Spodumene ore ultraviolet fluorescence preselection method based on deep learning

Abstract

The invention provides a spodumene ore ultraviolet fluorescence preselection method based on deep learning, and relates to the technical field of mineral processing and intelligent photoelectricity. According to the method, an advanced image processing algorithm and a pattern recognition technology are adopted, the content of Li2O in the spodumene ore is measured according to the photoluminescence characteristics of the spodumene ore, the spodumene ore can be divided into high-grade spodumene ore and low-grade spodumene ore based on the image of the spodumene ore under ultraviolet light, and the method has the advantages that the method is simple and convenient to operate, and the cost is low. The low-grade spodumene ore is separated from an ore sample by adjusting a control system of separation equipment, so that the grade of the whole spodumene ore is improved, the method has the characteristics of high efficiency, accuracy and automation, and compared with a traditional separation method, the separation method based on X-ray radiation can better utilize the characteristics of the spodumene ore, so that the separation efficiency of the spodumene ore is improved. The method can be applied to processing and utilization of spodumene ores, and is beneficial to improving the comprehensive utilization efficiency of lithium resources.

Description

Technical Field

[0001] The present invention belongs to the field of mineral processing and intelligent optoelectronic technology, and in particular relates to a spodumene ore ultraviolet fluorescence preselection method based on deep learning. Background Art

[0002] Spodumene is a lithium-containing aluminum silicate mineral. Due to its high lithium content and wide application fields, it is considered to be one of the important sources of lithium mineral resources. Lithium resources are of great strategic significance in industries such as new energy batteries, ceramics and glass. With the rapid development of the new energy industry, the demand for lithium resources has increased dramatically. However, most of the existing spodumene ores are of low grade, and the direct use of these low-grade spodumene ores is less economical. Therefore, how to efficiently sort low-grade spodumene ores and improve their grade and economic value has become a problem to be solved in the mining field.

[0003] Traditional spodumene ore separation methods mainly include gravity separation, magnetic separation, flotation and chemical separation. These methods have the following problems in practical application:

[0004] 1. Low efficiency: Traditional mineral processing methods have low efficiency in sorting low-grade spodumene ores, making it difficult to achieve large-scale and efficient sorting.

[0005] 2. High cost: The flotation method using chemical agents and the high-energy consumption magnetic separation method result in high beneficiation costs, which limits its application in low-grade spodumene ores.

[0006] 3. High energy consumption: Traditional mineral processing methods usually require a lot of energy consumption, which does not meet the current requirements of energy conservation and environmental protection.

[0007] 4. Complex process flow: Most traditional mineral processing methods have complex process flows, requiring multiple steps and a lot of manual intervention, which increases the difficulty and cost of operation.

[0008] Spodumene ore can emit red fluorescence under ultraviolet light with a wavelength of 365nm. This characteristic makes it possible to use optical methods to sort spodumene. Through ultraviolet irradiation and image processing technology, the lithium content in spodumene ore can be quickly and accurately detected, and sorted through automated equipment, which significantly improves sorting efficiency and accuracy and reduces costs and energy consumption. Therefore, the method of pre-selecting low-grade spodumene ore using ultraviolet fluorescence has broad application prospects and important practical significance. Summary of the invention

[0009] The technical problem to be solved by the present invention is to address the deficiencies of the above-mentioned prior art and to provide a deep learning-based spodumene ore ultraviolet fluorescence pre-selection method, which solves the technical problem of how to efficiently sort low-grade spodumene ore in the mining field and improve its grade and economic value.

[0010] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for ultraviolet fluorescence preselection of spodumene ore based on deep learning, which can significantly improve the grade and economic value of spodumene ore. The method utilizes the characteristic that spodumene ore exhibits red fluorescence under ultraviolet light with a wavelength of 365nm, and realizes efficient sorting of low-grade spodumene ore through advanced image processing algorithms and pattern recognition technology, comprising the following steps:

[0011] Step 1: Collect spodumene ore samples and mark them: irradiate the ore samples under ultraviolet light with a wavelength of 365nm to make the ore emit red fluorescence, and use a high-resolution camera to collect the fluorescence image of the ore under ultraviolet light to ensure the image quality and clarity of details. The collected fluorescence images are used as spodumene ore samples, including positive samples containing spodumene and negative samples without spodumene;

[0012] Step 2: Use image processing algorithms to process the spodumene ore sample and extract photoluminescence features to obtain useful feature data;

[0013] The collected spodumene ore samples are processed using image processing algorithms, and the specific steps include: noise filtering, edge detection and photoluminescence feature extraction;

[0014] The noise filtering is used to remove noise interference in the spodumene ore sample to improve the signal-to-noise ratio of the spodumene ore sample;

[0015] The edge detection is used to identify and extract edge features of spodumene ore samples;

