule, a vibration energy classification interval calibration submodule step, and a relative entropy-based correction submodule step; the grid division submodule step is used to divide the vibrated concrete area into several grids, and correspond the energy calculation results of the vibration energy distribution calculation module to specific grids; the vibration energy classification interval calibration submodule step is used to establish a mapping relationship between the vibration energy value and the vibration quality classification, and calibrate the interval range for each vibration quality classification of the vibration energy; the relative entropy-based correction submodule step is used to correct the vibration energy value through the image judgment result to obtain a more accurate real-time global vibration quality judgment result.

[0010] Preferably, the two-dimensional plane positioning sensor submodule system is composed of base stations, tags and a local host, the base stations are arranged around the vibrated concrete area, the tags are attached to the vibrator, and the local host is wirelessly connected to each base station and tag; the relative position of each base station is measured and input into the local host to complete the initialization of the two-dimensional plane positioning sensor submodule system; the acceleration sensor submodule system is composed of acceleration sensors and a local host, the acceleration sensors are respectively buried in the concrete or attached to and attached to the vibrator, and the local host is wirelessly connected to each acceleration sensor; the energy calculation model first organizes and summarizes the sensor data through the data conversion part, and then establishes the vibration energy distribution through the formula calculation part;

[0011] The data conversion section converts and integrates the data files recorded in real time by the sensor systems of the two-dimensional plane positioning sensor submodule and the acceleration sensor submodule, ultimately obtaining a file recording real-time two-dimensional plane positioning information, vibration acceleration information, and vibration frequency information of the vibrator. The file is input into the formula calculation section. The formula calculation section has pre-divided the concrete area that can be covered by the vibrator into several grids. According to the information recorded in the file and in combination with the model parameters, the concrete vibration energy of each grid is calculated according to the calculation formula; the calculation formula is:

[0012]

[0013] Among them, c1, ξ b ,ξ m is the concrete material parameter, ξ b is the boundary damping coefficient, ξ mis the material damping coefficient, c1 is an intermediate quantity determined by the concrete stiffness and damping, and both are model parameters; l is the center distance of the vibrator working position in the grid, obtained based on the two-dimensional plane positioning information of the vibrator; a is the vibration acceleration of the vibrator, obtained based on the vibration acceleration information; f is the operating frequency of the vibrator, obtained based on the vibration frequency information;

[0014] The process of measuring model parameters is as follows: insert nails evenly into the concrete, observe the nail that is affected the farthest by the vibrator and sink it, and thus determine the vibrator's influence radius R; according to the data of two acceleration sensors buried in the concrete, l , combine the two calculation formulas to solve ξ b ,ξ m , the calculation formula is:

[0015]

[0016] By comparing the vibrator in its incomplete insertion state, no-load state, and complete insertion state, c1 is obtained using a calculation formula; the calculation formula is:

[0017]

[0018] Where f′ is the vibration frequency in the incomplete insertion state; P e ' is the input power in the incomplete insertion state; P e0 is the input power in the no-load state; ρ is the concrete density; h is the vibrator insertion depth; a' is the vibration acceleration in the incomplete insertion state; r is the vibrator radius; R' is the influence radius in the incomplete insertion state.

[0019] Preferably, the lens plane of the camera is parallel to the surface of the vibrated concrete, and the camera captures images of the concrete surface at certain time intervals. The images of the concrete surface are transmitted by the camera to a local host for storage. The convolutional neural network model is loaded into the local host and, after being trained using a dataset of concrete vibration quality images, can provide a vibration quality classification judgment based on the input concrete surface images.

[0020] The convolutional neural network model training process is as follows: carry out several concrete vibrations, use the camera to capture multiple images of the concrete surface, and divide the images into grids, write labels for each divided image according to its actual vibration quality, and the images and their labels together constitute the concrete vibration quality image dataset; divide the dataset into a training set and a validation set; input the training set into the convolutional neural network model for training; and optimize the parameters of the convolutional neural network model using the validation set.

