Preparation method of high-calcium fluorite and barite paragenic ore combined inhibitor

Abstract

The invention relates to the technical field of mineral processing data processing, in particular to a preparation method of a high-calcium fluorite and barite paragenic ore combined inhibitor, which comprises the following steps: collecting multi-source process mineralogy data output by an X-ray fluorescence spectrometer, an automatic mineral analyzer and the like, converting the multi-source process mineralogy data into a structured format, and constructing a process mineralogy database; based on historical data of a database, a random forest algorithm is adopted to train a correlation model, the correlation rule of mineral characteristics such as calcite content, fluorite particle size distribution and barite monomer dissociation degree and the inhibitor ratio is learned, and parameters of the optimization model are fed back through a flotation test. The combined inhibitor solves the problem of selective separation caused by similar surface properties through a synergistic effect of covering active sites of fluorite calcium by modified water glass, regulating active sites of barite barium by citric acid and assisting in enhancing inhibition by modified starch, and solves the problem of low efficiency of multi-source data integration through structured data processing and model closed-loop optimization; the flotation separation efficiency and the concentrate grade are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of mineral processing data processing, and particularly to a preparation method of a combined inhibitor for a high-calcium fluorite and barite symbiotic ore. Background Art

[0002] The high-calcium fluorite and barite symbiotic ore often faces challenges in flotation separation. Due to the similarity in crystal structure and surface properties, the surface active sites of both are prone to non-selective adsorption with the collector, making it difficult to effectively separate the target minerals. A single inhibitor, due to limited action targets, usually can only exert an inhibitory effect on the surface characteristics of one of the minerals and is difficult to simultaneously regulate the binding ability of the two minerals with the collector. The design of a combined inhibitor needs to be based on the differences in surface ions of the two minerals - the surface of high-calcium fluorite is rich in calcium ions, and the surface of barite is mainly composed of barium ions. Inhibitors with different components can specifically interact with these two types of ions respectively: one component covers the calcium active sites on the surface of fluorite through chemical adsorption, reducing its affinity with the collector; the other component attaches to the barium active sites on the surface of barite through electrostatic or hydrogen bond interactions, changing its surface hydrophobicity. When the two components act synergistically, they respectively weaken the adsorption ability of fluorite and barite to the collector, but there are differences in the degree of inhibition, enabling one of the minerals to preferentially combine with the collector and float, while the other remains hydrophilic due to the inhibitory effect and stays in the pulp, ultimately achieving the selective separation of the two minerals.

[0003] In the process mineralogy research of high-calcium fluorite and barite symbiotic ore, the analysis data (such as the element content data output by an X-ray fluorescence spectrometer and the particle size distribution and monomer dissociation degree data output by a Mineral Liberation Analyzer (MLA)) are inefficient in data integration and correlation analysis due to incompatible formats and inconsistent dimensions, making it difficult to quickly establish the corresponding relationship between mineral characteristics and flotation process parameters. For example, the X-ray fluorescence spectrometer obtains tabular data of the element content of calcite, fluorite, and barite through chemical analysis, while the Mineral Liberation Analyzer (MLA) generates a bar chart of mineral particle size distribution and a matrix of the intergrowth relationship of monomer dissociation degree through image recognition and statistics. These two types of data are stored in the forms of text tables and graphical statistics respectively, and key parameters need to be manually extracted and the format converted before cross-comparison to correlate the relationship between calcite content, mineral particle size, and the dosage of flotation inhibitor. This process relies on manual operation, which is not only time-consuming but also may affect the accuracy of subsequent flotation process optimization due to data extraction errors, prolonging the cycle from data collection to process adjustment. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the present invention provides a preparation method of a combined inhibitor for high-calcium fluorite and barite symbiotic ore, which solves the separation problem of insufficient selectivity of a single inhibitor due to similar surface properties in high-calcium fluorite and barite symbiotic ore, and the problem of low integration and correlation analysis efficiency caused by incompatible data formats and inconsistent dimensions in multi-source analysis in process mineralogy research.

