Precise lossless identification method for real concentration of filling body

Abstract

The invention discloses an accurate lossless identification method for the real concentration of a filling body, belongs to the technical field of filling, and solves the problem that the real concentration of the filling body cannot be accurately measured after a goaf is maintained for a certain days. The method comprises the following steps: performing a hyperspectral scanning experiment on a test block by using a hyperspectral imager, extracting a surface spectral curve of the test block, collecting a modeling data set, dividing the modeling data set into a training set and a test set, performing iterative training on a concentration prediction model, loading a surface spectral curve of a filling body, executing the concentration prediction model, and outputting the concentration of the filling body. According to the method, the real concentration of the filling body is identified by utilizing a hyperspectral imaging technology, a sample does not need to be damaged, the method has the advantage of nondestructive detection, after the filling body concentration identification model is established, the corresponding concentration can be obtained after a test block with unknown concentration is scanned, the experiment detection period is greatly shortened, and the method has high efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of filling, and particularly relates to a method for accurately and non-destructively identifying the true concentration of a filling body. Background Technique

[0002] The full-tailings cemented filling body is a technology that mixes solid waste, cementitious materials, and water evenly to form a filling slurry, and then transports it to the goaf to control ground pressure. This technology solves the two major hazards of solid waste treatment and goaf collapse and is the main development trend in the mining field.

[0003] The mass concentration of the filling body is a key factor affecting its performance. The higher the concentration of the filling body, the higher the strength and denser the microstructure usually shown. During the preparation process of the filling body, its concentration can be accurately controlled and measured. However, once the filling body is transported to the goaf, affected by factors such as groundwater inflow or segregation, the actual concentration often deviates from the expected concentration. This deviation is one of the key reasons for the deviation of the filling body performance from the design standard. For the solidified filling body, it is very difficult to distinguish or identify the concentration through visual inspection or existing equipment. For example, in mining operations, the quality of the filling body is usually monitored by core sampling, and then the uniaxial compressive strength test is carried out to evaluate its strength. However, the concentration of the filling body is still unknown, and it is also impossible to determine whether the strength difference is due to the change in concentration. As is well known, it is easy to measure the concentration of fresh filling slurry, but the true concentration of the filling body after a certain number of days of curing in the goaf cannot be accurately determined. Therefore, quickly and non-destructively identifying the actual concentration of the filling body is crucial for accurately controlling the filling parameters and realizing intelligent filling. In view of the above problems, we propose a method for accurately and non-destructively identifying the true concentration of a filling body. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for accurately and non-destructively identifying the true concentration of a filling body in view of the deficiencies of the prior art, and solve the problem that the true concentration of the filling body after a certain number of days of curing in the goaf cannot be accurately determined.

[0005] In the prior art, it is easy to measure the concentration of fresh filling slurry, but the true concentration of the filling body after a certain number of days of goaf maintenance cannot be accurately measured. In view of the above problems, we propose a method for accurately and non-destructively identifying the true concentration of the filling body. When implementing the method, first measure tailings, cementitious materials and water according to the mixing ratio scheme, stir until uniform to obtain test blocks, take out the test blocks after curing for the curing age, conduct hyperspectral scanning experiments on the test blocks using a hyperspectral imager, and extract the spectral curve on the surface of the test blocks; at the same time, use the spectral reflectance as the input and the concentration of the test block as the output to train the algorithm, iteratively train the concentration prediction model, and output a converged concentration prediction model. Then, obtain a filling body with an unknown concentration, conduct a hyperspectral scanning experiment on the filling body using a hyperspectral imager, extract the spectral curve on the surface of the filling body, and finally use the spectral curve on the surface of the filling body as the input, execute the concentration prediction model, and output the concentration of the filling body. The embodiment of the present invention uses hyperspectral imaging technology to identify the true concentration of the filling body, without damaging the sample, having the advantage of non-destructive testing. After the filling body concentration identification model is established, the corresponding concentration can be obtained after scanning a test block with an unknown concentration, greatly shortening the experimental detection cycle and having high efficiency. Moreover, the present invention has good universality and scalability, and is applicable to all mine filling systems.

