A method for predicting the purity of high-purity quartz of the alaskite type and application thereof

Abstract

The application discloses a white granite type high-purity quartz purity prediction method and application thereof, and belongs to the technical field of rock purity prediction. The prediction method encodes and assigns one or more prediction parameters, such as main mica types, rock structures, rock structures, feldspar types, quartz types, solid inclusions, liquid inclusions, solid inclusion quantity, aluminum impurity element content grading, titanium and lithium element content grading, calcium and phosphorus element content grading, potassium and sodium element content grading and grain factor grading, and then predicts the quartz purity in the ore based on a random forest classifier prediction model or a support vector machine prediction model. The comparison between the actual measurement and the prediction result proves that the prediction method has accurate and reliable prediction results, realizes the replacement of manual detection by artificial intelligence prediction, greatly reduces the manpower and material resources, and provides an efficient and accurate prediction method for quartz purity prediction.

Description

Technical Field

[0001] This invention belongs to the field of rock purity prediction technology, specifically, it relates to a method for predicting the purity of high-purity quartz of granite type and its application. Background Technology

[0002] High-purity quartz has a dual meaning: as a material, it refers to quartz sand and its products with a SiO2 purity of 99.99% or higher; as a resource, it refers to high-purity quartz sand raw material minerals that can be processed, purified, and separated. Different types of high-purity quartz sand have a wide range of applications, but high-purity quartz sand of 4N8 grade (SiO2 purity of 99.998%) and above, required by emerging industries such as semiconductor chips and solar photovoltaics, is an indispensable and irreplaceable key basic material. Its raw material minerals are scarce resources formed under extremely harsh geological conditions and with extremely limited distribution. The West relies on the unique white granite-type high-purity quartz sand sourced from Spruce Pine in North Carolina, USA, for its supply.

[0003] More and more industries have higher requirements for the purity of quartz sand. Taking the IOTA-CG high-end product of Unimin Corporation in the United States as an example, the SiO2 purity needs to reach 99.9981%. At present, domestic white granite-type high-purity quartz raw materials can be made into high-purity quartz sand with SiO2 purity greater than 99.998%. However, due to different ore sources and different purification methods, the highest purity results are unstable.

[0004] However, previous studies relied on human experience to evaluate various parameters, resulting in low efficiency, lack of theoretical basis, poor interpretability, and limitations in further subdivision. In particular, due to the numerous potential factors and parameters from different ore sources, their high degree of numerical dispersion, and the difficulty in recognizing patterns, it is impossible for humans to predict the final purity of a specific sample before the preparation process begins. Summary of the Invention

[0005] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0006] To solve the above problems, the present invention adopts the following technical solution.

[0007] A method for predicting the purity of high-purity quartz in granite-type silica includes the following steps: Step 1: Encode and assign values ​​to the prediction results and prediction parameters; Step 2: Predict the purity of quartz based on the prediction model and the encoding assignment in Step 1; In step 1, the prediction parameters include one or more of the following: main mica type, rock structure, rock texture, feldspar type, quartz type, solid inclusions, liquid inclusions, number of solid inclusions, aluminum impurity element content classification, titanium and lithium element content classification, calcium and phosphorus element content classification, and potassium and sodium element content classification.

[0008] Preferably, in the above prediction method, the process of encoding and assigning values ​​to the prediction results in step 1 is as follows: For products with purity A ≥ 4N85, a value of 1 is assigned. For products with a purity of 4N8≤A<4N85, the value is assigned to 2. For products with a purity of 4N7≤A<4N8, the value is assigned to 3. In the case where the product purity is 4N5≤A<4N7, the value is assigned to 4; If the product purity A < 4N5, the value is assigned to 5.

[0009] Preferably, in the above prediction method, the prediction model includes one of a random forest classifier prediction model and a support vector machine prediction model.

[0010] Preferably, in the above prediction method, when the prediction model is a random forest classifier prediction model, the parameters of the model are set as follows: To balance computational efficiency and model accuracy, the number of decision trees is set to 80. The feature sampling ratio is set to 0.35; The tree depth limit is set to 5; The node splitting threshold is set to 15; The sample sampling ratio was set to 0.9; Bootstrap sampling is selected; The category weighting strategy is set to 'balanced'; External assessment option enabled; Parallel computing option 1.

[0011] Preferably, in the above prediction method, when the prediction model is a support vector machine prediction model, the parameters of the model are set as follows: Regularization parameter C: balances computational efficiency and model accuracy, with a value of 2.5-4.5; Dual validation strategy: Randomize independently according to the result category, conduct 10 independent training iterations, and conduct independent validation using leave-one-out method; RBF kernel function construction: The mathematical expression is as follows ; Where: K is the sample similarity function, and the variable is x. i and x j， x iFor the i-th sample record (vector), x j Let be the j-th sample record (vector); exp is the natural exponential function; γ is the kernel function coefficient, with a search range of 0.08-0.14; The square of the Euclidean distance between samples; Key feature selection threshold: The feature selection method is adopted, namely: mutual information score + correlation analysis. The score = 0.01 is selected as the threshold to remove individual invalid features.

[0012] Preferably, in the above prediction method, in step 1, the main mica type encoding assignment process is as follows: for the case where the ore raw material contains only one type of mica, the value is assigned as 1; for the case where the ore raw material contains two types of mica, the value is assigned as 2; and for the case where the ore raw material contains three types of mica, the value is assigned as 3.