[0005] To solve the above technical problems, the specific technical solutions of the present invention are as follows: The preparation method of the combined inhibitor for high-calcium fluorite and barite symbiotic ore provided by the present invention includes: Step 1, collecting process mineralogy data of high-calcium fluorite and barite symbiotic ore, where the process mineralogy data includes element content data output by an X-ray fluorescence spectrometer, particle size distribution data and monomer dissociation degree associated relationship data output by a mineral automatic analyzer, and corresponding data of inhibitor components, concentrate grade and recovery rate in historical flotation tests; Step 2, converting the process mineralogy data into a structured format. The tabular data in the process mineralogy data is stored as database columns through field mapping, and the graphic data extracts key parameters through image recognition and stores them as numerical fields, adding meta-information of sample numbers and collection timestamps to form a process mineralogy database; Step 3, based on the historical data of mineral characteristics and inhibitor components in the process mineralogy database, training an association model using a machine learning algorithm, where the mineral characteristic data includes calcite content, fluorite particle size distribution, and barite monomer dissociation degree; Step 4, receiving real-time mineral characteristic data of the ore to be processed, inputting the real-time mineral characteristic data of the ore to be processed into the association model, and obtaining the recommended inhibitor component ratio; Step 5, preparing a combined inhibitor by mixing raw materials according to the recommended ratio, recording the results of flotation tests, and feeding back the test results to the association model to update the model parameters. The flotation test results include concentrate grade, recovery rate, and mineral floating sequence data.

[0006] Furthermore, in the preparation method of the combined inhibitor for high-calcium fluorite and barite symbiotic ore of the present invention, Step 1 includes: Calcite, fluorite and barite content distribution data output by an X-ray fluorescence spectrometer; Proportion data of each mineral particle size interval output by a mineral automatic analyzer and associated ratio data of fluorite with barite and calcite; Particle size interval values and dissociation degree percentage parameters extracted from the graphic data output by a mineral automatic analyzer through image recognition.

[0007] Furthermore, in the preparation method of the combined inhibitor for high-calcium fluorite and barite symbiotic ore of the present invention, Step 3 includes: The correlation model is trained using the random forest algorithm, and the input variables are the calcite content, fluorite particle size distribution, and barite monomer dissociation degree data in the process mineralogy database; The historical data in the process mineralogy database is divided into a training set and a validation set in a ratio of 7:3. The training set is used for the correlation model to learn the correlation law between mineral characteristics and inhibitor ratio; The validation set is used to evaluate the prediction error of the correlation model, and the stability of the model is optimized by adjusting hyperparameters such as the depth of the tree and the feature sampling ratio.

[0008] Furthermore, in the method for preparing a combined inhibitor for high-calcium fluorite and barite symbiotic ore according to the present invention, step 4 includes: The real-time mineral characteristic data of the ore to be processed includes calcite content, fluorite particle size distribution range, and barite monomer dissociation degree; The real-time mineral characteristic data is input into the correlation model trained in step 3. If the calcite content is higher than the average threshold of the historical data in the process mineralogy database and the fluorite particle size is smaller than the median particle size of the historical data, the correlation model outputs the ratio of increasing the proportion of modified water glass; If the barite monomer dissociation degree is lower than the dissociation degree reference value of the historical data, the correlation model outputs the ratio of increasing the proportion of citric acid.

[0009] Furthermore, in the method for preparing a combined inhibitor for high-calcium fluorite and barite symbiotic ore according to the present invention, step 5 includes: The raw materials of the combined inhibitor are modified water glass and citric acid, and the raw material ratio is determined based on the ratio of modified water glass and citric acid recommended by the correlation model in step 4; The raw materials are mixed by mechanical stirring, and the stirring speed is controlled at 100 - 150 revolutions per minute to make the modified water glass and citric acid evenly dispersed; The stirring time is 15 - 20 minutes, and the mixing environment is at room temperature to make the raw materials fully mixed to form a homogeneous inhibitor system.