[0021] Preferably, the grid division submodule establishes a plane rectangular coordinate system with a corner point of the vibrated concrete area as the coordinate origin, and divides the vibrated concrete area into a plurality of grids of equal size with a fixed side length; the vibrated concrete grid is associated with the nearest vibrator coverage grid according to the distance, so that each vibration energy value obtained by the energy calculation model is sequentially mapped to the corresponding concrete grid;

[0022] The vibration energy classification interval calibration submodule uses support vector machine technology to establish a mapping relationship between vibration energy values ​​and vibration quality classifications, that is, the vibration energy-oriented vibration quality classification calibration interval range; carry out several concrete vibrations, use the camera to capture multiple concrete surface images and divide them into grids, and input them into the trained convolutional neural network model for vibration quality classification prediction; obtain the vibration energy value calculation results corresponding to each image from the energy calculation model as feature values, and use the image classification prediction results as labels to form several groups of one-dimensional data; input the one-dimensional data into the support vector machine to complete the calibration of the vibration energy classification interval, which is essentially the optimal classification boundary search for one-dimensional data under supervision; the support vector machine uses soft intervals, while looking for a decision boundary that can maximize the interval between different categories, allowing some points to be on the wrong side of the decision boundary, and introducing a penalty parameter to set the cost of misclassification;

[0023] The correction submodule based on relative entropy uses the image-based vibration quality judgment result to correct the vibration energy value of the local vibrated area around the vibrator insertion point, thereby obtaining a more accurate vibration quality judgment result; the random variable X is defined as the vibration quality judgment result, and its possible values ​​are X={x1,x2,...,x n}, corresponding to each vibration quality judgment classification, where n is the number of vibration quality classifications; the vibration quality judgment results based on vibration energy values ​​and the vibration quality judgment results based on images are regarded as two different but similar random distribution modes, namely P(X) and Q(X); for the vibration quality judgment result distribution P(X) based on vibration energy values, a standard normal distribution is established with the actual vibration energy cumulative value of the grid as the mean μ, and combined with the results of the vibration energy classification interval calibration, the probability p(x) of each quality judgment classification can be obtained. i ); For the image-based vibration quality judgment result distribution Q(X), the activation function of the convolutional neural network can directly output the probability of each category as the input image sample category, so the p′(x i );

[0024] The relative entropy KL of the two vibration quality judgment results is calculated according to the calculation formula of relative entropy, which is:

[0025]

[0026] The vibration energy is corrected by introducing the image judgment result according to the value of relative entropy. The calculation formula of the correction is:

[0027] E′=(1-KL)×E+KL×E I

[0028] Where E is the vibration energy value before correction; E′ is the vibration energy value after correction; E I is the converted vibration energy of the image judgment result, which is the midpoint value of the corresponding interval;

[0029] By determining the vibration quality classification corresponding to E′ based on the result of the vibration energy classification interval calibration, the vibration quality of the concrete grid corresponding to the current vibration energy can be obtained, and then the quality assessment result of the complete vibrated concrete area can be obtained.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] 1. This invention proposes an intelligent assessment system for the vibration quality of newly poured concrete. This system uses a multi-sensor process to calculate the distribution of concrete vibration energy. By deploying two-dimensional positioning sensors and acceleration sensors to collect several parameters of the concrete's real-time working conditions, the system performs calculations based on energy relationships, ultimately determining the distribution of concrete vibration energy as a basis for assessing concrete vibration quality. This vibration energy calculation process, based on the concrete's own working parameters and using rigorous formula derivation, represents a quantitative analysis of the concrete itself, offering high interpretability and resistance to interference from accidental factors.

[0032] 2. This invention proposes an intelligent assessment system for the vibration quality of freshly poured concrete. This system uses concrete surface images to determine vibration quality. By inputting surface images captured in real time by a camera into a trained convolutional neural network model to perform vibration quality classification and prediction, it can efficiently perform real-time concrete quality assessment. The convolutional neural network model is trained and tested on a dataset of concrete vibration quality images, achieving speed and accuracy far exceeding traditional manual assessment methods. Compared to one-dimensional sensor data, image data has a higher information density, thus providing a higher confidence level in the context of real-time monitoring.