[0013] Preferably, in the above prediction method, in step 1, the rock structure coding assignment process is as follows: for raw ore rock minerals with a particle size of more than 10 mm, a value of 1 is assigned; for raw ore rock minerals with a particle size between 5 and 10 mm, a value of 2 is assigned; and for raw ore rock minerals with a particle size of less than 5 mm, a value of 3 is assigned.

[0014] Preferably, in the above prediction method, in step 1, the rock structure coding assignment process is as follows: for the case of strong deformation of the raw ore, a value of 1 is assigned; for the case of weak deformation of the raw ore, a value of 2 is assigned; and for the case of no deformation of the raw ore, a value of 3 is assigned. The term "strong deformation" refers to a situation where the dynamic metamorphism is strong, the mineral crystals have undergone plastic deformation, are elongated, and exhibit obvious deformation characteristics. The term "weak deformation" refers to minerals that have undergone weak dynamic metamorphism and exhibit insignificant deformation characteristics in their crystals. The phrase "no deformation" refers to the absence of any alteration or degradation process, i.e., the absence of the phenomena described above.

[0015] Preferably, in the above prediction method, in step 1, the feldspar type coding assignment process is as follows: for the case where the ore raw material contains only one type of feldspar, the value is assigned as 1; for the case where the ore raw material contains two types of feldspar, the value is assigned as 2; for the case where the ore raw material contains three types of feldspar, the value is assigned as 3; and for other cases, the value is assigned as 4.

[0016] Preferably, in the above prediction method, in step 1, the quartz type coding assignment process is as follows: if the ore raw material contains one type of quartz, the value is assigned as 1; if the ore raw material contains two types of quartz, the value is assigned as 2; if the ore raw material contains three types of quartz, the value is assigned as 3; and in other cases, the value is assigned as 4.

[0017] Preferably, in the above prediction method, in step 1, the solid inclusion coding assignment process is as follows: if the solid inclusions in the raw ore do not contain dark minerals, the value is assigned as 1; if the solid inclusions in the raw ore contain dark minerals, the value is assigned as 2; if the solid inclusions in the raw ore contain mixed minerals, the value is assigned as 3.

[0018] Preferably, in the above prediction method, in step 1, the liquid inclusion coding assignment process is as follows: if the raw ore does not contain liquid inclusions, assign a value of 1; if the raw ore contains a small amount of liquid inclusions, assign a value of 2; if the raw ore contains a large amount of liquid inclusions, assign a value of 3. Small amount: refers to a liquid inclusion content of less than 8 per mm. 2 ; "More than 8 liquid inclusions / mm" 2 .

[0019] Preferably, in the above prediction method, in step 1, the solid inclusion quantity encoding and assignment process is as follows: for the case where the raw ore does not contain solid inclusions, the value is assigned as 1; for the case where the raw ore contains a small amount of solid inclusions, the value is assigned as 2; and for the case where the raw ore contains a large amount of solid inclusions, the value is assigned as 3. Small amount: refers to a solid inclusion content of less than 5 per cm³ 2 ; "More than 5 solid inclusions / cm³" 2 .

[0020] Preferably, in the above prediction method, in step 1, the classification of aluminum impurity element content is coded and assigned values. Specifically, for aluminum impurity element content below 8 × 10⁻⁶, the classification is coded and assigned values. -6 For the case of μg / g, a value of 1 is assigned; for aluminum impurity element content below 14×10 μg / g, the value is 1. -6 For the case of μg / g, the value is assigned to 2; for aluminum impurity element content below 20×10 -6 For the case of μg / g, the value is assigned to 3; for aluminum impurity element content below 30×10 μg / g, the value is assigned to 3. -6 For the case of μg / g, the value is assigned to 4; for aluminum impurity element content higher than 30×10 μg / g, the value is assigned to 4; -6 In this case, the value is assigned to 5; Preferably, in the above prediction method, in step 1, the classification of titanium and lithium content is encoded and assigned values. Specifically, the value is 1 for low titanium and low lithium; 2 for low titanium and high lithium; 3 for high titanium and low lithium; and 4 for high titanium and high lithium. Low titanium: refers to a titanium content ≤3μg / g; Low lithium: refers to a lithium content ≤1μg / g; High titanium: refers to a titanium content >3μg / g; High lithium: refers to a lithium content >1μg / g.

[0021] Preferably, in the above prediction method, in step 1, the classification of calcium and phosphorus content is coded and assigned values. Specifically, for low calcium and high phosphorus, the value is 2; for high calcium and low phosphorus, the value is 3; and for high calcium and high phosphorus, the value is 4. Low calcium: refers to a calcium content ≤3μg / g; Low phosphorus: refers to a phosphorus content ≤1μg / g; High calcium: refers to a calcium content >3μg / g. High phosphorus: refers to a phosphorus content >1μg / g.

[0022] Preferably, in the above prediction method, in step 1, the classification of potassium and sodium impurity element content is coded and assigned values. Specifically, for the case of low potassium and low sodium, the value is 1; for the case of high potassium and low sodium, the value is 2; for the case of low potassium and high sodium, the value is 3; and for the case of high potassium and high sodium, the value is 4. Low potassium: refers to a potassium content ≤1μg / g; Low sodium: refers to a sodium content ≤1μg / g; High potassium: refers to a potassium content >1μg / g; High sodium content refers to a sodium content greater than 1 μg / g.

