Green separation method and system for pegmatite type lithium ore with intelligent recognition function

Abstract

The invention discloses an intelligent recognition green separation method and system for pegmatite type lithium ores, and relates to the technical field of mineral recognition and separation. Training an intelligent identification and sorting decision model; ore pretreatment and intelligent roughing waste throwing; stage ore grinding and intelligent real-time separation; by means of the method, the high recovery rate of lithium minerals and high-grade concentrate output can be achieved.

Description

Technical Field

[0001] This invention relates to the field of mineral identification and sorting technology, specifically a green sorting method and system for intelligent pegmatite-type lithium ore. Background Technology

[0002] Pegmatite-type lithium deposits are an important source of lithium resources, with spodumene, lepidolite, and petalite being the main lithium minerals. Traditional separation methods mainly rely on the "grind-gravity separation-flotation" or "grind-flotation" process, which has the following inherent drawbacks: 1. The process is complex and inefficient: it requires multi-stage grinding and multi-stage separation, resulting in high energy consumption. The recovery rate of lithium minerals, especially those with fine particle size or complex symbiotic relationships with gangue minerals, is not ideal.

[0003] 2. High reagent consumption and environmentally unfriendly: The flotation process requires the use of large amounts of modifiers, collectors and frothers, which may cause water pollution and ecological damage, and does not meet the requirements of green mining development.

[0004] 3. Sorting accuracy depends on the homogeneity of the ore: Traditional processes are sensitive to fluctuations in the properties of the raw ore. When the ore's properties (such as mineral composition, dissemination characteristics, degree of crystallization, etc.) change, the sorting indicators are prone to fluctuation, requiring frequent adjustments to process parameters.

[0005] 4. Coarse tailings discharge, resource waste: In order to reduce over-grinding, the early tailings are discarded with coarse particles, which may result in the loss of lithium minerals that have been liberated or intergrowth in the tailings. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a green sorting method and system for intelligent identification of pegmatite-type lithium ore. It constructs a mineral gene feature library, uses an intelligent model to establish a mapping relationship between mineral gene features and optimal sorting behavior, and performs intelligent identification and sorting decisions in real time on the production line to achieve high recovery rate of lithium minerals and high-grade concentrate output.

[0007] The objective of this invention is achieved through the following technical solution: a green sorting method for pegmatite-type lithium ore with intelligent identification, the method comprising the following steps: S1. Constructing a mineral gene feature library: Representative ore samples were collected for the target pegmatite-type lithium deposit. Multidimensional analysis techniques (such as MLA / QEMSCAN, laser-induced breakdown spectroscopy (LIBS), hyperspectral imaging, and X-ray micro-area analysis) were used to systematically characterize the samples and extract the following: Mineralogical genes: types, contents, grain size, degree of liberation, symbiotic relationships, crystal structure and crystal orientation of lithium minerals and major gangue minerals.

[0008] Physicochemical characteristics: surface chemical properties of the target mineral (such as zero point of charge, surface energy), characteristic spectra (visible light-near infrared-short wave infrared, Raman, LIBS spectra), density, specific magnetization coefficient, conductivity, dielectric constant, etc.

[0009] Response behavior genes: Data on the differences in separation behavior exhibited by minerals with different gene characteristics in simulated or micro-scale unit operations such as gravity separation, magnetic separation, electrostatic separation, and flotation.

[0010] S2. Training of Intelligent Identification and Sorting Decision Model: Based on the mineral gene feature library constructed in step S1, the following intelligent model is trained using machine learning or deep learning algorithms: Real-time mineral particle identification model: The input is a particle flow image or spectral sequence acquired by an online detection system (such as a hyperspectral camera, LIBS probe, or high-speed vision system). The model can identify the main lithium mineral species (such as spodumene and lepidolite) in each particle or particle group in real time and accurately, as well as their surface characteristics and their association with gangue.

[0011] The sorting path decision model takes the identified particle genetic characteristics (such as mineral type, co-occurrence, surface contamination, particle size, etc.) as input, and combines them with preset optimization objectives (such as maximum recovery rate, highest grade, lowest energy consumption, minimum reagent dosage, or their comprehensive benefits). From a predefined sorting method library (including but not limited to intelligent dry screening, sensor-based intelligent waste disposal, intelligent photoelectric sorting, intelligent high-voltage electrostatic separation, intelligent magnetic separation, intelligent foam flotation, intelligent heavy medium cyclone, etc.), the model dynamically determines the most effective single or combined sorting method and its optimal operating parameters (such as magnetic field strength, electrostatic separation voltage, flotation reagent regime and dosage) for the particle or similar particle group.