[0010] Furthermore, the method for preparing a combined inhibitor for high-calcium fluorite and barite symbiotic ore according to the present invention further includes: The combined inhibitor prepared in step 5 is applied to the flotation test, and the floating sequence of fluorite and barite, the calcium fluoride content, barium sulfate content, and calcite residue content in the concentrate are recorded during the test; The flotation result data recorded in the test is compared and analyzed with the ratio recommended by the correlation model in step 4 to generate a feedback adjustment signal for adjusting the characteristic weight coefficient of the mineral characteristics on the inhibitor ratio in the correlation model in step 3; After the correlation model is updated, the inhibitor ratio recommendation logic for the subsequent ore to be processed is optimized based on the updated characteristic weight coefficient to improve the matching degree between the recommended ratio and the actual flotation effect.

[0011] Furthermore, the preparation method of the combined inhibitor for high-calcium fluorite and barite symbiotic ore according to the present invention further includes: The modified starch covers the calcium active sites on the surface of fluorite to assist in enhancing the inhibitory effect of the modified water glass on fluorite; The modified water glass covers the calcium active sites on the surface of fluorite through chemical adsorption to reduce the affinity between fluorite and the collector, keeping the surface of fluorite hydrophilic; Citric acid attaches to the barium active sites on the surface of barite through electrostatic or hydrogen bond interactions to change the surface hydrophobicity of barite and regulate its binding ability with the collector; The three act synergistically in the homogeneous inhibitor system formed in step 5. By differentially inhibiting the adsorption ability of fluorite and barite to the collector, one of the minerals floats preferentially and the other remains in the pulp for selective separation.

[0012] Advantages of the present invention; The present invention converts multi-source process mineralogy data into a unified format through structured data processing technology and constructs a database, solving the problem of low integration efficiency caused by incompatible multi-source data formats and inconsistent dimensions in the traditional method, providing a standardized data basis for subsequent correlation analysis; Based on the correlation model trained by the random forest algorithm, combined with real-time mineral property data, the inhibitor ratio is dynamically recommended, realizing the adaptive adjustment of the ratio scheme to different ore characteristics and improving the pertinence of inhibitor selection; Modified water glass, citric acid and modified starch form a surface hydrophobicity difference with strongly hydrophilic fluorite and weakly hydrophobic barite through synergistic adsorption at different active sites (calcium active sites and barium active sites) in the homogeneous system, effectively solving the problem of insufficient selectivity caused by similar surface properties of a single inhibitor, and finally improving the flotation separation efficiency and concentrate grade of high-calcium fluorite and barite. Description of the Drawings

[0013] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for the embodiments will be briefly introduced below. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the drawings.

[0014] Figure 1 It is a flowchart of a preparation method of a combined inhibitor for high-calcium fluorite and barite symbiotic ore provided by an embodiment of the present invention. Detailed Embodiments

[0015] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention. The technical solutions provided by each embodiment of the present invention will be described in detail below with reference to the drawings. To better understand the objectives of the present invention, the present invention will be further described in detail below.

[0016] Please refer to Figure 1 , the method for preparing a combined inhibitor for high-calcium fluorite and barite symbiotic ore provided by the present invention includes: Step 1: Collect the process mineralogical data of high-calcium fluorite and barite symbiotic ore. The process mineralogical data includes the elemental content data output by an X-ray fluorescence spectrometer, the particle size distribution data and the monomer dissociation degree associated relationship data output by a mineral automatic analyzer, and the corresponding data of inhibitor components, concentrate grade, and recovery rate in historical flotation tests. Step 2: Convert the process mineralogical data into a structured format. The tabular data in the process mineralogical data is stored as database columns through field mapping, and the graphic data extracts key parameters through image recognition and stores them as numerical fields. Additional sample numbers and collection timestamp meta-information are added to form a process mineralogical database. Step 3: Based on the historical data of mineral characteristics and inhibitor components in the process mineralogical database, use a machine learning algorithm to train an association model. The mineral characteristic data includes calcite content, fluorite particle size distribution, and barite monomer dissociation degree. Step 4: Receive the real-time mineral characteristic data of the ore to be processed, input the real-time mineral characteristic data of the ore to be processed into the association model, and obtain the recommended inhibitor component ratio. Step 5: Prepare a combined inhibitor by mixing raw materials according to the recommended ratio, record the flotation test results, and feedback the test results to the association model to update the model parameters. The flotation test results include concentrate grade, recovery rate, and mineral floating order data.