[0023] In addition, the present invention also provides an application of the above-mentioned method for predicting the purity of high-purity quartz of granite type.

[0024] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) This invention enables the separate study of high-purity quartz raw materials of leucogranite type based on existing real data and geological and mineral theories, distinguishing them from other types such as granite pegmatite, and thus realizing the prediction of quartz product purity supported by intelligent algorithms. It effectively solves the problems of not being able to quickly determine the processing potential, the inability to conduct manual evaluation or the low efficiency in the purification process of a large number of products, avoids the problem of poor interpretability of artificial intelligence algorithms without geological significance, is conducive to the prediction and evaluation of the processing potential of different batches of ore, and is conducive to significantly reducing the cost caused by processing batches of samples with low potential and failing to obtain high-purity results.

[0025] (2) This invention overcomes the problems of classifying different ore sources and algorithm prediction. It is beneficial to confirm the highest purity endowment after processing with high confidence in the early stage of processing, realize the determination of the raw material processing precision potential, significantly reduce the waste of manpower and material resources in the processing process, and also help to lock in specific high-purity sample entities. It provides a method and technology for locking in high-purity quartz sand samples in advance, improves the ability to discover high-concentration mineral products, thereby improving the production capacity of concentrate products and realizing the overall economic value of high-purity quartz processing. Attached Figure Description

[0026] Figure 1 This is a confusion matrix diagram generated by the random forest classifier in Example 1; Figure 2 This is a precision-recall curve (PR curve) of the random forest classifier in Example 1. Figure 3 The receiver operating characteristic (ROC) curve of the random forest classifier in Example 1 is shown. Figure 4 This is a heatmap showing the correlation of features predicted by the support vector machine in Example 2. in: Prediction category 1 represents product purity A≥4N85; Prediction category 2 represents product purity 4N8≤A<4N85; Predicted purity of Category 3 products: 4N7 ≤ A < 4N8; Predicted purity of product category 4: 4N5≤A<4N7; Predicted purity of product category 5: A < 4N5; Figure 5 The presents the receiver operating characteristic (ROC) curve of the support vector machine classifier in Example 2. Figure 6 This is a precision-recall curve (PR curve) of the support vector machine classifier in Example 2. Detailed Implementation

[0027] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0028] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0029] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that mutually excludes other embodiments. The present invention provides the following embodiments.

[0030] Example 1 This embodiment provides a prediction method for predicting the purity of high-purity quartz of granite type. The prediction method includes the following steps: Step 1, assign the encoding value; Specifically, encoding and assignment include encoding and assigning values ​​to the prediction results and prediction parameters.

[0031] The process of encoding and assigning values ​​to the prediction results is as follows: The highest output purity of all test samples is qualitatively coded as follows: 1 for final product purity A ≥ 4N85; 2 for final product purity 4N8 ≤ A < 4N85; 3 for final product purity 4N7 ≤ A < 4N8; 4 for final product purity 4N5 ≤ A < 4N7; and 5 for final product purity A < 4N5.

[0032] In addition, the process of encoding and assigning values ​​to the prediction parameters is as follows: In this embodiment, the prediction parameters include the main mica type, rock structure, rock texture, feldspar type, quartz type, solid inclusions, liquid inclusions, number of solid inclusions, aluminum impurity element content classification, titanium and lithium element content classification, calcium and phosphorus element content classification, and potassium and sodium element content classification. By encoding and assigning values ​​to the above prediction parameters, the purity of quartz can be predicted using the prediction model. The specific process is as follows.

[0033] The main mica types in this type of ore raw material are qualitatively coded as follows: if the ore raw material contains only one type of mica, a value of 1 is assigned; if the ore raw material contains two types of mica, a value of 2 is assigned; and if the ore raw material contains three types of mica, a value of 3 is assigned. The rock structure of this type of ore raw material is qualitatively coded as follows: for raw ore rock minerals with a particle size of more than 10 mm, a value of 1 is assigned; for raw ore rock minerals with a particle size between 5 and 10 mm, a value of 2 is assigned; and for raw ore rock minerals with a particle size of less than 5 mm, a value of 3 is assigned. The rock structure of this type of ore raw material is qualitatively coded as follows: strong deformation of the raw ore is assigned a value of 1; weak deformation of the raw ore is assigned a value of 2; and no deformation of the raw ore is assigned a value of 3. In this embodiment, strong deformation refers to strong dynamic metamorphism, in which the mineral crystals have undergone plastic deformation and are elongated, with obvious deformation characteristics; weak deformation refers to weak dynamic metamorphism, in which the deformation characteristics of the mineral crystals are not obvious; and no deformation refers to no metamorphism, i.e., no of the above phenomena have occurred.