[0033] 3. The present invention proposes an intelligent assessment system for the vibration quality of newly poured concrete. Based on the numerical correction process of vibration energy of relative entropy theory, the energy data and image data are integrated with each other. The high precision and high information density characteristics of the image data are used to compensate for the possible errors of the energy data. The rapidity and real-time calculation of the energy data compensate for the shortcomings of the image data's insufficient coverage and slow update speed. Together, a more accurate real-time global vibration quality judgment result is obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A flowchart of an intelligent assessment system for the vibration quality of newly poured concrete provided by an embodiment of the present invention;

[0035] Figure 2 This is a system structure diagram of an intelligent assessment system for the vibration quality of newly poured concrete provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0037] The present invention proposes an intelligent assessment system for the vibration quality of newly poured concrete, which realizes real-time and accurate assessment of the concrete vibration quality. Figure 1 and Figure 2 In order to illustrate that the method of the present invention can accurately and effectively complete the vibration quality assessment of freshly poured concrete, an embodiment of the present invention will be given below.

[0038] See also Figures 1 to 2 The present invention provides an intelligent assessment system for the vibration quality of newly poured concrete, and the technical solution is as follows:

[0039] In the embodiment of this application, the system proposed in this invention is used to describe in detail the process of evaluating the vibration quality of newly poured concrete. In the embodiment of this application, the concrete vibration quality evaluation is aimed at the pouring of a concrete slab at a construction site. The total pouring volume of the concrete slab is 100×100×20 cubic centimeters. Figure 1 and Figure 2 The content describes in detail the vibration quality assessment process in the concrete area; Figure 1The workflow of the method proposed in the present invention includes: S10. Vibration energy distribution calculation module; S20. Surface image classification and recognition module; S30. Data fusion module. S10. Vibration energy distribution calculation module includes: S11. Arrange multiple sensors to collect real-time data during the concrete vibration process; S12. Measure model parameters through experiments and pre-input them into the formula calculation part; S13. Complete the format conversion of multi-sensor data and calculate the vibration energy according to the formula. S20. Surface image classification and recognition module includes: S21. Take the concrete surface image and input it into the convolutional neural network model; S22. Construct a concrete vibration quality image dataset for network training; S23. The convolutional neural network model gives a three-category vibration quality judgment based on the input image. S30. Data fusion module includes: S31. Establish a coordinate system for the concrete area and divide it into grids corresponding to vibration energy; S32. Use support vector machines to define intervals for vibration energy for each vibration quality classification; S33. Use relative entropy theory to correct the vibration energy value based on the image judgment results. Figure 2 The system structure proposed by the present invention includes: vibration energy distribution calculation module, surface image classification and recognition module and data fusion module; Figure 1 and Figure 2 The following describes the contents:

[0040] The present invention provides an intelligent assessment system for the vibration quality of newly poured concrete, which specifically includes:

[0041] The vibration energy distribution calculation module is used to collect and transmit multi-sensor data, and perform theoretical distribution calculation of vibration energy distribution based on the sensor data, corresponding to the above-mentioned step S10; it includes: a two-dimensional plane positioning sensor submodule, an acceleration sensor submodule and an energy calculation model; the two-dimensional plane positioning sensor submodule is used to measure the two-dimensional plane position of the vibrator in real time during the vibration process, and the acceleration sensor submodule is used to measure the vibration acceleration of the vibrator itself and the concrete in real time during the vibration process, corresponding to the above-mentioned step S11; the energy calculation model is used to calculate the vibration energy distribution of the vibrated concrete by processing the various sensor data obtained according to the theoretical formula, and includes: a data conversion part and a formula calculation part.