[0017] The method for preparing a combined inhibitor for high-calcium fluorite and barite symbiotic ore provided by the present invention realizes the dynamic optimization of inhibitor ratio through a data-driven intelligent process. The core steps and subordinate technical solutions are supplemented as follows: Process mineralogy data collection is the starting point of the method, and multi-source heterogeneous data needs to be systematically obtained to support subsequent analysis. Specifically, the X-ray fluorescence spectrometer (XRF) outputs the content distribution data of calcite, fluorite, and barite through chemical element analysis, usually presented in the form of a text table, recording the mass percentage of each mineral in the ore (such as calcite content 15%, fluorite content 40%); the Mineral Liberation Analyzer (MLA) outputs two types of data through image recognition and statistical analysis: one is the proportion of each mineral particle size in the intervals of 0.074-0.15mm, 0.15-0.3mm, etc. (particle size distribution data), and the other is the intergrowth ratio of fluorite with barite and calcite (monomer dissociation degree intergrowth relationship data), and the latter represents the symbiotic degree between different minerals in the form of a matrix (such as the intergrowth ratio of fluorite and barite is 25%); historical flotation test data records the ratios of inhibitor components such as modified water glass and citric acid in past tests (such as 3:1, 2.5:1.5) and the grades of calcium fluoride and barium sulfate in the corresponding concentrates (such as calcium fluoride grade 92%, barium sulfate grade 88%). All data is labeled with a sample number (such as S20250515-01) and the collection time (such as May 15, 2025, 10:00) during collection to achieve data traceability.

[0018] The multi-source data collected needs to be converted into a unified structured format to construct a process mineralogy database. For the tabular data output by XRF, through field mapping, column names such as "calcite content" and "fluorite content" are mapped to the "calcite_content" and "fluorite_content" fields of the database and stored as numerical data; the graphical data output by MLA (such as particle size distribution histograms, intergrowth relationship heatmaps) is processed through image recognition technology: the OCR (Optical Character Recognition) is used to extract the numerical values in each particle size interval in the histogram (such as the proportion in the 0.1-0.5mm interval is 60%), and the percentage values in the intergrowth relationship matrix are recognized through an image segmentation algorithm (such as the fluorite-barite intergrowth ratio is 25%) and converted into numerical fields for storage; after all data is appended with sample numbers and timestamp meta-information, it is stored in a relational database (such as MySQL) in the structure of "sample number - collection time - data type - data value" to form a process mineralogy database that can be queried and associated.

[0019] Based on the historical data in the process mineralogy database, a random forest algorithm is used to train an association model to learn the association rules between mineral properties and inhibitor ratios. The input variables of the model are the calcite content in the database (such as 18%), the fluorite particle size distribution (such as the proportion in the range of 0.1 - 0.5 mm is 60%), and the barite monomer dissociation degree (such as 92%). The output variable is the ratio of modified sodium silicate to citric acid (such as 3:1). During training, 80 groups of historical data in the database are divided into a training set (56 groups) and a validation set (24 groups) according to 7:3: The training set is used for the model to learn the positive correlation between the increase in calcite content and the increase in the proportion of modified sodium silicate, and the negative correlation between the decrease in barite dissociation degree and the increase in the proportion of citric acid; The validation set is used to evaluate the prediction error of the model (such as the deviation between the predicted ratio of 3:1 and the actual optimal ratio of 2.8:1.2), and optimize the model stability by adjusting the tree depth of the random forest (such as from 5 layers to 7 layers) and the feature sampling ratio (such as from 0.7 to 0.8). Finally, an association model that can accurately map mineral properties to inhibitor ratios is obtained.