[0006] The present invention is implemented as follows. A method for accurately and non-destructively identifying the true concentration of a filling body, S10, measure tailings, cementitious materials and water according to the mixing ratio scheme, stir until uniform to obtain test blocks, and perform curing treatment on the test blocks;

[0007] S20, take out the test blocks after curing for the curing age, conduct a hyperspectral scanning experiment on the test blocks using a hyperspectral imager, perform reflectance calibration and preprocessing on the spectral data, further divide the region of interest, and extract the spectral curve on the surface of the test blocks;

[0008] S30, collect a modeling data set, divide the modeling data set into a training set and a test set, use the spectral reflectance as the input and the concentration of the test block as the output to train the algorithm, iteratively train the concentration prediction model, and output a converged concentration prediction model;

[0009] S40, obtain a filling body with an unknown concentration, conduct a hyperspectral scanning experiment on the filling body using a hyperspectral imager, perform reflectance calibration and preprocessing on the spectral data, further divide the region of interest, and extract the spectral curve on the surface of the filling body;

[0010] S50, load the spectral curve on the surface of the filling body, use the spectral curve on the surface of the filling body as the input, execute the concentration prediction model, and output the concentration of the filling body.

[0011] The variables in the mixing ratio scheme include concentration, ash-sand ratio, water-cement ratio, and cementitious material ratio.

[0012] The hyperspectral imager is not less than 10 -1λ, the number of channels is not less than 100.

[0013] In step S20, the curing age of the test block is not less than 28 days. During the curing age, the number of scans for a single test block is not less than 5 times, and 2 - 4 surfaces of each test block are scanned. Each test block surface is divided into not less than 100 regions of interest. The reflectivity of each region of interest is obtained from the average reflectivity values of not less than 20 pixels, and the spectral curve of each filling body is the average reflectivity of all regions of interest.

[0014] In step S30, the proportion of the test set is not less than 20% of the total data set, and the accuracy of the selected concentration prediction model in the test set is not less than 85%.

[0015] In step S40, the filling bodies with unknown concentrations include laboratory - prepared and stope - cored test blocks. If the filling body is soaked by water, it is air - dried for 2 - 6 hours before hyperspectral scanning.

[0016] Compared with the prior art, the embodiments of the present application mainly have the following beneficial effects:

[0017] The embodiments of the present invention use hyperspectral imaging technology to identify the true concentration of the filling body without damaging the sample, having the advantage of non - destructive testing. After the filling body concentration identification model is established, the corresponding concentration can be obtained after scanning the test block with unknown concentration, greatly shortening the experimental detection period and having high efficiency. Moreover, the present invention has good universality and popularizability and is applicable to all mine filling systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a schematic flowchart of the implementation of the method for accurately and non - destructively identifying the true concentration of the filling body provided by the present invention.

[0019] Figure 2 shows the hyperparameter optimization result diagram in the concentration prediction model. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs; the terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above - mentioned drawings are intended to cover non - exclusive inclusion. The terms "first", "second", etc. in the specification and claims of this application or the above - mentioned drawings are used to distinguish different objects and are not used to describe a specific order.

[0021] References to "embodiments" in this specification mean that the specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of the present application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0022] In the prior art, it is easy to measure the concentration of fresh filling slurry, but it is impossible to accurately measure the true concentration of the filling body after a certain number of days of goaf maintenance. In view of the above problems, we propose a method for accurately and non-destructively identifying the true concentration of the filling body. When implementing the method, first measure tailings, cementitious materials, and water according to the mixing ratio scheme, stir until uniform to obtain test blocks, take out the test blocks after curing for the curing age, perform hyperspectral scanning experiments on the test blocks using a hyperspectral imager, and extract the spectral curve on the surface of the test blocks. At the same time, use the spectral reflectance as the input and the test block concentration as the output to train the algorithm, iteratively train the concentration prediction model, and output a converged concentration prediction model. Then, obtain the filling body with an unknown concentration, perform hyperspectral scanning experiments on the filling body using a hyperspectral imager, extract the spectral curve on the surface of the filling body, and finally use the spectral curve on the surface of the filling body as the input, execute the concentration prediction model, and output the filling body concentration. The embodiments of the present invention use hyperspectral imaging technology to identify the true concentration of the filling body, without damaging the sample, having the advantage of non-destructive testing. After the filling body concentration identification model is established, the corresponding concentration can be obtained after scanning the test block with an unknown concentration, greatly shortening the experimental detection cycle and having high efficiency. Moreover, the present invention has good universality and scalability, and is applicable to all mine filling systems.