[0042] Specifically, the two-dimensional plane positioning sensor submodule consists of four base stations, a tag and a local host. The base stations are arranged around the vibrated concrete area and form the four corners of a rectangle. The tag is attached to the vibrator. The local host is wirelessly connected to each base station and tag. The relative position of each base station is measured and input into the local host to complete the initialization of the two-dimensional plane positioning sensor submodule. The acceleration sensor submodule consists of three acceleration sensors and a local host. Two of the acceleration sensors are buried in the concrete, and one of the acceleration sensors is attached to the vibrator. The local host is wirelessly connected to each sensor.

[0043] In the embodiment of the present application, the energy calculation model first organizes and summarizes the sensor data through the data conversion part, and then establishes the vibration energy distribution through the formula calculation part. The data conversion part converts and integrates the data files recorded in real time by the two-dimensional plane positioning sensor submodule and the acceleration sensor submodule, and finally obtains a csv file, which records the real-time two-dimensional plane positioning information, vibration acceleration information and vibration frequency information of the vibrator; the csv file is input into the formula calculation part; the formula calculation part has pre-divided the concrete area that the vibrator can cover into several square grids with a side length of 5 cm, and according to the information recorded in the csv file, combined with the model parameters, the concrete vibration energy calculation of each grid is completed according to the calculation formula; the calculation formula is:

[0044]

[0045] Among them, c1, ξ b ,ξ m is a concrete material parameter, which belongs to the model parameter; l is the center distance of the vibrator working position in the grid, which is obtained based on the two-dimensional plane positioning information of the vibrator; a is the vibration acceleration of the vibrator, which is obtained based on the vibration acceleration information; f is the working frequency of the vibrator, which is obtained based on the vibration frequency information;

[0046] The specific model parameter measurement process is as follows: insert nails evenly into the concrete, observe the nail that the vibrator can affect and sink the farthest, and thus determine the vibrator's impact radius R; according to the data of two acceleration sensors buried in the concrete a l , combine the two calculation formulas to solve ξ b ,ξ m , the calculation formula is:

[0047]

[0048] By comparing the vibrator in its incomplete insertion state, no-load state, and complete insertion state, c1 is obtained using a calculation formula; the calculation formula is:

[0049]

[0050] Where f′ is the vibration frequency in the incomplete insertion state; P e ' is the input power in the incomplete insertion state; P e0 is the input power in the no-load state; ρ is the concrete density; h is the vibrator insertion depth; a' is the vibration acceleration in the incomplete insertion state; r is the vibrator radius; R' is the influence radius in the incomplete insertion state.

[0051] In this embodiment, R is 0.5m, ξ b is 0.7699, ξ m is 0.0296, c1 is 423138.2, and f is 200Hz.

[0052] Preferably, the surface image classification and recognition module is used to collect surface images of vibrated concrete and input them into a convolutional neural network model, and classify and determine the vibration quality based on the images, corresponding to the above-mentioned step S20; and includes: a camera and a convolutional neural network model; the camera is installed on the vibrator, with the lens plane parallel to the surface of the vibrated concrete, and captures concrete surface images at certain time intervals, and the concrete surface images are transmitted by the camera to a local host for storage, corresponding to the above-mentioned step S21; the convolutional neural network model is located in the local host, and after being trained with a concrete vibration quality image dataset, can give a three-category vibration quality judgment for the input concrete surface image, including: unqualified, medium, and qualified, corresponding to the above-mentioned step S23;

[0053] Specifically, the training process of the convolutional neural network model is as follows: carry out several concrete vibrations, use the camera to capture multiple images of the concrete surface, and divide them into several square grids with a side length of 5 cm, write a label for each divided image according to its actual vibration quality, and each image and its label together constitute the concrete vibration quality image dataset; divide the dataset into a training set and a validation set, input the training set into the convolutional neural network model for training, and optimize the parameters of the convolutional neural network model through the validation set, corresponding to the above-mentioned step S22.

[0054] Table 1 shows the classification of the concrete vibration quality image dataset according to vibration quality, as well as the division of the training set and the validation set.

[0055] Table 1 Classification and division of concrete vibration quality image dataset

[0056] Number of images Unqualified medium qualified total training set 1390 2235 1153 4778 Validation set 347 559 288 1194 total 1737 2794 1441 5972

[0057] Table 2 shows the vibration quality classification results of each convolutional neural network model on the same concrete vibration quality image dataset.