[0020] The real-time mineral property data of the ore to be processed needs to be input into the association model according to the structured format of the process mineralogy database to obtain the dynamically recommended inhibitor ratio. The real-time data includes the calcite content of the current batch of ore (such as 20%), the fluorite particle size distribution (such as the proportion in the range of 0.1 - 0.5 mm is 55%), and the barite monomer dissociation degree (such as 88%), and its format is the same as the historical data in the database (such as "calcite_content = 20%", "fluorite_particle_size = 55%", "barite_liberation = 88%"). The model outputs the ratio based on the association rules obtained from training: If the calcite content (20%) is higher than the average threshold (17%) of the historical data in the database and the fluorite particle size (0.1 - 0.5 mm) is smaller than the historical median particle size (0.3 mm), the model outputs a ratio that increases the proportion of modified sodium silicate (such as 3.2:0.8); If the barite dissociation degree (88%) is lower than the historical reference value (90%), the model outputs a ratio that increases the proportion of citric acid (such as 2.5:1.5).

[0021] After the recommended ratio is determined, it is necessary to prepare the combined inhibitor by mechanically stirring and mixing the raw materials, and optimize the model through the feedback of flotation tests. During specific preparation, weigh the modified sodium silicate and citric acid according to the recommended ratio (such as 3.2:0.8), and mechanically stir at a speed of 120 revolutions per minute for 18 minutes in a normal temperature environment to make the two raw materials evenly disperse to form a homogeneous system. Apply the prepared inhibitor to the flotation test, and record the floating order of fluorite and barite (such as fluorite floating within 10 minutes and barite floating after 20 minutes), the calcium fluoride content in the concentrate (such as 93%), the barium sulfate content (such as 89%), and the calcite residue (such as 2%). After comparing the test data with the ratio (3.2:0.8) recommended by the model, a feedback adjustment signal is generated (such as the weight coefficient of the calcite content to the proportion of modified sodium silicate is adjusted from 0.4 to 0.5) to update the feature weights of the associated model. After the model is updated, the recommended ratio of the inhibitor for the subsequent ore to be processed will be based on the adjusted weight coefficient to improve the matching degree between the recommended ratio and the actual flotation effect.

[0022] Specifically, in the method for preparing the combined inhibitor for high-calcium fluorite and barite symbiotic ore according to the present invention, step 1 includes: The calcite, fluorite and barite content distribution data output by the X-ray fluorescence spectrometer; The data of the proportion of each mineral particle size interval and the intergrowth ratio data of fluorite with barite and calcite output by the mineral automatic analyzer; The particle size interval value and dissociation degree percentage parameter extracted by image recognition from the graphic data output by the mineral automatic analyzer.

[0023] In the method for preparing the combined inhibitor for high-calcium fluorite and barite symbiotic ore according to the present invention, the collection of process mineralogy data in step 1 needs to systematically obtain multi-source heterogeneous data to provide a basis for subsequent data integration and model training.

[0024] The calcite, fluorite and barite content distribution data output by the X-ray fluorescence spectrometer (XRF) is obtained by exciting the elements in the ore sample with X-rays to generate characteristic fluorescence, and the mass percentage of each mineral is obtained after spectral analysis. During specific implementation, the ore sample is ground to 200 mesh (particle size about 0.074mm) and then pressed into a tablet, and put into the XRF equipment for element analysis. The equipment converts the element signal into mineral content data through a calibration curve. For example, the test results of a certain sample may show that the calcite content is 18%, the fluorite content is 45%, and the barite content is 22%. Such data is usually output in the form of a two-dimensional table, and the column names include fields such as "mineral type" and "content (mass percentage)", directly reflecting the relative proportion of target minerals and impurity minerals in the ore.

[0025] The proportion data of each mineral particle size interval output by the Mineral Liberation Analyzer (MLA) is obtained by collecting images and identifying the composition of ore particles through a Scanning Electron Microscope (SEM) and an Energy Dispersive Spectrometer (EDS), and the distribution of each mineral in different particle size ranges is analyzed by combining statistical algorithms. For example, the equipment divides ore particles into three intervals of 0.074 - 0.15mm, 0.15 - 0.3mm, and 0.3 - 0.6mm, and statistically calculates the proportion of the number of particles of fluorite, barite, and calcite in each interval (such as fluorite accounting for 60% in the 0.15 - 0.3mm interval). Such data is usually presented in the form of a bar chart or a line chart, intuitively reflecting the mineral particle size distribution characteristics.