[0012] S3. Ore Pretreatment and Intelligent Roughing and Waste Disposal: After crushing, the raw ore does not directly enter the fine grinding operation. Instead, it is first screened and intelligently pre-selected. Intelligent photoelectric separators or hyperspectral separators are used. Based on the real-time mineral particle recognition model trained in S2, the blocky ore (e.g., 10-100mm) after medium or fine crushing is scanned and identified. The lithium-rich ore blocks, lithium-poor blocks or waste rock blocks that have been basically liberated are separated, realizing the pre-disposal of coarse particles and greatly reducing the amount fed into the mill.

[0013] S4. Staged Grinding and Intelligent Real-Time Sorting: The pre-selected ore undergoes staged grinding, with an intelligent online sorting unit installed after each grinding stage. The grinding product, after particle size classification, forms multiple particle size streams. Each particle size stream is transmitted through a transmission device equipped with online detection sensors (such as a hyperspectral imager or an online LIBS analyzer). Real-time Identification and Decision-Making: Online detection data is transmitted in real-time to the mineral particle real-time identification model and sorting path decision model trained in S2. The models instantly complete particle identification and sorting decisions, separating the particles into concentrate, middlings, and tailings.

[0014] Intelligent sorting: Based on decision commands, programmable intelligent actuators such as high-speed air valve arrays, guide plates, high-pressure electrodes, magnetic separation drums, and micro flotation column clusters guide particles from different particle sizes into different collection channels. For example: Qualified lithium concentrate particles that have been separated into individual particles → directly enter the concentrate pool.

[0015] Lithium-rich intergrowth particles → Return to this section or previous section for re-grinding.

[0016] Low lithium content or gangue particles → discarded as tailings or middlings.

[0017] For particles with specific properties (such as surface contamination or magnetic differences), decisions are made to use specific electrical or magnetic separation parameters for finer selection.

[0018] S5. Closed-loop feedback and model optimization: During the sorting process, the products (concentrate, middlings, tailings) of each sorting unit are continuously collected and analyzed online (such as portable XRF and online particle size analysis).

[0019] The actual sorting results are compared with the model prediction results to generate feedback data.

[0020] Using feedback data, incremental learning or online optimization is performed on the real-time mineral particle identification model and sorting path decision model in S2 periodically or in real time, so that the model can adapt to the small fluctuations in ore genetic characteristics and keep the sorting process in the optimal state at all times.

[0021] S6. Resource Utilization and Wastewater Purification: The coarse tailings produced can be used to prepare building materials or backfill materials, depending on their mineral composition. For the very small amount of flotation reagents that must be used, environmentally friendly reagents are preferred.

[0022] Wastewater generated during the sorting process is recycled to the greatest extent possible after coagulation, sedimentation, adsorption and advanced oxidation treatment, achieving near-zero discharge.

[0023] A green sorting system for pegmatite-type lithium ore with intelligent identification, implementing the above method, is provided. The system includes a mineral gene feature analysis platform, a model training and computing server, an intelligent roughing and waste disposal unit, a crushing module and a staged grinding loop, a cluster of intelligent real-time sorting units in series, an online analysis and feedback module, a central control system, and a tailings resource utilization and wastewater treatment module. The mineral gene feature analysis platform is used in step S1, the model training and computing server is used to carry and run the real-time mineral particle identification model and sorting path decision model in step S2, the crushing module and staged grinding loop are used in step S5, the tailings resource utilization and wastewater treatment module is used in step S6, and the central control system is used to coordinate the operation of each module and the data flow.

[0024] The intelligent coarse selection and waste disposal unit includes a feeder, a sensor array, and an intelligent actuator, which is used in step S3.

[0025] Each unit of the intelligent real-time sorting unit cluster includes an online detection module, an intelligent decision-making module, and a programmable sorting execution module. The intelligent decision-making module communicates with the model training and computing server for step S4.