[0058] Table 2 Comparison of vibration quality classification of each convolutional neural network model

[0059]

[0060]

[0061] The comparison results of the vibration quality classification of each convolutional neural network model shown in Table 2 show that ShuffleNetV2 has the advantages of high classification accuracy, short classification time, and lighter weight. The optimal convolutional neural network model is ShuffleNetV2; the subsequent data fusion module will use the classification results of the optimal model.

[0062] Preferably, the data fusion module is used to combine the vibration energy data and the image judgment data through formula calculation to obtain the final concrete vibration quality judgment result, corresponding to the above-mentioned step S30; it includes: a grid division submodule, a vibration energy classification interval calibration submodule, and a relative entropy-based correction submodule; the grid division submodule is used to divide the vibrated concrete area into several square grids with a side length of 5 cm, and the energy calculation results of the vibration energy distribution calculation module are corresponded to specific grids, corresponding to the above-mentioned step S31; the vibration energy classification interval calibration submodule is used to establish a mapping relationship between the vibration energy value and the vibration quality classification, and calibrate the interval range for each vibration quality classification of the vibration energy, corresponding to the above-mentioned step S32; the relative entropy-based correction submodule is used to correct the vibration energy value through the image judgment result to obtain a more accurate real-time global vibration quality judgment result, corresponding to the above-mentioned step S33.

[0063] Specifically, the grid division submodule establishes a plane rectangular coordinate system with the lower left corner of the vibrated concrete area as the coordinate origin, and divides the vibrated concrete area into a number of equally sized grids with a fixed side length of 5 cm. The vibrated concrete grids are associated with the nearest grid within the vibrator coverage range based on their distance, thereby sequentially mapping each vibration energy value obtained by the energy calculation model to a specific concrete grid.

[0064] In an embodiment of the present application, the vibration energy classification interval calibration submodule uses support vector machine technology to establish a mapping relationship between vibration energy values ​​and vibration quality classifications, that is, the vibration energy calibration interval range for three vibration quality classifications; several concrete vibrations are carried out, and the camera is used to capture multiple concrete surface images and divide them into grids, and the images are input into the trained convolutional neural network model for vibration quality classification prediction; the vibration energy value calculation results corresponding to each image are taken from the energy calculation model as feature values, and the classification prediction results of the images are used as labels to form several groups of one-dimensional data; the one-dimensional data is input into the support vector machine to complete the calibration of the vibration energy classification interval, which is essentially the optimal classification boundary search for one-dimensional data under supervision; the support vector machine uses soft intervals, and while looking for a decision boundary that maximizes the interval between different categories, it allows some points to be on the wrong side of the decision boundary, and introduces a penalty parameter to set the cost of misclassification; please refer to Table 3, which shows the vibration energy interval demarcation results for three vibration quality classifications.

[0065] Table 3 Delineation of vibration energy intervals for three vibration quality classifications

[0066]

[0067] Preferably, the correction submodule based on relative entropy uses the image-based vibration quality judgment result to correct the vibration energy value of the local vibrated area around the vibrator insertion point, thereby obtaining a more accurate vibration quality judgment result; define the random variable X as the vibration quality judgment result, and its possible value is X={x1, x2, x3}, corresponding to the three vibration quality judgment results respectively; regard the vibration quality judgment result based on the vibration energy value and the image-based vibration quality judgment result as two different but similar random distribution modes, namely P(X) and Q(X); for the vibration quality judgment result distribution P(X) based on the vibration energy value, establish a standard normal distribution with the true vibration energy cumulative value of the grid as the mean μ, and combine the results of the vibration energy classification interval calibration to obtain the probability p(x) of each of the three quality judgment results. i ); For the image-based vibration quality judgment result distribution Q(X), the activation function of the convolutional neural network can directly output the probability of each category as the input image sample category, so the three quality judgment results p′(x i );

