A filling slurry concentration monitoring system and method based on real-time spectral information

Abstract

The present invention provides a filling slurry concentration monitoring system and method based on real-time spectral information, which belongs to the field of mine filling technology. The system includes a visualization module, a stirring tank, a hyperspectral scanning device, an image processing module and a neural network learning module, wherein the hyperspectral scanning device is arranged above the stirring tank, and the hyperspectral scanning device collects spectral data in the filling slurry and transmits it to the image processing module to process the spectral image. The processed data is compared with the conventional measured data of the filling slurry spectral database in the neural network learning module to establish a connection between the spectral data and the concentration, thereby further obtaining a slurry concentration detection model, and finally realizing the filling slurry concentration monitoring. The system has a simple structure and a high degree of automation, and can realize non-contact measurement of the filling slurry concentration. It provides an effective device and method for realizing real-time feedback of the concentration during the stirring process of the filling slurry, and provides important support for the development of metal ore filling mining technology.

Description

Technical Field

[0001] The present invention relates to the field of mine filling technology, and in particular to a filling slurry concentration monitoring system and method based on real-time spectral information. Background Art

[0002] While mineral resources ensure the sustainable development of the national economy, the problem of environmental pollution caused by resource exploitation is becoming increasingly prominent. With the increase in resource exploitation efforts, the traditional extensive mining development model has shown the characteristics of "high exploitation, low utilization, and high emissions", which not only occupies a large amount of land, but also leads to the continuous deterioration of environmental pollution in mining areas, serious pollution of solid waste, water bodies and heavy metals, and frequent mine safety accidents, affecting the sustainable and healthy development of my country's metal mining industry.

[0003] Backfill technology can minimize the discharge of solid waste (all tailings, slag, and waste rock), increasing its utilization rate to 50% to 100%. There is no dehydration underground, no dust is raised on the surface, and harmful metal ions are cemented and sealed in the backfilled mining area without precipitation, achieving zero discharge of solid waste and wastewater from mines. There is no or minimal construction of tailings ponds and waste rock dumps in mining areas, achieving the coordinated development of resource development and ecological protection. At the same time, backfill mining can effectively prevent and control the collapse of goaf areas and eliminate the hazards of tailings ponds, realizing a new disaster management model of "one waste to treat two hazards."

[0004] Backfill mining technology is a key approach to achieving green development of metal mines. Concentration is a key factor affecting the three major process steps: slurry mixing and preparation, pipeline transportation, and stope filling. However, accurately measuring and controlling backfill slurry concentration is extremely difficult, leading to large fluctuations in slurry concentration during transportation and causing serious problems such as pipe blockage, bursts, and instability of the backfill.

[0005] The slurry concentration test methods generally include direct drying method, concentration pot determination method, online concentration measurement method, etc. The drying method and concentration pot method have long test cycles and complex processes. Currently, they are only used in a few mines or during the commissioning of filling systems. Online concentration detectors can be divided into two types: invasive and non-invasive. Among them, the most common invasive types include photoelectric concentration meters, differential pressure concentration meters, U-tube concentration meters, vibration concentration meters, etc. Non-invasive concentration meters are mostly based on ray attenuation method, ultrasonic method, microwave method, conductivity method, capacitance method, etc. However, considering the monitoring accuracy, monitoring medium conditions, monitoring cost, and the degree of disturbance to transportation, the aforementioned online concentration monitoring methods are not suitable for real-time monitoring of filling slurry concentration.

[0006] In summary, due to the constraints of convenience, accuracy, safety, applicability and other aspects, the traditional slurry concentration measurement method cannot meet the requirements of precision, automation and intelligence of modern filling mining technology, resulting in the difficulty in achieving accurate measurement of filling slurry concentration, which has become a bottleneck problem restricting the intelligent control of concentration in the filling slurry preparation process. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to provide a filling slurry concentration monitoring system and method based on real-time spectral information, so as to realize real-time monitoring of the concentration of the filling slurry during stirring, thereby promoting the intelligentization of the filling slurry preparation process.

[0008] The system includes a visualization module, a stirring tank, a hyperspectral scanning device, an image processing module, a filling slurry spectrum database and a neural network learning module, wherein the hyperspectral scanning device is arranged above the stirring tank, the hyperspectral scanning device is connected to the image processing module, the image combing module is connected to the neural network learning module, the data in the filling slurry spectrum database is called by the neural network learning module, and the neural network learning module is connected to the visualization module.