[0034] The feldspar type of this type of ore raw material is qualitatively coded as follows: if the ore raw material contains only one type of feldspar, a value of 1 is assigned; if the ore raw material contains two types of feldspar, a value of 2 is assigned; if the ore raw material contains three types of feldspar, a value of 3 is assigned; and if other cases are assigned a value of 4. The quartz type of this type of ore raw material is qualitatively coded as follows: if the ore raw material contains one type of quartz, a value of 1 is assigned; if the ore raw material contains two types of quartz, a value of 2 is assigned; if the ore raw material contains three types of quartz, a value of 3 is assigned; and if the other cases are not, a value of 4 is assigned. The solid inclusions of this type of ore raw material are qualitatively coded as follows: if the solid inclusions in the raw ore do not contain dark minerals, a value of 1 is assigned; if the solid inclusions in the raw ore contain dark minerals, a value of 2 is assigned; and if the solid inclusions in the raw ore contain mixed minerals, a value of 3 is assigned. The liquid inclusion status of this type of ore raw material is qualitatively coded as follows: 1 is assigned to the case where the raw ore contains no liquid inclusions; 2 is assigned to the case where the raw ore contains a small amount of liquid inclusions; and 3 is assigned to the case where the raw ore contains a large amount of liquid inclusions. In this embodiment, "small amount" refers to a liquid inclusion content of less than 8 inclusions / mm². 2 "More than": refers to a liquid inclusion content higher than 8 per mm. 2 .

[0035] The quantity of solid inclusions in this type of ore raw material is qualitatively coded as follows: 1 is assigned to the case where the raw ore contains no solid inclusions; 2 is assigned to the case where the raw ore contains a small amount of solid inclusions; and 3 is assigned to the case where the raw ore contains a large amount of solid inclusions. In this embodiment, "small amount" refers to a solid inclusion content of less than 5 inclusions / cm³. 2 "More than 5 solid inclusions / cm³" 2 .

[0036] The grading of aluminum impurity content (DQP-Al) in this type of ore raw material is qualitatively coded. Specifically, for aluminum impurity content below 8 × 10⁻⁶, the grading is as follows:-6 For the case of μg / g, a value of 1 is assigned; for aluminum impurity element content below 14×10 μg / g, the value is 1. -6 For the case of μg / g, the value is assigned to 2; for aluminum impurity element content below 20×10 -6 For the case of μg / g, the value is assigned to 3; for aluminum impurity element content below 30×10 μg / g, the value is assigned to 3. -6 For the case of μg / g, the value is assigned to 4; for aluminum impurity element content higher than 30×10 μg / g, the value is assigned to 4; -6 In the case of μg / g, the value is assigned to 5.

[0037] The titanium and lithium content (Ti, Li) of this type of ore raw material is qualitatively coded as follows: low titanium and low lithium is assigned a value of 1; low titanium and high lithium is assigned a value of 2; high titanium and low lithium is assigned a value of 3; and high titanium and high lithium is assigned a value of 4. In this embodiment, low titanium means titanium content ≤ 3 μg / g, and low lithium means lithium content ≤ 1 μg / g; high titanium means titanium content > 3 μg / g, and high lithium means lithium content > 1 μg / g.

[0038] The calcium and phosphorus content (Ca, P) of this type of ore raw material is qualitatively coded as follows: low calcium and high phosphorus are assigned a value of 2; high calcium and low phosphorus are assigned a value of 3; and high calcium and high phosphorus are assigned a value of 4. In this embodiment, low calcium means calcium content ≤ 3 μg / g and low phosphorus means phosphorus content ≤ 1 μg / g; high calcium means calcium content > 3 μg / g and high phosphorus means phosphorus content > 1 μg / g.

[0039] The potassium and sodium impurity element content (K, Na) of this type of ore raw material is qualitatively coded as follows: low potassium and low sodium are assigned a value of 1; high potassium and low sodium are assigned a value of 2; low potassium and high sodium are assigned a value of 3; and high potassium and high sodium are assigned a value of 4. In this embodiment, low potassium means potassium content ≤ 1 μg / g, and low sodium means sodium content ≤ 1 μg / g; high potassium means potassium content > 1 μg / g, and high sodium means sodium content > 1 μg / g.

[0040] Step 2: Predict the purity of quartz using a random forest classifier prediction model.

[0041] In step 2, the parameters of the random forest classifier are first set. In this embodiment, the specific parameter settings are as follows.

[0042] The number of decision trees (n_estimators) balances computational efficiency and model accuracy, and is set to 80. The feature sampling ratio (max_features) is set to 0.35; The tree depth limit (max_depth) is set to 5; The node splitting threshold (min_samples_split) is set to 15. The sample sampling ratio (max_samples) is set to 0.9; Select "Yes" for Bootstrap sampling. Class weighting strategy (class_weight): Select 'balanced'; Out-of-bag evaluation (oob_score) is enabled. Parallel computation (n_jobs): Select 1 (TRUE).

[0043] In this embodiment, the random forest classifier defines the confusion matrix using random numbers, automatically calculates feature weights, and reads the code with the highest output purity as the classification label to form a classification label list. The model is pre-trained on the data, and a model record file is generated. Based on the visualized calculation results, each parameter (classifier) ​​is further adjusted until the accuracy reaches above 80%.

[0044] In this embodiment, taking 33 groups of samples as an example, the purity of quartz in the samples is predicted through the above steps. The purity of quartz in each sample is then tested manually, and the results are compared with the predicted values. Specific results are shown in Table 1, where the confusion matrix results are as follows: Figure 1 As shown in the figure, the predicted PR curve is as follows: Figure 2 As shown in the figure, the ROC curve is as follows: Figure 3 As shown.