[0068] The relative entropy KL of the two vibration quality judgment results is calculated according to the calculation formula of relative entropy, which is:

[0069]

[0070] The vibration energy is corrected by introducing the image judgment result according to the value of relative entropy. The calculation formula of the correction is:

[0071] E′=(1-KL)×E+KL×E I

[0072] Where E is the vibration energy value before correction; E′ is the vibration energy value after correction; E I is the converted vibration energy of the image judgment result, which is the midpoint value of the corresponding interval;

[0073] By determining the vibration quality classification corresponding to E′ based on the results of the vibration energy classification interval calibration, the vibration quality of the concrete zone corresponding to the current vibration energy can be obtained, and then the quality assessment result of the complete vibrated concrete area can be obtained; by comparing the obtained vibration quality assessment result with the actual situation, the accuracy of the method proposed in the present invention can be evaluated.

[0074] Table 4 shows the accuracy evaluation results of the method proposed in the present invention when the vibration energy data is not corrected by the data fusion module.

[0075] Table 4 Accuracy evaluation of the method proposed in the present invention before correction

[0076]

[0077] Table 5 shows the accuracy evaluation results of the method proposed in the present invention when the vibration energy data is corrected by the data fusion module.

[0078] Table 5 Accuracy evaluation of the revised method proposed in the present invention

[0079]

[0080] A comparison of Tables 4 and 5 shows that the accuracy of the method before correction was 89.9%, with 10.1% of incorrect judgments, including 0.1% of large judgment errors, that is, qualified was mistakenly judged as unqualified, or unqualified was mistakenly judged as qualified; the accuracy of the method after correction was 99.3%, with only 0.7% of incorrect judgments, and no large judgment errors occurred. It can be seen that the data fusion module can greatly improve the accuracy of vibration quality judgment; the method proposed in the present invention can more accurately evaluate the vibration quality of freshly poured concrete.

[0081] The intelligent assessment system for the vibration quality of newly poured concrete proposed in the present invention realizes real-time and accurate concrete vibration quality assessment from multi-sensor data acquisition, surface image classification and recognition to multimodal data fusion; the vibration energy distribution calculation module collects several parameters of the real-time working conditions of concrete, calculates based on energy relationships, and obtains the distribution of concrete vibration energy; the surface image classification and recognition module captures image data of the concrete surface through a camera and inputs it into a convolutional neural network model to achieve classification prediction; the data fusion module corrects the theoretical calculation results based on the image classification results, obtains a more complete global vibration energy distribution through multimodal data fusion, and further infers the final vibration quality assessment result. The method proposed in the present invention quantitatively analyzes the concrete vibration process based on theoretical formulas, and the scheme has high interpretability. At the same time, by introducing image data to perform prediction, the information density of vibration quality inference is guaranteed, which can significantly improve the accuracy of vibration quality assessment of newly poured concrete, effectively reduce quality risks and post-construction repair costs, and can be applied to various newly poured concrete construction scenarios.

[0082] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent assessment system for the vibration quality of newly poured concrete, characterized in that: The system includes: vibration energy distribution calculation module, surface image classification and recognition module and data fusion module; The vibration energy distribution calculation module is used to collect and transmit multi-sensor data and perform theoretical distribution calculations of vibration energy distribution based on the sensor data. It includes: a two-dimensional plane positioning sensor submodule, an acceleration sensor submodule, and an energy calculation model. The two-dimensional plane positioning sensor submodule is used to measure the two-dimensional plane position of the vibrator in real time during the vibration process; the acceleration sensor submodule is used to measure the vibration acceleration of the vibrator itself and the interior of the concrete in real time during the vibration process. The energy calculation model is used to process the obtained sensor data according to a theoretical formula to calculate the vibration energy distribution of the vibrated concrete, and includes: a data conversion part and a formula calculation part. The surface image classification and recognition module is used to collect surface images of vibrated concrete and input them into a convolutional neural network to classify and determine the vibration quality based on the images. It includes a camera and a convolutional neural network model. The camera is installed on the vibrator and is used to horizontally capture surface images of the vibrated concrete. The convolutional neural network model is used to classify the surface images of the vibrated concrete captured by the camera according to the vibration quality. The data fusion module is used to combine the vibration energy data and the image judgment data through formula calculation to obtain the final concrete vibration quality judgment result; it includes: a grid division submodule, a vibration energy classification interval calibration submodule, and a relative entropy-based correction submodule; the grid division submodule is used to divide the vibrated concrete area into several grids, and correspond the energy calculation results of the vibration energy distribution calculation module to specific grids; the vibration energy classification interval calibration submodule is used to establish a mapping relationship between the vibration energy value and the vibration quality classification, and calibrate the interval range for each vibration quality classification; the relative entropy-based correction submodule is used to correct the vibration energy value through the image judgment result to obtain a more accurate real-time global vibration quality judgment result.