[0016] The photoluminescence feature extraction is used to extract the photoluminescence region of the spodumene ore sample for subsequent analysis and sorting;

[0017] Step 3: Use the convolutional neural network in the deep learning algorithm to build a deep learning model for intelligent photoelectric pre-selection of spodumene ore: After the spodumene ore sample is processed, the pattern recognition technology is applied to analyze the extracted photoluminescence characteristics, and the content of Li2O in the spodumene ore sample is determined based on the photoluminescence characteristics;

[0018] The prepared spodumene ore sample data is divided into a training set and a validation set, the training set is used to train the spodumene ore intelligent photoelectric pre-selection deep learning model, and the intelligent photoelectric pre-selection deep learning model is tuned according to the performance of the validation set, so that the spodumene ore intelligent photoelectric pre-selection deep learning model can accurately identify and classify spodumene ores of different grades;

[0019] The performance of the validation set includes: accuracy, precision, recall and recognition time of the validation set;

[0020] Step 4: Automatically separate low-grade spodumene ore by adjusting the control system of the sorting equipment;

[0021] The control system of the sorting equipment includes an image analysis module, a sorting instruction module and an execution module;

[0022] The image analysis module is used to receive and analyze the processed spodumene ore sample data to determine the grade of the spodumene ore sample;

[0023] Based on the fluorescence image of spodumene ore under ultraviolet light, the grade of spodumene ore samples is divided into high-grade spodumene ore and low-grade spodumene ore through the trained spodumene ore intelligent photoelectric pre-selection deep learning model; the high-grade spodumene ore is retained for further processing, while the low-grade spodumene ore is separated;

[0024] The sorting instruction module issues a sorting instruction according to the result of the image analysis module;

[0025] The execution module performs the actual separation operation on the low-grade spodumene ore according to the sorting instruction.

[0026] The beneficial effects of adopting the above technical solution are: the present invention provides a deep learning-based ultraviolet fluorescence pre-selection method for spodumene ore, which can achieve the sorting index of pre-enriched concentrate Li2O ≥ 2.0% and Li2O recovery rate ≥ 95%. Compared with the traditional sorting method, this method has the advantages of high efficiency, precision, automation and energy saving, and the sorting speed is significantly improved, which is suitable for large-scale spodumene ore sorting operations; photoluminescence characteristics are used for detection and classification, and the sorting accuracy is high; the whole process has a high degree of automation, reducing manual operations and labor costs; significantly reducing energy consumption and being more environmentally friendly.

[0027] The method of the present invention can be widely applied to the processing and utilization of spodumene ore, improve the comprehensive utilization efficiency of lithium resources, and has broad application prospects and economic value. The method is not only suitable for the beneficiation process of spodumene ore, but can also be promoted and applied to the separation of other ores with specific photoluminescence characteristics, providing new ideas and new methods for the development of mining separation technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A schematic diagram of the implementation process of the ultraviolet fluorescence sorting machine provided by an embodiment of the present invention;

[0029] Figure 2 A schematic flow chart of a method for ultraviolet fluorescence preselection of spodumene ore based on deep learning provided in an embodiment of the present invention;

[0030] Figure 3An embodiment of the present invention provides a fluorescent image of an ore irradiated under ultraviolet light. DETAILED DESCRIPTION

[0031] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0032] The ore used in the embodiment of the present invention is from Dahongliutan, Xinjiang, and the sorting equipment used is a self-developed ultraviolet fluorescence sorting machine, such as Figure 1 As shown, the adopted intelligent photoelectric pre-selection deep learning model is built based on the Keras toolbox in the Python 3.6 environment. In the training stage of the intelligent photoelectric pre-selection deep learning model, the Learning Rate is 1e-4, the Downsample factor is 16, the Optimizer is Adam, the Momentum is 0.9, the batch size is 8, the epoch is 200, and the callback is ReduceLR On Plateau (factor=0.5, patience=3), which is used to monitor the effective loss. The Early stop (Min_Delta=0, patience=10) algorithm is used to monitor the effective loss of the segmented intelligent photoelectric pre-selection deep learning model; the ultraviolet fluorescence sorting machine includes a feeding device, a belt conveyor, a fluorescent light-emitting device, an image acquisition device and a mechanical separation device; finally, the intelligent photoelectric pre-selection deep learning model is trained in the Nvidia RTX 2080Ti and CUDA10.1, cudnn7.3.1 environment.