[0045] Table 1. Results of quartz purity prediction using the random forest classifier

[0046] The results above show that the objective accuracy of the prediction results in this embodiment is 90%. Result type 5 is high-purity quartz below 4N5 (A < 99.995% purity), result type 4 is high-purity quartz of 4N7 (99.995% ≤ A < 99.997% purity), result type 3 is high-purity quartz of 4N8 (99.997% ≤ A < 99.998% purity), result type 2 is high-purity quartz of 4N85 (99.998% ≤ A < 99.9985% purity), and result type 1 is high-purity quartz above 4N85 (A ≥ 99.9985% purity). This embodiment achieves more refined classification and prediction for high-purity quartz samples of 99.997%-99.999%.

[0047] Example 2 This embodiment provides a prediction method for predicting the purity of high-purity quartz of granite type. The prediction method includes the following steps: Step 1, assign the encoding value; Specifically, encoding and assignment include encoding and assigning values ​​to the prediction results and prediction parameters.

[0048] The process of encoding and assigning values ​​to the prediction results is as follows: The highest output purity of all test samples is qualitatively coded as follows: 1 for final product purity A ≥ 4N85; 2 for final product purity 4N8 ≤ A < 4N85; 3 for final product purity 4N7 ≤ A < 4N8; 4 for final product purity 4N5 ≤ A < 4N7; and 5 for final product purity A < 4N5.

[0049] In addition, the process of encoding and assigning values ​​to the prediction parameters is as follows: In this embodiment, the prediction parameters include the main mica type, rock structure, rock texture, feldspar type, quartz type, solid inclusions (dark minerals), liquid inclusions, number of solid inclusions, aluminum impurity element content classification, titanium and lithium element content classification, calcium and phosphorus element content classification, and potassium and sodium element content classification. By encoding and assigning values ​​to the above prediction parameters, the purity of quartz can be predicted using the prediction model. The specific process is as follows.

[0050] The main mica types in this type of ore raw material are qualitatively coded as follows: if the ore raw material contains only one type of mica, a value of 1 is assigned; if the ore raw material contains two types of mica, a value of 2 is assigned; and if the ore raw material contains three types of mica, a value of 3 is assigned. The rock structure of this type of ore raw material is qualitatively coded as follows: for raw ore rock minerals with a particle size of more than 10 mm, a value of 1 is assigned; for raw ore rock minerals with a particle size between 5 and 10 mm, a value of 2 is assigned; and for raw ore rock minerals with a particle size of less than 5 mm, a value of 3 is assigned. The rock structure of this type of ore raw material is qualitatively coded as follows: strong deformation of the raw ore is assigned a value of 1; weak deformation of the raw ore is assigned a value of 2; and no deformation of the raw ore is assigned a value of 3. In this embodiment, strong deformation refers to strong dynamic metamorphism, in which the mineral crystals have undergone plastic deformation and are elongated, with obvious deformation characteristics; weak deformation refers to weak dynamic metamorphism, in which the deformation characteristics of the mineral crystals are not obvious; and no deformation refers to no metamorphism, i.e., no of the above phenomena have occurred.

[0051] The feldspar type of this type of ore raw material is qualitatively coded as follows: if the ore raw material contains only one type of feldspar, a value of 1 is assigned; if the ore raw material contains two types of feldspar, a value of 2 is assigned; if the ore raw material contains three types of feldspar, a value of 3 is assigned; and if other cases are assigned a value of 4. The quartz type of this type of ore raw material is qualitatively coded as follows: if the ore raw material contains one type of quartz, a value of 1 is assigned; if the ore raw material contains two types of quartz, a value of 2 is assigned; if the ore raw material contains three types of quartz, a value of 3 is assigned; and if the other cases are not, a value of 4 is assigned. The presence of dark minerals in solid inclusions of this type of ore raw material is qualitatively coded as follows: if the solid inclusions in the raw ore do not contain dark minerals, a value of 1 is assigned; if the solid inclusions in the raw ore contain dark minerals, a value of 2 is assigned; and if the solid inclusions in the raw ore contain mixed minerals, a value of 3 is assigned. The liquid inclusion status of this type of ore raw material is qualitatively coded as follows: 1 is assigned to the case where the raw ore contains no liquid inclusions; 2 is assigned to the case where the raw ore contains a small amount of liquid inclusions; and 3 is assigned to the case where the raw ore contains a large amount of liquid inclusions. In this embodiment, "small amount" refers to a liquid inclusion content of less than 8 inclusions / mm². 2 "More than": refers to a liquid inclusion content higher than 8 per mm. 2 .

[0052] The quantity of solid inclusions in this type of ore raw material is qualitatively coded as follows: 1 is assigned to the case where the raw ore contains no solid inclusions; 2 is assigned to the case where the raw ore contains a small amount of solid inclusions; and 3 is assigned to the case where the raw ore contains a large amount of solid inclusions. In this embodiment, "small amount" refers to a solid inclusion content of less than 5 inclusions / cm³. 2 "More than 5 solid inclusions / cm³" 2 .

[0053] The grading of aluminum impurity content (DQP-Al) in this type of ore raw material is qualitatively coded. Specifically, for aluminum impurity content below 8 × 10⁻⁶, the grading is as follows: -6 For the case of μg / g, a value of 1 is assigned; for aluminum impurity element content below 14×10 μg / g, the value is 1. -6 For the case of μg / g, the value is assigned to 2; for aluminum impurity element content below 20×10 -6 For the case of μg / g, the value is assigned to 3; for aluminum impurity element content below 30×10 μg / g, the value is assigned to 3. -6 For the case of μg / g, the value is assigned to 4; for aluminum impurity element content higher than 30×10 μg / g, the value is assigned to 4; -6 In the case of , the value is assigned to 5.