[0026] The associated proportion data of fluorite with barite and calcite output by the Mineral Liberation Analyzer is obtained by identifying the symbiotic relationship of different minerals in ore particles. The associated proportion is defined as the percentage of the number of a certain mineral particle symbiotic with other mineral particles in the total number of particles of this mineral (such as 25% of fluorite particles symbiotic with barite and 15% symbiotic with calcite). The MLA uses image segmentation technology to distinguish the boundaries of different minerals, combines composition analysis to determine the associated relationship, and finally outputs it in the form of a matrix or a heat map, such as "fluorite - barite associated proportion 25%" and "fluorite - calcite associated proportion 15%". Such data is used to evaluate the degree of monomer dissociation of minerals and is a key indicator of the difficulty of flotation separation.

[0027] The graphic data (such as particle size distribution bar chart, associated proportion heat map) output by the Mineral Liberation Analyzer needs to extract key parameters through image recognition technology. Specifically, when implementing, the Optical Character Recognition (OCR) technology is used to extract the particle size interval values (such as "0.15 - 0.3mm") and the corresponding proportions (such as "60%") from the coordinate axes and annotations of the bar chart, and the percentage values of the associated proportion (such as "25%") are identified from the color gradient of the heat map through an image segmentation algorithm. The extracted parameters are stored as numerical fields after format conversion (such as "particle_size_range: 0.15 - 0.3mm" and "proportion: 60%"), which together with the tabular data output by XRF constitute the original data layer of the process mineralogy database.

[0028] Specifically, for the preparation method of the combined inhibitor for high - calcium fluorite and barite symbiotic ore described in the present invention, step 3 includes: Training an association model using the random forest algorithm, with the input variables being the calcite content, fluorite particle size distribution, and barite monomer dissociation degree data in the process mineralogy database; Dividing the historical data in the process mineralogy database into a training set and a validation set according to a ratio of 7:3. The training set is used for the association model to learn the association rules between mineral characteristics and inhibitor ratios; The validation set is used to evaluate the prediction error of the correlation model, and the model stability is optimized by adjusting hyperparameters such as the depth of the tree and the feature sampling ratio.

[0029] In the preparation method of the combined inhibitor for high-calcium fluorite and barite symbiotic ore described in the present invention, the correlation model training in step 3 realizes the intelligent correlation learning between mineral properties and inhibitor ratio through the random forest algorithm, and its specific implementation needs to combine links such as data preprocessing, model construction, and parameter optimization.

[0030] The input variables of the correlation model are derived from the structured data in the process mineralogy database, specifically including calcite content (such as mass percentage values of 15%, 18%, 20%, etc.), fluorite particle size distribution (such as statistical values of the proportion in the 0.1 - 0.5 mm interval of 50%, 60%, etc.), and barite monomer dissociation degree (such as reciprocals of the intergrowth ratios of 85%, 90%, 92%, etc.). These variables are stored in the database in the form of numerical fields (such as "calcite_content", "fluorite_size_distribution", "barite_liberation"), and after being read through a programming interface (such as the pandas library in Python), they are converted into a feature matrix recognizable by the model. Each row represents a historical test sample, and each column corresponds to a mineral property variable.

[0031] Before model training, the historical data in the process mineralogy database needs to be divided into a training set and a validation set in a ratio of 7:3. For example, if the database contains 100 groups of historical test data (each group of data includes calcite content, fluorite particle size distribution, barite dissociation degree, and the corresponding inhibitor ratio), then 70 groups are randomly selected as the training set (for model learning), and the remaining 30 groups are used as the validation set (for model evaluation). When dividing the data, the consistency of the sample distribution needs to be maintained (such as the same ratio of high, medium, and low calcite content samples in the training set and the validation set) to avoid overfitting or underfitting of the model due to data deviation.