[0033] The feeding device is responsible for feeding the ore into the belt conveyor evenly and continuously to ensure the stability of the entire pre-selection process;

[0034] The belt conveyor conveys the ore from the feeding device to below the fluorescent light-emitting device, ensuring that the ore can be irradiated by ultraviolet light;

[0035] The fluorescent light emitting device excites the fluorescent components in the ore by emitting ultraviolet light with a wavelength of 365nm, causing it to emit red fluorescence;

[0036] The image acquisition device uses a camera to collect images or signals of the ore under ultraviolet light, so as to identify and determine the composition of the ore;

[0037] The mechanical separation device uses an air spray valve to separate the ore containing spodumene from other ores according to the recognition result provided by the image acquisition device, thereby achieving preliminary ore sorting;

[0038] In this embodiment, a spodumene ore ultraviolet fluorescence preselection method based on deep learning is provided. Figure 2 As shown, the following steps are included:

[0039] Step 1: Collect spodumene ore samples and mark them: Collect a large number of ore samples from Xinjiang Dahongliutan, and irradiate a large number of ore samples under ultraviolet light with a wavelength of 365nm to make the ore emit red fluorescence. Use a high-resolution camera to collect the fluorescence image of the ore under ultraviolet light, such as Figure 3 As shown, to ensure the image quality and clarity of details, the collected fluorescence images are used as spodumene ore samples, including positive samples containing spodumene and negative samples not containing spodumene;

[0040] Step 2: Data preprocessing and photoelectric feature extraction of spodumene ore samples, including image processing algorithms and feature extraction engineering, to obtain useful feature data: Use advanced image processing algorithms to preprocess the collected spodumene ore samples, and the specific steps include: noise filtering, edge detection and photoluminescence feature extraction;

[0041] The noise filtering can remove the noise interference in the spodumene ore sample and improve the signal-to-noise ratio of the spodumene ore sample;

[0042] The edge detection is used to identify and extract edge features of spodumene ore samples;

[0043] The photoluminescence feature extraction is to extract the photoluminescence area of ​​the spodumene ore sample for subsequent analysis and sorting;

[0044] Step 3: Use the convolutional neural network in the deep learning algorithm to build a deep learning model for intelligent photoelectric pre-selection of spodumene ore: After the spodumene ore sample is processed, the pattern recognition technology is applied to analyze the extracted photoluminescence characteristics, and the content of Li2O in the spodumene ore sample is determined based on the photoluminescence characteristics;

[0045] The prepared spodumene ore sample data is divided into a training set and a validation set, the training set is used to train the spodumene ore intelligent photoelectric pre-selection deep learning model, and the intelligent photoelectric pre-selection deep learning model is tuned according to the performance of the validation set, so that the spodumene ore intelligent photoelectric pre-selection deep learning model can accurately identify and classify spodumene ores of different grades;

[0046] The performance of the validation set includes: accuracy, precision, recall and recognition time of the validation set;

[0047] Step 4: Automatically separate low-grade spodumene by adjusting the control system of the sorting equipment:

[0048] The sorting device includes an image analysis module, a sorting instruction module and an execution module;

[0049] The image analysis module is used to receive and analyze the processed spodumene ore sample data fluorescent image to determine the grade of the spodumene ore sample;

[0050] Based on the fluorescence image of spodumene ore under ultraviolet light, the grade of spodumene ore samples is divided into high-grade spodumene ore and low-grade spodumene ore through the trained spodumene ore intelligent photoelectric pre-selection deep learning model; the high-grade spodumene ore is retained for further processing, while the low-grade spodumene ore is separated;

[0051] The sorting instruction module issues a sorting instruction according to the result of the image analysis module;

[0052] The execution module performs the actual separation operation on the spodumene ore according to the sorting instruction;

[0053] In this embodiment, after the ore is crushed, an ultraviolet fluorescence sorter is used to sort the spodumene ore sample, which specifically includes the following steps:

[0054] Step 4.1: Ore crushing: Use a crusher (such as a jaw crusher, cone crusher, etc.) to crush large pieces of spodumene ore into smaller particles of 10-50 mm, and transfer the crushed spodumene ore to the secondary crushing equipment through a belt conveyor. In the secondary crushing equipment cone crusher, the spodumene ore is further crushed into a particle size of 3-10 mm to ensure uniform particle size;

[0055] Step 4.2: Ore feeding: The crushed spodumene ore is evenly fed into the belt conveyor, and the spodumene ore in the belt conveyor is evenly fed onto the conveyor belt through the vibrating feeder;

[0056] The vibrating feeder has an adjustment function to control the feeding amount. During the feeding process, care should be taken to avoid blockage and overload to ensure the stability of the belt conveyor.