[0054] The titanium and lithium content (Ti, Li) of this type of ore raw material is qualitatively coded as follows: low titanium and low lithium is assigned a value of 1; low titanium and high lithium is assigned a value of 2; high titanium and low lithium is assigned a value of 3; and high titanium and high lithium is assigned a value of 4. In this embodiment, low titanium means titanium content ≤ 3 μg / g, and low lithium means lithium content ≤ 1 μg / g; high titanium means titanium content > 3 μg / g, and high lithium means lithium content > 1 μg / g.

[0055] The calcium and phosphorus content (Ca, P) of this type of ore raw material is qualitatively coded as follows: low calcium and high phosphorus are assigned a value of 2; high calcium and low phosphorus are assigned a value of 3; and high calcium and high phosphorus are assigned a value of 4. In this embodiment, low calcium means calcium content ≤ 3 μg / g and low phosphorus means phosphorus content ≤ 1 μg / g; high calcium means calcium content > 3 μg / g and high phosphorus means phosphorus content > 1 μg / g.

[0056] The potassium and sodium impurity element content (K, Na) of this type of ore raw material is qualitatively coded as follows: low potassium and low sodium are assigned a value of 1; high potassium and low sodium are assigned a value of 2; low potassium and high sodium are assigned a value of 3; and high potassium and high sodium are assigned a value of 4. In this embodiment, low potassium means potassium content ≤ 1 μg / g, and low sodium means sodium content ≤ 1 μg / g; high potassium means potassium content > 1 μg / g, and high sodium means sodium content > 1 μg / g.

[0057] Step 2: Predict the purity of quartz using a support vector machine algorithm prediction model.

[0058] In step 2, the support vector machine parameters are designed as follows.

[0059] Regularization parameter (C): Balances computational efficiency and model accuracy, with a value of 2.5-4.5; Dual validation strategy: Eliminate the influence of randomness; Randomize independently according to the result category, use 10 independent training iterations, and use leave-one-out method for independent validation; Final optimal parameters: Based on multiple rounds of training, the class weights were finely adjusted. For this sample set, C=2.5 achieved the best results.

[0060] RBF kernel function construction: The mathematical expression is K(x_i, x_j) = exp(-γ * ||x_i - x_j||²); Where: K is the sample similarity function, and the variable is x. i and x j， x iFor the i-th sample record (vector), x j Let be the j-th sample record (vector); exp is the natural exponential function; γ is the kernel function coefficient, with a search range of 0.08-0.14; ||x i - x j ||² represents the squared Euclidean distance between samples; Key feature selection threshold: The feature selection method is adopted, namely: mutual information score + correlation analysis. The score = 0.01 is selected as the threshold to remove individual invalid features.

[0061] The model is pre-trained on the data, and a model log file is generated. Based on the visualized calculation results, the classifier is further refined until the accuracy reaches above 80.36%.

[0062] In this embodiment, the quartz purity of 33 samples in Example 1 was predicted using a support vector machine (SVM) algorithm. The prediction results are shown in Table 2. Furthermore, the heatmap of the correlation between the SVM prediction features in this embodiment is shown below. Figure 4 As shown in the figure, the predicted ROC curve is as follows: Figure 5 As shown in the figure, the predicted PR curve is as follows: Figure 6 As shown.

[0063] Table 2. Support Vector Machine Prediction Results of Quartz Purity Main types of mica rock structure Rock structure Feldspar type Quartz type solid inclusions Liquid inclusions Number of solid inclusions DQP-AI (Grading) Ti, Li (gradation) Ca, P (gradation) K, Na (gradation) Known results Prediction results Is it correct? 2 2 3 3 2 3 1 2 2 1 9 1 2 2 yes 2 1 3 3 2 3 1 1 2 1 6 2 2 2 yes 2 1 3 3 1 1 1 1 2 2 5 4 3 3 yes 2 2 2 1 1 1 1 1 3 2 9 3 3 3 yes 2 2 4 3 2 1 1 1 5 4 8 4 5 5 yes 1 2 3 3 2 1 2 1 5 1 8 4 5 4 no 1 1 4 1 1 1 1 1 4 1 5 3 4 4 yes 2 1 3 3 2 3 1 1 5 1 8 4 4 4 yes 1 2 3 1 3 1 1 1 5 3 8 4 5 5 yes 2 2 3 3 2 3 1 1 5 3 9 4 5 5 yes 2 2 3 3 2 1 1 1 5 3 7 4 5 5 yes 2 2 3 3 2 3 1 1 5 3 7 4 5 5 yes 2 2 3 3 2 3 1 1 5 3 8 4 5 5 yes 2 1 3 3 2 1 1 1 5 3 9 4 5 5 yes 2 1 3 3 2 3 1 1 5 1 5 3 4 4 yes 1 1 3 3 1 1 1 1 5 1 5 4 4 4 yes 1 2 3 1 2 1 2 2 5 1 5 4 4 4 yes 2 1 3 3 1 1 1 2 5 1 5 4 4 4 yes 2 1 3 3 1 1 1 1 5 1 5 4 4 4 yes 2 1 3 3 2 1 1 1 5 1 6 4 4 4 yes 2 2 3 3 1 3 2 1 3 2 9 3 3 3 yes 1 2 3 1 1 1 1 1 3 2 9 3 3 3 yes 2 1 3 3 2 3 1 1 3 2 9 3 3 3 yes 1 2 2 3 2 1 1 1 3 2 6 3 3 3 yes 1 2 4 3 2 1 1 1 3 2 6 3 3 3 yes 2 2 4 3 2 1 1 1 3 2 9 3 3 3 yes 2 2 3 3 2 3 1 1 2 1 6 2 2 2 yes 2 2 3 3 2 3 2 1 2 1 9 2 2 2 yes 2 2 3 3 2 3 1 1 2 1 5 2 2 2 yes 2 2 3 3 2 3 1 1 2 1 4 3 2 2 yes 2 2 3 3 2 3 1 1 2 1 6 3 2 3 no 2 1 3 3 2 1 1 1 2 1 5 4 2 3 no 2 3 3 1 1 1 1 1 2 1 5 4 2 3 no The results above show that the objective accuracy rate of the prediction results in this embodiment is 87%. Result type 5 is high-purity quartz below 4N5 (A < 99.995% purity), result type 4 is high-purity quartz of 4N7 (99.995% ≤ A < 99.997% purity), result type 3 is high-purity quartz of 4N8 (99.997% ≤ A < 99.998% purity), result type 2 is high-purity quartz of 4N85 (99.998% ≤ A < 99.9985% purity), and result type 1 is high-purity quartz above 4N85 (A ≥ 99.9985% purity). This embodiment achieves more refined classification and prediction for high-purity quartz samples of 99.997%-99.999%.