[0023] The embodiments of the present invention provide a method for accurately and non-destructively identifying the true concentration of the filling body, Figure 1 shows a schematic implementation flow diagram of the method for accurately and non-destructively identifying the true concentration of the filling body. The method for accurately and non-destructively identifying the true concentration of the filling body specifically includes:

[0024] S10, measure tailings, cementitious materials, and water according to the mixing ratio scheme, stir until uniform to obtain test blocks, and perform curing treatment on the test blocks;

[0025] It should be noted that the variables in the mixing ratio scheme include concentration, ash-sand ratio, water-cement ratio, cementitious material ratio, etc. The concentration range and gradient are adjusted according to mine production. When preparing the test block mixture, generally refer to the current industry standard "Standard for Test Methods of Basic Properties of Building Mortars" JGJ / T 70.

[0026] S20, take out the test blocks after curing for the curing age, perform hyperspectral scanning experiments on the test blocks using a hyperspectral imager, perform reflectance calibration and preprocessing on the spectral data, further divide the region of interest, and extract the spectral curve on the surface of the test blocks;

[0027] In this embodiment, the hyperspectral imager is not less than 10 -1 λ, the number of channels is not less than 100, the curing age of the test block is not less than 28 days, the number of scans of a single test block during the curing age is not less than 5 times, each test block is scanned on 2-4 surfaces, each test block surface is divided into not less than 100 regions of interest, the reflectance of each region of interest is obtained from the average reflectance values of not less than 20 pixels, and the spectral curve of each filling body is the average reflectance of all regions of interest.

[0028] It should be noted that taking a domestic metal mine as the background, the set ratio of cement to tailings is 1:4, and the mass concentration gradients are set to 61%, 64%, 67%, 70%, and 73% respectively. According to this mixing ratio scheme, the weighed tailings, cement, and water are poured into the mixing bucket and stirred until evenly mixed, and then the filling slurry is poured into a mold of 7.07×7.07×7.07 cm. After waiting for 24 hours, demolding is carried out, and immediately the test block is placed in a standard curing box for curing. When the curing age reaches 3 days, 7 days, 11 days, 14 days, 17 days, 21 days, 25 days, and 28 days, spectral scanning experiments are carried out.

[0029] The resolution of the hyperspectral imager is 7 nm, and it consists of 256 bands in total. The whole instrument is placed in a closed black box to prevent external light from affecting the experiment. 4 halogen lamps are installed inside the black box as light sources. The Specview software is used to control the spectrometer, including the focal length adjustment and exposure adjustment of the camera. Before testing, it is calibrated with a 100% reflectance whiteboard, considering that the absorption of all bands is 0. In order to be displayed at the center of the image during imaging, the filling body test block is placed directly below the hyperspectral camera, and the test blocks are scanned one by one according to the numbers.

[0030] The scanned files are imported into the Specview software for pixel-by-pixel reflectance calibration. At the same time, spectral smoothing processing is carried out to reduce the influence of noise and spiky spectra. The ENVI software is used to extract the spectra of the corrected spectra. Taking the entire surface of the filling body as the research object, 5×5 = 25 pixel squares are selected as regions of interest. A single region of interest is regarded as a whole, and the average spectrum is calculated as the spectral value of this region. In order to reduce errors, about 200 regions of interest are extracted from each filling body surface, and the average of these 200 spectra is used as the high-light curve of this test block. To remove specific noise or redundant information and extract a more pure spectral signal, local maximum points are found in the data, and the maximum points are connected to form an upper envelope line. The original data is divided by the upper envelope line to obtain the reflectance value of each band.

[0031] S30. Collect the modeling data set, divide the modeling data set into a training set and a test set, use the spectral reflectance as the input and the concentration of the test block as the output to train the algorithm, iteratively train the concentration prediction model, select the algorithm with the best accuracy and robustness as the concentration prediction model, and output the converged concentration prediction model;

[0032] It should be noted that the proportion of the test set is not less than 20% of the total data set, and the accuracy of the selected concentration prediction model in the test set is not less than 85%.

[0033] S40. Obtain the filling body with unknown concentration, perform a hyperspectral scanning experiment on the filling body using a hyperspectral imager, perform reflectance calibration and preprocessing on the spectral data, further divide the region of interest, and extract the spectral curve of the filling body surface;

[0034] In the embodiment of the present invention, the filling body with unknown concentration includes laboratory-prepared and stope-core-taking test blocks. If the filling body is soaked in water, it is air-dried for 2-6 hours before hyperspectral scanning.

[0035] S50. Load the spectral curve of the filling body surface, use the spectral curve of the filling body surface as the input, execute the concentration prediction model, and output the concentration of the filling body.