[0026] The beneficial effects of this invention are: This invention systematically characterizes the "genetic" features of ore, establishes a mapping relationship between genetic features and optimal sorting behavior using an intelligent model, and performs intelligent identification and sorting decisions in real time on the production line to achieve high recovery rate of lithium minerals and high-grade concentrate output, while minimizing reagent usage, energy consumption and environmental impact. Attached Figure Description

[0027] Figure 1 This is a general process flow diagram of the present invention; Figure 2 A schematic diagram illustrating the structure and working principle of the intelligent real-time sorting unit; Figure 3 This is a block diagram of the sorting system. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0029] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0030] In one embodiment of this application, a typical pegmatite-type spodumene deposit is taken as an example: Ore properties: The main mineral composition of the raw ore is spodumene (18%), quartz (30%), feldspar (mainly albite and microcline, 25%), muscovite and biotite (15%), with the remainder being small amounts of beryl, tantalite, columbite, tourmaline, etc. The spodumene grains are unevenly distributed, with coarse grains reaching 5 mm and fine grains less than 0.1 mm. It is closely associated with albite, quartz, and mica. Some spodumene edges or fissures are aegirized and stained with iron, resulting in complex surface properties.

[0031] like Figures 1 to 2 As shown, a green sorting method for pegmatite-type lithium ore with intelligent identification is described. The method includes the following steps: S1. Constructing a mineral gene feature library: 1. Sampling and Preparation: A total of 200 kg of representative core and block ore samples were collected from different areas and strata in the mine. After coarse crushing, the samples were prepared into slides, block samples, and powder samples for different analyses.

[0032] 2. Multidimensional Joint Analysis: Mineralogy and Liberation Genes: Thirty slide samples were analyzed using an automated mineral parameter analysis system (MLA). Precise content, grain size distribution, and degree of liberation data for minerals such as spodumene, albite, quartz, and mica were obtained. Key findings: In the +0.3 mm grain size range, approximately 40% of spodumene had liberated from gangue monomers; in the -0.1 mm grain size range, spodumene mainly exhibited complex intergrowth with albite and mica. Three common crystal orientations of spodumene and their contact boundary types with gangue (smooth, serrated, and reactive edges) were recorded.

[0033] Physicochemical genes: Spectral characteristics: The surface of the bulk samples was scanned using a hyperspectral imager (VNIR-SWIR, 400-2500nm) to establish a standard spectral library for spodumene (characteristic absorption peaks located at ~1080nm, ~1160nm, ~1380nm, ~2200nm), altered spodumene (nephrite alteration leading to shift and weakening of the 2200nm peak), aegirine (iron-induced strong absorption), and mica (~1400nm, ~2200nm characteristic).

[0034] Surface chemistry: The zero electrical points of pure spodumene, nepheline spodumene, and albite in the pH range of 3-11 were determined to be 2.5, 4.5, and 2.0, respectively, by using a Zeta potentiometer.

[0035] Electrical and magnetic properties: Tests using a dielectric separator and a vibrating sample magnetometer revealed that spodumene, quartz, and feldspar are non-conductive minerals with slight differences in dielectric constant; aegirine and biotite exhibit weak magnetism (specific magnetic susceptibility approximately 35 × 10⁻⁶). -6 m³ / kg).

[0036] Response behavior genes: Micro-batch experiments were conducted, and particles of different gene types (such as pure spodumene, aegirized spodumene, and spodumene-albite intergrowths) were tested using high-voltage electrical separation (15-25kV), strong magnetic separation (background magnetic field strength 1.0T), and micro-flotation (sodium oleate system, pH=8). The results showed that pure spodumene exhibited the best conductivity at 22kV; aegirized spodumene was effectively captured under a 1.0T magnetic field; and the recovery rate of intergrowths in flotation depended on the intergrowth ratio.

[0037] 3. Construct a digital feature library: Integrate all the above data, including the image features, spectral curves, chemical composition, physical properties and corresponding sorting behavior tags of mineral particles, into a cloud database to form the "digital genetic ID card" of the deposit.

[0038] S2. Training of Intelligent Recognition and Sorting Decision Model: 1. Data Preprocessing and Labeling: Tens of thousands of particle data entries (including hyperspectral images, composition data, and behavioral tags) exported from the gene feature library are cleaned, enhanced, and compared with digital features. Each data entry is labeled with its "true identity" (e.g., single spodumene, 30% nephrite-altered spodumene, 60% spodumene-quartz intergrowth) and its "optimal fate" under various sorting conditions (e.g., high-voltage electrostatic separation - conductor products, weak magnetic separation - magnetic products, flotation - concentrate).