[0064] The above description, in conjunction with specific embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, several simple deductions or substitutions can be made without departing from the concept of the present invention, and all such deductions or substitutions should be considered to fall within the scope of protection defined by the claims submitted herein.

Claims

1. A method for predicting the purity of high-purity quartz of granite type, characterized in that, Includes the following steps: Step 1: Encode and assign values ​​to the prediction results and prediction parameters; Step 2: Predict the purity of quartz based on the prediction model and the encoding assignment in Step 1; In step 1, the prediction parameters include one or more of the following: main mica type, rock structure, rock texture, feldspar type, quartz type, solid inclusions, liquid inclusions, number of solid inclusions, aluminum impurity element content classification, titanium and lithium element content classification, calcium and phosphorus element content classification, and potassium and sodium element content classification.

2. The method for predicting the purity of high-purity quartz of granite type according to claim 1, characterized in that, In step 1, the process of encoding and assigning values ​​to the prediction results is as follows: For products with purity A ≥ 4N85, a value of 1 is assigned. For products with a purity of 4N8≤A<4N85, the value is assigned to 2. For products with a purity of 4N7≤A<4N8, the value is assigned to 3. In the case where the product purity is 4N5≤A<4N7, the value is assigned to 4; If the product purity A < 4N5, the value is assigned to 5.

3. The method for predicting the purity of high-purity quartz of granite type according to claim 2, characterized in that, The prediction model includes one of the random forest classifier prediction model and the support vector machine prediction model.

4. The method for predicting the purity of high-purity quartz of granite type according to claim 3, characterized in that, When the prediction model is a random forest classifier, the parameters of the model are set as follows: To balance computational efficiency and model accuracy, the number of decision trees is set to 80. The feature sampling ratio is set to 0.35; The tree depth limit is set to 5; The node splitting threshold is set to 15; The sample sampling ratio was set to 0.9; Bootstrap sampling is selected; The category weighting strategy is set to 'balanced'; External assessment option enabled; Parallel computing option 1.

5. The method for predicting the purity of high-purity quartz of granite type according to claim 3, characterized in that, When the prediction model is a support vector machine prediction model, the parameters of the model are set as follows: Regularization parameter C: balances computational efficiency and model accuracy, with a value of 2.5-4.5; Dual validation strategy: Randomize independently according to the result category, conduct 10 independent training iterations, and conduct independent validation using leave-one-out method; RBF kernel function construction: The mathematical expression is as follows ; Where: K is the sample similarity function, and the variable is... and x j， x i For the i-th sample record (vector), x j Let j be the j-th sample record (vector); exp is the natural exponential function; The search range for kernel function coefficients is 0.08-0.14; The square of the Euclidean distance between samples; Key feature selection threshold: The feature selection method is adopted, namely: mutual information score + correlation analysis. The score = 0.01 is selected as the threshold to remove individual invalid features.

6. The method for predicting the purity of high-purity quartz of granite type according to claim 1, characterized in that, In step 1, the main mica type encoding assignment process is as follows: for the case where the ore raw material contains only one type of mica, the value is assigned as 1; for the case where the ore raw material contains two types of mica, the value is assigned as 2; and for the case where the ore raw material contains three types of mica, the value is assigned as 3.

7. The method for predicting the purity of high-purity quartz of granite type according to claim 1, characterized in that, In step 1, the rock structure coding assignment process is as follows: for raw ore rock minerals with a particle size of more than 10 mm, the value is assigned as 1; for raw ore rock minerals with a particle size between 5 and 10 mm, the value is assigned as 2; and for raw ore rock minerals with a particle size of less than 5 mm, the value is assigned as 3.