[0009] The hyperspectral scanning device was installed on a steel bracket directly above the open stirring tank at a distance of 30 cm to 40 cm from the liquid surface to facilitate data acquisition.

[0010] The hyperspectral scanning device includes a spectrometer and an area array camera. The spectrometer is placed in front of the area array camera, facing the stirring tank, and halogen lamp light sources are set on both sides of the hyperspectral scanning device.

[0011] The image processing module uses chemometric spectral analysis to perform spectral matching on the spectral information obtained by the hyperspectral scanning device to identify and analyze sample components. It also enables quantitative calculation of mineral composition and concentration.

[0012] Transfer learning and zero-shot learning mechanisms are introduced into the neural network learning module. As the filling slurry spectral database becomes richer, the inversion model can be continuously improved and the accuracy of concentration measurement can be enhanced.

[0013] The process of establishing the filling slurry spectrum database is as follows: by sampling the slurry in the stirring tank and performing X-ray fluorescence spectroscopy analysis to obtain the mineral and elemental composition and proportion, the filling slurry concentration is measured and calibrated using the drying method, and the measurement results are used as the benchmark value, and the benchmark value set is used as the database.

[0014] The data source of the neural network learning module is the mineral composition, particle size distribution and concentration quantitatively calculated by the image processing module.

[0015] In this system, a hyperspectral scanning device collects spectral data (hyperspectral images and reflectance spectra) from the filler slurry and transmits it to a computer. This device uses non-contact monitoring, eliminating the need to disrupt the slurry's internal structure. The image processing module processes the spectral images, employing dedicated image processing software and a built-in inversion model to identify hyperspectral images as slurry characteristic parameters. The neural network learning module compares conventional measured data, establishing a relationship between spectral data and concentration to further develop a filler slurry concentration monitoring model, ultimately achieving high-precision monitoring of the filler slurry concentration.

[0016] The application method of the monitoring system comprises the following steps:

[0017] S1: Establishing a spectral database of filling slurry:

[0018] Samples are taken from the stirring tank, and the mineral and elemental composition and ratio of the filling slurry are analyzed using X-ray fluorescence spectroscopy. The concentration of the filling slurry is measured and calibrated using a drying method. The measured concentration values ​​and the corresponding mineral components are used as reference values. The reference values ​​are collected as a database of filling slurry spectra. Sampling is continuously performed during the stirring process to improve the database.

[0019] S2: Install the hyperspectral scanning device:

[0020] A bracket is used to place the hyperspectral scanning device vertically above the stirring tank. A semiconductor refrigeration and high-pressure air dual cooling device, as well as high-pressure air curtain technology, are used to protect the lens of the hyperspectral scanning device. This reduces interference with the data extraction process caused by mud splashing, dust, and water vapor in the stirring tank, thereby improving the adaptability of the device.

[0021] S3: Obtaining filling slurry composition:

[0022] Each mineral component in the filler slurry emits radiation of different wavelengths under the illumination of a hyperspectral scanning device. The spectrometer disperses the broad-wavelength mixed light into single-wavelength light of different frequencies and projects the dispersed light onto an area array camera. The area array camera then performs photoelectric conversion, ultimately obtaining spectral lines arranged by wavelength or frequency. The spectral line intensity of the absorption line is used as the characteristic vector of the mineral component, forming a learning sample set for filler slurry composition detection.

[0023] S4: Establish a relationship model between mineral content and slurry concentration:

[0024] A neural network learning module is used to train the learning sample set of mineral composition detection obtained in S3, fit the feature vectors and the data in the filling slurry spectrum database, and establish the relationship between mineral composition content and absorption line intensity, as well as the relationship between slurry concentration and absorption line intensity;

[0025] S5: Visualization output:

[0026] After the S4 model is established, the concentration of the filling slurry is analyzed using the spectral data in the stirring tank during the filling process, and the concentration monitoring results are output through the visualization module;

[0027] S6: Filling slurry concentration control:

[0028] According to the concentration value output by the visualization module, the front end feed of the mixing tank is adjusted as needed, thereby achieving precise control of the filling slurry concentration during the mixing process.