[0036] The embodiment of the present invention uses hyperspectral imaging technology to identify the true concentration of the filling body, without damaging the sample, having the advantage of non-destructive testing. After the concentration identification model of the filling body is established, the corresponding concentration can be obtained after scanning the test block with unknown concentration, greatly shortening the experimental detection cycle and having high efficiency. Moreover, the present invention has good universality and popularizability and is applicable to all mine filling systems.

[0037] It should be noted that when training the concentration prediction model, the modeling data set is divided into a 70% training set and a 30% test set, and the support vector machine algorithm is used for modeling. In the support vector machine algorithm, the adjustment of hyperparameters is crucial for the performance of the model. The hyperparameters of the concentration prediction model include C and Gamma. In this embodiment, the value range of C is (0.1, 1.0, 10, 100, 200, 500), and the value range of Gamma is (0.001, 0.01, 0.1, 1, 10). The grid optimization algorithm combined with 5-fold cross-validation is used to optimize the hyperparameters. Figure 2 The figure showing the optimization results of the hyperparameters in the concentration prediction model is shown. When the hyperparameters C = 200 and Gamma = 10, the accuracy is the highest. The accuracies of the training set and the test set are 99.31% and 98.23% respectively.

[0038] In summary, the present invention provides a method for accurately and non-destructively identifying the true concentration of backfill. In the embodiments of the present invention, hyperspectral imaging technology is used to identify the true concentration of backfill, without the need to damage the sample, which has the advantage of non-destructive testing. After the backfill concentration identification model is established, the corresponding concentration can be obtained by scanning the test block with unknown concentration, greatly shortening the experimental detection period and having high efficiency. Moreover, the present invention has good universality and promotability and is applicable to all mine backfill systems.

[0039] It should be noted that for the foregoing embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, some steps may be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0040] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the protection scope of the invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on these embodiments, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art can still, without conflict, make combinations, additions, deletions or other adjustments to the features in the embodiments of the present invention according to the situation without creative efforts, so as to obtain different technical solutions that are essentially not divorced from the concept of the present invention, and these technical solutions also belong to the scope of protection of the present invention.

Claims

1. A method for accurately and non-destructively identifying the true concentration of a filling body, characterized in that: include: S10, measuring tailings, cementing materials and water according to the mixing ratio scheme, stirring until uniform, obtaining test blocks, and performing curing treatment on the test blocks; S20, taking out the test block after curing for the curing age, performing a hyperspectral scanning experiment on the test block using a hyperspectral imager, calibrating and preprocessing the spectral data for reflectance, further dividing the region of interest, and extracting the spectral curve of the test block surface; S30, collecting a modeling data set, dividing the modeling data set into a training set and a test set, using spectral reflectance as input and test block concentration as output to train an algorithm, iteratively training a concentration prediction model, and outputting a converged concentration prediction model; S40, obtaining a filling body of unknown concentration, performing a hyperspectral scanning experiment on the filling body using a hyperspectral imager, calibrating and preprocessing the spectral data for reflectance, further dividing the region of interest, and extracting a spectral curve of the filling body surface; S50, loading the filling body surface spectral curve, taking the filling body surface spectral curve as input, executing the concentration prediction model, and outputting the filling body concentration.

2. The method for accurately and non-destructively identifying the true concentration of a filling body according to claim 1, characterized in that: The variables in the mix ratio include concentration, cement-sand ratio, water-cement ratio, and binder ratio.

3. The method for accurately and non-destructively identifying the true concentration of a filling body according to claim 2, characterized in that: The hyperspectral imager is not less than 10 -1 λ, the number of channels is not less than 100.

4. The method for accurately and non-destructively identifying the true concentration of a filling body according to claim 1, characterized in that: In step S20, the curing age of the test block is not less than 28 days, the single test block is scanned no less than 5 times during the curing age, and each test block is scanned on 2-4 surfaces. The surface of each test block is divided into no less than 100 regions of interest, and the reflectivity of each region of interest is obtained by the average reflectivity value of no less than 20 pixels. The spectral curve of each filling body is the average reflectivity of all regions of interest.

5. The method for accurately and non-destructively identifying the true concentration of a filling body according to claim 1, characterized in that: In step S30, the test set accounts for no less than 20% of the total data set, and the accuracy of the selected concentration prediction model in the test set is no less than 85%.

6. The method for accurately and non-destructively identifying the true concentration of a filling body according to any one of claims 2 to 5, characterized in that: In step S40, the filling body of unknown concentration includes laboratory preparation and field core test blocks. If the filling body is soaked in water, the filling body is air-dried for 2-6 hours before high-spectral scanning.

Images