2. The intelligent assessment method for the vibration quality of newly poured concrete according to claim 1, characterized in that: The two-dimensional plane positioning sensor submodule consists of a base station, a tag and a local host. The base stations are arranged around the vibrated concrete area, the tags are attached to the vibrator, and the local host is connected to each base station and tag in a wireless manner; the relative position of each base station is measured and input into the local host to complete the initialization of the two-dimensional plane positioning sensor submodule.

3. The intelligent assessment method for the vibration quality of freshly poured concrete according to claim 1, characterized in that: The acceleration sensor submodule is composed of an acceleration sensor and a local host. The acceleration sensors are respectively embedded in concrete and attached to the vibrator. The local host is connected to each acceleration sensor in a wireless manner.

4. The intelligent assessment method for the vibration quality of freshly poured concrete according to claim 1, characterized in that: The energy calculation model first organizes and summarizes the sensor data through the data conversion part, and then establishes the vibration energy distribution through the formula calculation part; The data conversion section converts and integrates the data files recorded in real time by the two-dimensional plane positioning sensor submodule and the acceleration sensor submodule, ultimately obtaining a file recording real-time two-dimensional plane positioning information, vibration acceleration information, and vibration frequency information of the vibrator. The file is input into the formula calculation section. The formula calculation section has pre-divided the concrete area that can be covered by the vibrator into several grids. According to the information recorded in the file and in combination with the model parameters, the concrete vibration energy of each grid is calculated according to the calculation formula. The calculation formula is: Among them, c1, ξ b ,ξ m is the concrete material parameter, ξ b is the boundary damping coefficient, ξ m is the material damping coefficient, c1 is an intermediate quantity determined by the stiffness and damping of concrete, and both are model parameters; l is the center distance of the vibrator working position in the grid, which is obtained based on the two-dimensional plane positioning information of the vibrator; a is the vibration acceleration of the vibrator, which is obtained based on the vibration acceleration information; f is the operating frequency of the vibrator, which is obtained based on the vibration frequency information.

5. The intelligent assessment method for the vibration quality of freshly poured concrete according to claim 4, characterized in that: The process of measuring model parameters is as follows: insert nails evenly into the concrete, observe the nail that is affected the farthest by the vibrator and sink it, and thus determine the vibrator's influence radius R; according to the data of two acceleration sensors buried in the concrete, l , combine the two calculation formulas to solve ξ b ,ξ m , the calculation formula is: By comparing the vibrator in its incomplete insertion state, no-load state, and complete insertion state, c1 is obtained using a calculation formula; the calculation formula is: Where f′ is the vibration frequency in the incomplete insertion state; P e ' is the input power in the incomplete insertion state; P e0 is the input power in the no-load state; ρ is the concrete density; h is the vibrator insertion depth; a' is the vibration acceleration in the incomplete insertion state; r is the vibrator radius; R' is the influence radius in the incomplete insertion state.

6. The intelligent assessment method for the vibration quality of freshly poured concrete according to claim 1, characterized in that: The lens plane of the camera is parallel to the surface of the vibrated concrete, and the camera captures images of the concrete surface at certain time intervals. The concrete surface images are transmitted by the camera to a local host for storage. The convolutional neural network model is loaded into the local host and, after being trained using a concrete vibration quality image dataset, can give a vibration quality classification judgment based on the input concrete surface images.