[0029] The beneficial effects of the above technical solution of the present invention are as follows:

[0030] In the above scheme, a hyperspectral scanning device is used to perform online analysis of the mineral composition of the filling material and extract spectral information. Spectral analysis methods such as chemometrics and deep learning are combined to perform spectral matching on the tested filling material to identify and analyze the sample components and achieve concentration analysis based on spectral data. This is an effective filling material concentration monitoring system and method based on spectral recognition patterns. This filling material concentration monitoring system based on real-time spectral information has a simple structure and a high degree of automation. It can achieve non-contact measurement of filling material concentration, providing an effective device and method for achieving real-time feedback on concentration during the filling process, and providing important support for the development of metal ore filling mining technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Schematic diagram of the structure of the filling slurry concentration monitoring system based on real-time spectral information of the present invention;

[0032] Figure 2 This is a basic structural diagram of the hyperspectral scanning device of the present invention.

[0033] Among them: 1- visualization module; 2- stirring tank; 3- hyperspectral scanning device; 4- image processing module; 5- filling slurry spectrum database; 6- neural network learning module. DETAILED DESCRIPTION

[0034] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0035] The present invention provides a filling slurry concentration monitoring system and method based on real-time spectral information.

[0036] like Figure 1As shown, the system includes a visualization module 1, a stirring tank 2, a hyperspectral scanning device 3, an image processing module 4, a filling slurry spectrum database 5 and a neural network learning module 6, wherein the hyperspectral scanning device 3 is arranged above the stirring tank 2, the hyperspectral scanning device 3 is connected to the image processing module 4, the image combing module 4 is connected to the neural network learning module 6, the data in the filling slurry spectrum database 5 is called by the neural network learning module 6, and the neural network learning module 6 is connected to the visualization module 1.

[0037] The hyperspectral scanning device 3 is installed on a steel bracket just above the open stirring tank at a distance of 30 cm to 40 cm from the liquid surface to facilitate data acquisition.

[0038] like Figure 2 As shown, the hyperspectral scanning device 3 includes a spectrometer, an area array camera, an optical lens, a cooling device and an air curtain dustproof housing. The spectrometer is placed in front of the area array camera, facing the stirring tank. Halogen lamps are set on both sides of the hyperspectral scanning device 3 to provide light sources.

[0039] The image processing module 4 recognizes the image data obtained by the hyperspectral scanning device as characteristic parameters of the filling slurry and simultaneously realizes the quantitative calculation of the mineral components and concentrations.

[0040] Transfer learning and zero-shot learning mechanisms are introduced into the neural network learning module 6. As the filling slurry spectrum database 5 is enriched, the inversion model can be continuously improved and the accuracy of concentration measurement can be improved.

[0041] The process of establishing the filling slurry spectrum database 5 is as follows: by sampling the slurry in the stirring tank and performing X-ray fluorescence spectroscopy analysis to obtain the mineral and element composition and proportion, and using the drying method to measure and calibrate the filling slurry concentration, the measurement results are used as reference values, and the reference value collection is used as the database.

[0042] The data source of the neural network learning module 6 is the mineral composition, particle size distribution and concentration quantitatively calculated by the image processing module 4.

[0043] The application method of the monitoring system comprises the following steps:

[0044] S1: Establishing a spectral database of filling slurry:

[0045] Samples are taken from the stirring tank, and the mineral and elemental composition and ratio of the filling slurry are analyzed using X-ray fluorescence spectroscopy. The concentration of the filling slurry is measured and calibrated using a drying method. The measured concentration values ​​and the corresponding mineral components are used as reference values. The reference values ​​are collected as a database of filling slurry spectra. Sampling is continuously performed during the stirring process to improve the database.

[0046] S2: Install the hyperspectral scanning device:

[0047] A bracket is used to place the hyperspectral scanning device vertically above the stirring tank, and a semiconductor refrigeration and high-pressure air dual cooling device as well as high-pressure air curtain technology are used to protect the lens of the hyperspectral scanning device;

[0048] S3: Obtaining filling slurry composition:

[0049] Each mineral component in the filler slurry emits radiation of different wavelengths under the illumination of a hyperspectral scanning device. The spectrometer disperses the broad-wavelength mixed light into single-wavelength light of different frequencies and projects the dispersed light onto an area array camera. The area array camera then performs photoelectric conversion, ultimately obtaining spectral lines arranged by wavelength or frequency. The spectral line intensity of the absorption line is used as the characteristic vector of the mineral component, forming a learning sample set for filler slurry composition detection.

[0050] S4: Establish a relationship model between mineral content and slurry concentration:

[0051] The neural network learning module is used to train the learning sample set of mineral composition detection obtained in S3, fit the characteristic vector and the data in the filling slurry spectrum database, and establish the relationship between mineral composition content and absorption line intensity, as well as the relationship between slurry concentration and absorption line intensity.