[0039] 2. Model Architecture and Training: A real-time mineral particle identification model is developed, combining an improved VGG16 convolutional neural network with a Transformer module (i.e., combining the VGG16 convolutional neural network with a Transformer module to achieve feature enhancement). The input is a 256x256 pixel image patch with 16 feature bands captured by an online hyperspectral camera. The model outputs the mineral type and co-occurrence estimate for each image patch. Using 80% of the data for training and 20% for validation, the model achieves a 98.7% accuracy rate in identifying spodumene particles on the test set, with a co-occurrence estimation error of less than ±5%.

[0040] The sorting path decision model employs a deep reinforcement learning (DRL) framework, specifically the DDPG algorithm. The state space is the particle feature vector output by the recognition model; the action space is a continuous or discrete combination of sorting methods (photoelectric, electric, magnetic, flotation) and parameters (voltage, magnetic field strength, reagent dosage); and the reward function is set as a comprehensive economic benefit function. It balances the value of metal recycling, energy costs, and reagent costs. Through millions of simulated sorting iterations, the model can ultimately recommend the most efficient sorting scheme for a given group of particles with specific characteristics.

[0041] S3. Ore Pretreatment and Intelligent Roughing and Waste Disposal: The raw ore is crushed to -100mm ("-" and "~" both mean less than or equal to) through a two-stage crushing and one-stage grinding circuit. The 40-100mm lump ore is fed into the TOMRA COM Tertiary intelligent sorting machine. The intelligent sorting machine's built-in hyperspectral imaging system quickly scans each stone and transmits the spectral data to the mineral particle real-time identification model deployed on the edge server for instant judgment. Category A (lithium-rich spodumene blocks with distinct surface spectral characteristics): Guided concentrate channel.

[0042] Category B (lean ore or waste rock, mainly felsic, without spodumene characteristic spectrum): directional tailings channel.

[0043] Results: With a processing capacity of 150 t / h, approximately 55% of the waste rock is removed in advance, the Li2O grade in the waste rock is less than 0.1%, and the lithium metal loss rate is less than 2%. The amount of ore fed into the mill is reduced by more than half, resulting in significant energy savings.

[0044] S4. Stage grinding and intelligent real-time sorting: The ore after waste disposal enters the closed-loop process of "three-stage grinding - four-stage intelligent sorting". The four-stage intelligent sorting is divided into several sorting units, and each sorting unit is equipped with a flow channel.

[0045] 1. First stage grinding and separation: The ore is ground to -3mm. After classification, the +0.5mm particle size ("+" means greater than or equal to) enters the intelligent high-voltage electrostatic separation unit. An online LIBS probe performs a rapid composition scan of the ore stream, identifying monomeric spodumene and intergrowths with good conductivity. The separation path decision model dynamically adjusts the electrostatic separator voltage (20-25kV) based on the ratio of iron (Fe) to aluminum (Al) elemental intensity on the particle surface (reflecting the degree of anephrite alteration). Qualified conductive particles (concentrate 1) are separated, while non-conductive particles (mainly quartz and feldspar) are discarded as tailings 1, and the intermediate products are returned to this stage for regrinding.

[0046] 2. Second-stage grinding and separation: The tailings and some middlings from the first stage are ground to -0.5mm. After classification, the 0.1-0.5mm particle size enters the combined intelligent separation unit. This unit integrates hyperspectral imaging lines and electromagnetic induction separation. The identification model distinguishes: (a) weakly magnetic nephrite-altered spodumene and biotite; (b) non-magnetic pure spodumene and feldspar. The decision model controls the electromagnetic separator to first separate (a) type magnetic products (as middlings 2, which will be processed separately later). The pure spodumene particles in the non-magnetic products are precisely separated by high-speed air valve array to obtain concentrate 2.

[0047] 3. Third-stage grinding and separation: The fine-grained middlings and intergrowths from the second stage are ground to -0.1 mm (P80 = 0.074 mm). At this point, the minerals are basically fully liberated, but the particle size is fine and the surface is complex. The ore enters the intelligent microbubble flotation column cluster. The pulp state is monitored in real time by an online particle size analyzer and potentiometer. Based on the identified surface properties of spodumene in the pulp (overall degree of nephrite alteration), the decision model dynamically adjusts the addition rate of the collector (environmentally friendly modified fatty acids) and the amount of frother, and optimizes the bubble size and flushing water flow rate in each zone of the flotation column. This achieves highly selective recovery of fine-grained spodumene (concentrate 3).