[0057] Step 4.3: Ore transmission: The spodumene ore is transmitted from the feeding device to the detection area of ​​the fluorescent light-emitting device through a belt conveyor. The belt conveyor needs to ensure that the spodumene ore remains stable during the transmission process to avoid shaking and slipping. The transmission speed needs to be adjusted according to the processing capacity of the detection device to ensure the accuracy and efficiency of the detection;

[0058] Step 4.4: Ore detection: Use ultraviolet fluorescence equipment to detect the content of Li2O in spodumene ore. Install the ultraviolet fluorescence detection equipment above the conveyor belt. When the spodumene ore passes through the detection equipment, the ultraviolet light source irradiates the surface of the spodumene ore to stimulate the unique fluorescence of the spodumene ore. Use the spodumene ore intelligent photoelectric pre-selection deep learning model to classify the spodumene ore into high-grade spodumene ore and low-grade spodumene ore for real-time analysis and processing;

[0059] This embodiment includes spodumene ore separation: the control system of the detection equipment will automatically classify the spodumene ore according to the fluorescence intensity of the spodumene ore, and divide the ore into two categories: high-grade spodumene ore and low-grade spodumene ore. A pneumatic sorting device or a mechanical sorting device is installed at the end of the conveyor belt. When the ore passes through the sorting device, the pneumatic injection device or the mechanical arm pushes the high-grade spodumene ore into a predetermined collection area, and the low-grade spodumene ore continues to be transported to another collection area to ensure the sorting accuracy and efficiency. By adjusting the equipment parameters to optimize the sorting effect, Li2O ≥ 2.0%, the above-mentioned Li2O recovery rate ≥ 95%.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A spodumene ore ultraviolet fluorescence preselection method based on deep learning, characterized in that: The following steps are involved: Step 1: Collect the fluorescence image of the ore under ultraviolet light as the spodumene ore sample and mark it, including the positive sample containing spodumene and the negative sample not containing spodumene; Step 2: Use image processing algorithms to process the spodumene ore sample and extract photoluminescence features to obtain useful feature data; Step 3: Use the convolutional neural network in the deep learning algorithm to build a deep learning model for intelligent photoelectric pre-selection of spodumene ore: After the spodumene ore sample is processed, the pattern recognition technology is applied to analyze the extracted photoluminescence characteristics, and the content of Li2O in the spodumene ore sample is determined based on the photoluminescence characteristics; Step 4: Automatically separate low-grade spodumene ore by adjusting the control system of the sorting equipment.

2. A method for ultraviolet fluorescence preselection of spodumene ore based on deep learning according to claim 1, characterized in that: The specific method of step 1 is: Spodumene ore samples were collected and marked: the ore samples were irradiated under ultraviolet light with a wavelength of 365nm to make the ore emit red fluorescence, and a high-resolution camera was used to capture the fluorescence image of the ore under ultraviolet light to ensure the image quality and clarity of details. The collected fluorescence images were used as spodumene ore samples, including positive samples containing spodumene and negative samples not containing spodumene.

3. A method for preselecting spodumene ore by ultraviolet fluorescence based on deep learning according to claim 2, characterized in that: The step 2 uses an image processing algorithm to process the collected spodumene ore sample, and the specific steps include: noise filtering, edge detection and photoluminescence feature extraction; The noise filtering is used to remove noise interference in the spodumene ore sample to improve the signal-to-noise ratio of the spodumene ore sample; The edge detection is used to identify and extract edge features of spodumene ore samples; The photoluminescence feature extraction is used to extract the photoluminescence area of ​​the spodumene ore sample for subsequent analysis and sorting.

4. The method for ultraviolet fluorescence preselection of spodumene ore based on deep learning according to claim 1, characterized in that: The step 3 divides the prepared spodumene ore sample data into a training set and a validation set, uses the training set to train the spodumene ore intelligent photoelectric pre-selection deep learning model, and tunes the intelligent photoelectric pre-selection deep learning model according to the performance of the validation set, so that the spodumene ore intelligent photoelectric pre-selection deep learning model can accurately identify and classify spodumene ores of different grades.

5. A method for ultraviolet fluorescence preselection of spodumene ore based on deep learning according to claim 4, characterized in that: The performance of the verification set described in step 3 includes: the accuracy, precision, recall rate and recognition time of the verification set.

6. The method for ultraviolet fluorescence preselection of spodumene ore based on deep learning according to claim 1, characterized in that: The control system of the sorting equipment in step 4 includes an image analysis module, a sorting instruction module and an execution module; The image analysis module is used to receive and analyze the processed spodumene ore sample data to determine the grade of the spodumene ore sample; The sorting instruction module issues a sorting instruction according to the result of the image analysis module; The execution module performs the actual separation operation on the low-grade spodumene ore according to the sorting instruction.

7. The method for ultraviolet fluorescence preselection of spodumene ore based on deep learning according to claim 1, characterized in that: The image analysis module is based on the fluorescent image of spodumene ore under ultraviolet light, and divides the grade of spodumene ore samples into two categories: high-grade spodumene ore and low-grade spodumene ore through a trained spodumene ore intelligent photoelectric pre-selection deep learning model; the high-grade spodumene ore is retained for further processing, while the low-grade spodumene ore is separated.

Images