[0052] S5: Visualization output:

[0053] After the S4 model is established, the concentration of the filling slurry is analyzed using the spectral data in the stirring tank during the filling process, and the concentration monitoring results are output through the visualization module;

[0054] S6: Filling slurry concentration control:

[0055] According to the concentration value output by the visualization module, the front end feed of the mixing tank is adjusted as needed, thereby achieving precise control of the filling slurry concentration during the mixing process.

[0056] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. An application method of a filling slurry concentration monitoring system based on real-time spectral information, characterized in that: The steps are as follows: S1: Establishing a spectral database of filling slurry: Samples are taken from the stirring tank, and the mineral and elemental composition and ratio of the filling slurry are analyzed using X-ray fluorescence spectroscopy. The concentration of the filling slurry is measured and calibrated using a drying method. The measured concentration values ​​and the corresponding mineral components are used as reference values. The reference values ​​are collected as a database of filling slurry spectra. Sampling is continuously performed during the stirring process to improve the database. S2: Install the hyperspectral scanning device: A bracket is used to place the hyperspectral scanning device vertically above the stirring tank, and a semiconductor refrigeration and high-pressure air dual cooling device as well as high-pressure air curtain technology are used to protect the lens of the hyperspectral scanning device; S3: Obtaining filling slurry composition: Each mineral component in the filler slurry emits radiation of different wavelengths under the illumination of a hyperspectral scanning device. The spectrometer disperses the broad-wavelength mixed light into single-wavelength light of different frequencies and projects the dispersed light onto an area array camera. The area array camera then performs photoelectric conversion, ultimately obtaining spectral lines arranged by wavelength or frequency. The spectral line intensity of the absorption line is used as the characteristic vector of the mineral component, forming a learning sample set for filler slurry composition detection. S4: Establish a relationship model between mineral content and slurry concentration: A neural network learning module is used to train the learning sample set of mineral composition detection obtained in S3, fit the feature vectors and the data in the filling slurry spectrum database, and establish the relationship between mineral composition content and absorption line intensity, as well as the relationship between slurry concentration and absorption line intensity; S5: Visualization output: After the S4 model is established, the concentration of the filling slurry is analyzed using the spectral data in the stirring tank during the filling process, and the concentration monitoring results are output through the visualization module; S6: Filling slurry concentration control: According to the concentration value output by the visualization module, the front end feed of the mixing tank is adjusted as needed, thereby achieving precise control of the filling slurry concentration during the mixing process; The monitoring system includes a visualization module, a stirring tank, a hyperspectral scanning device, an image processing module, a filling slurry spectrum database and a neural network learning module, wherein the hyperspectral scanning device is arranged above the stirring tank, the hyperspectral scanning device is connected to the image processing module, the image processing module is connected to the neural network learning module, the data in the filling slurry spectrum database is called by the neural network learning module, and the neural network learning module is connected to the visualization module; The hyperspectral scanning device is installed on a steel bracket directly above the open stirring tank at a distance of 30 cm to 40 cm from the liquid surface to facilitate data acquisition; The process of establishing the filler slurry spectrum database is as follows: by sampling the slurry in the stirring tank and performing X-ray fluorescence spectroscopy analysis to obtain the mineral and elemental composition and proportion, the concentration of the filler slurry is measured and calibrated using a drying method, and the measurement results are obtained as reference values, and the reference value set is used as the database; the hyperspectral scanning device includes a spectrometer and an area array camera. The spectrometer is placed at the front end of the area array camera, facing the stirring tank, and halogen lamp light sources are set on both sides of the hyperspectral scanning device.

2. The filling slurry concentration monitoring system based on real-time spectral information according to claim 1 is characterized in that: The image processing module uses chemometric spectral analysis to perform spectral matching on the spectral information obtained by the hyperspectral scanning device, so as to identify and analyze the sample components and realize quantitative calculation of mineral components and concentrations.

3. The filling slurry concentration monitoring system based on real-time spectral information according to claim 1 is characterized in that: The neural network learning module introduces transfer learning and zero-shot learning mechanisms. As the filling slurry spectrum database is enriched, the inversion model can be continuously improved to enhance the accuracy of concentration measurement.

4. The filling slurry concentration monitoring system based on real-time spectral information according to claim 1 is characterized in that: The data source of the neural network learning module is the mineral composition, particle size distribution and concentration obtained by quantitative calculation of the image processing module.

Images