[0048] S5. Closed-Loop Feedback and Model Optimization: Online X-ray fluorescence (XRF) analyzers are installed on the concentrate and tailings chutes of each sorting unit to detect the content of elements such as Li, Na, K, and Fe in the product every 5 minutes. Actual recovery rate and grade data are compared with the predicted values ​​of the sorting path decision model at that time. Deviation data, along with the ore genetic characteristics at that time (obtained from online sensors), are input into the cloud model weekly for incremental learning. This allows the model to adaptively adjust the sorting strategy and maintain stable indicators when faced with minor fluctuations in ore properties caused by changes in the mine face.

[0049] S6. Resource utilization and wastewater purification: Tailings: The coarse-grained (-100+40mm) waste rock produced in step S3, after testing, is found to be non-toxic and harmless, and is all transported to the supporting building materials plant for the production of non-fired bricks or roadbed materials. The fine-grained flotation tailings, after thickening and dewatering, have their filter cake used for underground backfilling.

[0050] Wastewater: A small amount of wastewater generated during flotation is collected and sent to the treatment station. It undergoes three stages of treatment: neutralization precipitation (removal of heavy metal ions), ozone oxidation (degradation of residual organic matter), and activated carbon adsorption. All the purified water is reused for grinding and flotation operations, achieving zero discharge.

[0051] Overall effect: The average technical and economic indicators of this embodiment after one quarter of operation are compared with those of the traditional single flotation process as follows:

[0052] The above embodiments fully demonstrate that the present invention deeply integrates mineral gene theory with artificial intelligence recognition and decision-making, and achieves efficient, green and intelligent sorting of pegmatite-type lithium ore through the innovative process of "intelligent coarse polishing - staged grinding - real-time decision-making and sorting", resulting in significant technical, economic and social environmental benefits.

[0053] Figure 1This invention demonstrates a complete technical route, from constructing a mineral gene feature library and model training, to intelligent roughing and waste disposal, staged grinding and real-time sorting, and finally to closed-loop feedback optimization. The core sorting process (steps S3 and S4) illustrates how model-based intelligent decision-making drives the physical sorting process and achieves continuous optimization through feedback loops.

[0054] Figure 2 The core architecture of the intelligent real-time sorting unit was demonstrated. Minerals flow sequentially through an online detection zone on a conveyor belt, where sensor arrays (such as hyperspectral cameras) rapidly collect particle characteristic data and upload it to a central decision controller. The AI ​​model within the controller instantly completes identification and decision-making, issuing precise control commands to downstream programmable actuators (such as high-voltage electrodes, magnetic separators, and valve arrays) to sort particles of different properties into different product channels, achieving a millisecond-level closed loop of "perception-decision-execution".

[0055] In another embodiment of this application: like Figure 3 As shown in the previous embodiment, a green sorting system for pegmatite-type lithium ore with intelligent identification is implemented. The system includes a mineral gene feature analysis platform, a model training and computing server, an intelligent roughing and waste disposal unit, a crushing module and a staged grinding circuit, a cluster of intelligent real-time sorting units connected in series, an online analysis and feedback module, a central control system, and a tailings resource utilization and wastewater treatment module. The discharge port of the crushing module is connected to the feed port of the intelligent roughing and waste disposal unit, the staged grinding circuit is connected to the discharge port of the intelligent roughing and waste disposal unit, the cluster of intelligent real-time sorting units includes several intelligent real-time sorting units, which are connected in series after each grinding section of the staged grinding circuit, the online analysis and feedback module is used to analyze the sorted products and feed them back to the model training and computing server, and the central control unit is communicatively connected to each of the above units and modules. The mineral gene feature analysis platform is used in step S1, the model training and computing server is used to carry and run the real-time mineral particle identification model and sorting path decision model in step S2, the crushing module and stage grinding loop are used in step S5, the tailings resource utilization and wastewater treatment module is used in step S6, and the central control system is used to coordinate the operation of each module and data flow.

[0056] The intelligent coarse selection and waste disposal unit includes a feeder, a sensor array, and an intelligent actuator, which is used in step S3.

[0057] Each unit in the intelligent real-time sorting unit cluster includes an online detection module, an intelligent decision-making module, and a programmable sorting execution module, which are used in step S4.