8. The method for predicting the purity of high-purity quartz of granite type according to claim 1, characterized in that, In step 1, the rock structure coding assignment process is as follows: for raw ore that undergoes strong deformation, a value of 1 is assigned; for raw ore that undergoes weak deformation, a value of 2 is assigned; and for raw ore that does not undergo deformation, a value of 3 is assigned. The term "strong deformation" refers to a situation where the dynamic metamorphism is strong, the mineral crystals have undergone plastic deformation, are elongated, and exhibit obvious deformation characteristics. The term "weak deformation" refers to minerals that have undergone weak dynamic metamorphism and exhibit insignificant deformation characteristics in their crystals. The phrase "no deformation" refers to the absence of any alteration or degradation process, i.e., the absence of the phenomena described above.

9. The method for predicting the purity of high-purity quartz of granite type according to claim 1, characterized in that, In step 1, the feldspar type coding process is as follows: if the ore raw material contains only one type of feldspar, the value is 1; if the ore raw material contains two types of feldspar, the value is 2; if the ore raw material contains three types of feldspar, the value is 3; and for other cases, the value is 4.

10. The method for predicting the purity of high-purity quartz of granite type according to claim 1, characterized in that, In step 1, the quartz type coding assignment process is as follows: if the ore raw material contains one type of quartz, the value is 1; if the ore raw material contains two types of quartz, the value is 2; if the ore raw material contains three types of quartz, the value is 3; and for other cases, the value is 4.

11. The method for predicting the purity of high-purity quartz of granite type according to claim 1, characterized in that, In step 1, the solid inclusion coding assignment process is as follows: if the solid inclusions in the raw ore do not contain dark minerals, the value is assigned as 1; if the solid inclusions in the raw ore contain dark minerals, the value is assigned as 2; if the solid inclusions in the raw ore contain mixed minerals, the value is assigned as 3.

12. The method for predicting the purity of high-purity quartz of granite type according to claim 1, characterized in that, In step 1, the liquid inclusion coding assignment process is as follows: if the raw ore does not contain liquid inclusions, the value is 1; if the raw ore contains a small amount of liquid inclusions, the value is 2; if the raw ore contains a large amount of liquid inclusions, the value is 3. Small amount: refers to a liquid inclusion content of less than 8 per mm. 2 ; "More than 8 liquid inclusions / mm" 2 .

13. The method for predicting the purity of high-purity quartz of granite type according to claim 1, characterized in that, In step 1, the solid inclusion quantity coding process is as follows: if the raw ore does not contain any solid inclusions, the value is 1; if the raw ore contains a small amount of solid inclusions, the value is 2; if the raw ore contains a large amount of solid inclusions, the value is 3. Small amount: refers to a solid inclusion content of less than 5 per cm³ 2 ; "More than 5 solid inclusions / cm³" 2 .

14. The method for predicting the purity of high-purity quartz of granite type according to claim 1, characterized in that, In step 1, the grading of aluminum impurity element content is coded and assigned values. Specifically, for aluminum impurity element content below 8 × 10⁻⁶, the grading is coded and assigned values. -6 For the case of μg / g, a value of 1 is assigned; for aluminum impurity element content below 14×10 μg / g, the value is 1. -6 For the case of μg / g, the value is assigned to 2; for aluminum impurity element content below 20×10 -6 For the case of μg / g, the value is assigned to 3; for aluminum impurity element content below 30×10 μg / g, the value is assigned to 3. -6 For the case of μg / g, the value is assigned to 4; for aluminum impurity element content higher than 30×10 μg / g, the value is assigned to 4; -6 In the case of μg / g, the value is assigned to 5.

15. The method for predicting the purity of high-purity quartz of granite type according to claim 1, characterized in that, In step 1, the grading of titanium and lithium content is coded and assigned values. Specifically, the value is 1 for low titanium and low lithium; 2 for low titanium and high lithium; 3 for high titanium and low lithium; and 4 for high titanium and high lithium. Low titanium: refers to a titanium content ≤3μg / g; Low lithium: refers to a lithium content ≤1μg / g; High titanium: refers to a titanium content >3μg / g; High lithium: refers to a lithium content >1μg / g.

16. The method for predicting the purity of high-purity quartz of granite type according to claim 1, characterized in that, In step 1, the calcium and phosphorus content is graded and coded with values: 2 for low calcium and high phosphorus, 3 for high calcium and low phosphorus, and 4 for high calcium and high phosphorus. Low calcium: refers to a calcium content ≤3μg / g; Low phosphorus: refers to a phosphorus content ≤1μg / g; High calcium: refers to a calcium content >3μg / g. High phosphorus: refers to a phosphorus content >1μg / g.

17. The method for predicting the purity of high-purity quartz of granite type according to claim 1, characterized in that, In step 1, the classification of potassium and sodium impurity element content is coded and assigned values. Specifically, the value is 1 for low potassium and low sodium; 2 for high potassium and low sodium; 3 for low potassium and high sodium; and 4 for high potassium and high sodium. Low potassium: refers to a potassium content ≤1μg / g; Low sodium: refers to a sodium content ≤1μg / g; High potassium: refers to a potassium content >1μg / g; High sodium content refers to a sodium content greater than 1 μg / g.

18. An application of the purity prediction method for high-purity quartz of granite type as described in any one of claims 1-17.

Images