7. The intelligent assessment method for the vibration quality of freshly poured concrete according to claim 6, characterized in that: The convolutional neural network model training process is as follows: carry out several concrete vibrations, use the camera to capture multiple concrete surface images, and divide the images into grids, write labels for each divided image according to its actual vibration quality, and each image and its label together constitute the concrete vibration quality image dataset; divide the dataset into a training set, a validation set, and a test set; input the training set into the convolutional neural network model for training; optimize the parameters of the convolutional neural network model through the validation set; and evaluate the accuracy of the convolutional neural network model after parameter optimization through the test set.

8. The intelligent assessment method for the vibration quality of freshly poured concrete according to claim 1, characterized in that: The grid division submodule establishes a plane rectangular coordinate system with a corner point of the vibrated concrete area as the coordinate origin, and divides the vibrated concrete area into several grids of equal size with a fixed side length; the vibrated concrete grid is associated with the nearest vibrator coverage area grid according to the distance, so that the various vibration energy values ​​obtained by the energy calculation model are mapped to the corresponding concrete grids in sequence.

9. The intelligent assessment method for the vibration quality of freshly poured concrete according to claim 1, characterized in that: The vibration energy classification interval calibration submodule uses support vector machine technology to establish a mapping relationship between vibration energy values ​​and vibration quality classifications, that is, the vibration energy-oriented vibration quality classification calibration interval range; several concrete vibrations are carried out, and multiple images of the concrete surface are captured using the camera and divided into grids, and the images are input into the trained convolutional neural network model for vibration quality classification prediction; The numerical calculation results of the vibration energy corresponding to each image are obtained from the energy calculation model as feature values, and the classification prediction results of the image are used as labels to form several groups of one-dimensional data; the one-dimensional data is input into the support vector machine to complete the calibration of the vibration energy classification interval, which is essentially the search for the optimal classification boundary of one-dimensional data under supervision; the support vector machine uses soft intervals, and while searching for the decision boundary that maximizes the interval between different categories, it allows some points to be on the wrong side of the decision boundary, and introduces a penalty parameter to set the cost of misclassification.

10. The intelligent assessment method for the vibration quality of freshly poured concrete according to claim 1, characterized in that: The relative entropy-based correction submodule uses the image-based vibration quality determination result to correct the vibration energy value of the local vibrated area around the vibrator insertion point, thereby obtaining a more accurate vibration quality determination result; Define the random variable X as the result of the vibration quality judgment, and its possible value is X={x1,x2,...,x n }, corresponding to each vibration quality judgment classification, where n is the number of vibration quality classifications; the vibration quality judgment results based on vibration energy values ​​and the vibration quality judgment results based on images are regarded as two different but similar random distribution modes, namely P(X) and Q(X); for the vibration quality judgment result distribution P(X) based on vibration energy values, a standard normal distribution is established with the actual vibration energy cumulative value of the grid as the mean μ, and combined with the results of the vibration energy classification interval calibration, the probability p(x) of each quality judgment classification can be obtained. i ); For the image-based vibration quality judgment result distribution Q(X), the activation function of the convolutional neural network can directly output the probability of each category as the input image sample category, so the p′(x i ); The relative entropy KL of the two vibration quality judgment results is calculated according to the calculation formula of relative entropy, which is: The vibration energy is corrected by introducing the image judgment result according to the value of relative entropy. The calculation formula of the correction is: E′=(1-KL)×E+KL×E I Where E is the vibration energy value before correction; E′ is the vibration energy value after correction; E I is the converted vibration energy of the image judgment result, which is the midpoint value of the corresponding interval; By determining the vibration quality classification corresponding to E′ based on the result of the vibration energy classification interval calibration, the vibration quality of the concrete grid corresponding to the current vibration energy can be obtained, and then the quality assessment result of the complete vibrated concrete area can be obtained.

Images