[0058] Figure 3The presentation showcases the hardware components and data flow relationships of the entire sorting system. The system centers on a central control system, coordinating the mineral gene analysis platform, AI server, and main production line. The production line employs a modular design, sequentially comprising pretreatment, intelligent roughing and waste disposal, staged grinding, and a cluster of interconnected intelligent real-time sorting units. An online analysis module monitors the sorted products, feeding data back to the AI ​​server for model optimization, forming an intelligent closed loop. Tailings and wastewater treatment modules ensure the entire process is environmentally friendly.

[0059] The above description is merely an embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A green sorting method for pegmatite-type lithium ore with intelligent identification, characterized in that: Includes the following steps: S1. Construct a mineral gene feature library; S2. Training of intelligent identification and sorting decision model: Training a real-time mineral particle identification model and a sorting path decision model based on a mineral gene feature library; S3. Ore pretreatment and intelligent roughing and waste disposal: The raw ore is crushed, and the mineral particle real-time identification model identifies and separates low-value blocky materials. S4. Stage grinding and intelligent real-time sorting: The pre-selected ore is subjected to stage grinding, and an intelligent real-time sorting unit is set after each stage of grinding. The online detection sensor detects the mineral particle data after grinding and transmits it to the mineral particle real-time identification model and the sorting path decision model in real time to complete particle identification and sorting decision. According to the decision instructions, the particles are guided to the collection channel by the programmable intelligent actuator. S5. Closed-loop feedback and model optimization: The collected mineral particles are analyzed online, and the actual sorting effect is compared with the model prediction effect to form feedback data. The feedback data is used to perform incremental learning or online optimization of the real-time mineral particle identification model and the sorting path decision model on a regular or real-time basis.

2. The intelligent identification method for green sorting of pegmatite-type lithium ore according to claim 1, characterized in that: In step S1, core mineral gene features are extracted and a digital feature library is constructed.

3. The intelligent identification method for green sorting of pegmatite-type lithium ore according to claim 2, characterized in that: The core mineral gene features include mineralogy genes, physicochemical genes, and response behavior genes.

4. The intelligent identification method for green sorting of pegmatite-type lithium ore according to claim 1, characterized in that: In step S2, the input to the real-time mineral particle identification model is the particle flow image or spectral sequence obtained by the online detection system. The sorting path decision model takes the genetic characteristics of the particles identified by the real-time mineral particle identification model as input, and combines them with the preset target to determine a single or combined sorting method and its operating parameters for the particle from a predefined sorting method library.

5. The intelligent identification method for green sorting of pegmatite-type lithium ore according to claim 1, characterized in that: In step S3, after the raw ore is crushed, it is first screened and intelligently pre-selected. Based on the real-time mineral particle identification model, the blocky ore after medium or fine crushing is scanned and identified to separate the lithium-rich ore blocks, lithium-poor or waste rock blocks that have been liberated, so as to achieve the pre-disposal of coarse particles.

6. The intelligent identification method for green sorting of pegmatite-type lithium ore according to claim 1, characterized in that: It also includes step S6: resource utilization and wastewater purification, where the wastewater generated during the sorting process in step S4 is treated by coagulation, sedimentation, adsorption and advanced oxidation.

7. A green sorting system for pegmatite-type lithium ore with intelligent identification, implementing the method of claim 1, characterized in that: It includes a mineral gene feature analysis platform, a model training and computing server, an intelligent roughing and waste disposal unit, a crushing module and a staged grinding circuit, a cluster of intelligent real-time sorting units connected in series, an online analysis and feedback module, and a central control system. The discharge port of the crushing module is connected to the feed port of the intelligent roughing and waste disposal unit, and the staged grinding circuit is connected to the discharge port of the intelligent roughing and waste disposal unit. The cluster of intelligent real-time sorting units includes several intelligent real-time sorting units, which are connected in series after each grinding section of the staged grinding circuit.

8. The intelligent identification green sorting system for pegmatite-type lithium ore according to claim 7, characterized in that: The intelligent coarse selection and waste disposal unit includes a feeder, a sensor array, and an intelligent actuator.

9. A green sorting system for pegmatite-type lithium ore with intelligent identification according to claim 7, characterized in that: The intelligent real-time sorting unit includes an online detection module, an intelligent decision-making module, and a programmable sorting execution module. The intelligent decision-making module communicates with the model training and computing server.

10. A green sorting system for pegmatite-type lithium ore with intelligent identification according to claim 7, characterized in that: It also includes modules for tailings resource utilization and wastewater treatment.

Images