[0032] The training set is used for the correlation model to learn the correlation law between mineral properties and inhibitor ratio. Specifically in implementation, a random forest algorithm is used to construct the model: the algorithm extracts multiple sub-samples from the training set through bootstrap sampling, and each sub-sample generates a decision tree. Each node of the tree selects the optimal splitting condition (such as "calcite content > 17%") based on the information gain of the mineral property variable (such as calcite content), and finally outputs the inhibitor ratio (such as modified water glass:citric acid = 3:1) through the voting results of multiple trees. For example, samples with a higher calcite content (> 17%) in the training set mostly correspond to a higher proportion of modified water glass in the ratio (such as 3:1), and the model learns such patterns to establish a positive correlation between calcite content and the proportion of modified water glass.

[0033] The validation set is used to evaluate the prediction error of the association model and optimize the model stability. After the model training is completed, the mineral property data in the validation set (such as calcite content 19%, fluorite particle size distribution 55%, barite dissociation degree 88%) are input into the model to obtain the predicted inhibitor ratio (such as predicted ratio 3.2:0.8), and the error is calculated by comparing it with the actual optimal ratio in the validation set (such as actual ratio 3:1) (such as absolute deviation 0.2:0.2). If the error exceeds the preset threshold (such as 0.3:0.3), the hyperparameters of the random forest are adjusted: for example, increasing the depth of the tree (adjusting from the default 5 layers to 7 layers) to enhance the model's ability to capture complex associations, or reducing the feature sampling ratio (adjusting from 0.8 to 0.7) to reduce the influence of noise. Through repeated adjustment and validation, an association model with a prediction error stably within an acceptable range (such as deviation ≤ 0.2:0.2) is finally obtained.

[0034] Specifically, in the method for preparing the combined inhibitor for high-calcium fluorite and barite symbiotic ore described in the present invention, step 4 includes: The real-time mineral property data of the ore to be processed includes calcite content, fluorite particle size distribution range, and barite monomer dissociation degree; The real-time mineral property data is input into the association model trained in step 3. If the calcite content is higher than the average threshold of the historical data in the process mineralogy database and the fluorite particle size is smaller than the median particle size of the historical data, the association model outputs a ratio for increasing the proportion of modified water glass. If the barite monomer dissociation degree is lower than the dissociation degree reference value of the historical data, the association model outputs a ratio for increasing the proportion of citric acid.

[0035] In the method for preparing the combined inhibitor for high-calcium fluorite and barite symbiotic ore described in the present invention, step 4 realizes the intelligent recommendation of the inhibitor ratio of the ore to be processed through the association model, and its core lies in the structured input of the real-time mineral property data and the dynamic decision-making logic of the model.

[0036] The real-time mineral property data of the ore to be processed needs to be in the same format as the historical data in the process mineralogy database, specifically including three dimensions: calcite content (expressed as a mass percentage, e.g., 19%), fluorite particle size distribution range (expressed as the proportion of each particle size range, e.g., the proportion in the 0.1 - 0.5 mm range is 55%), and barite monomer dissociation degree (expressed as the reciprocal of the intergrown proportion, e.g., 88%). These data are collected by the same detection equipment as in Step 1: the calcite content is obtained by detection with an X-ray fluorescence spectrometer (XRF), and the fluorite particle size distribution and barite monomer dissociation degree are analyzed and output by a mineral automatic analyzer (MLA), and are converted into a structured format (such as "calcite_content = 19%", "fluorite_size = 55%", "barite_liberation = 88%") through image recognition and field mapping, which is exactly the same as the field names and data types of the historical data in the database.

[0037] After the real-time mineral property data is input into the correlation model trained in Step 3, the model makes dynamic decisions based on the statistical characteristics of the historical data. The average threshold and median particle size of the historical data are obtained by statistical calculation of the calcite content and fluorite particle size distribution of all historical samples in the process mineralogy database: for example, the average value of the calcite content of 100 groups of historical samples in the database is 17% (average threshold), and the median particle size of the fluorite particle size distribution is 0.3 mm (that is, the fluorite particle size of 50% of the samples is less than 0.3 mm). If the calcite content (19%) of the ore to be processed is higher than the average threshold (17%) and the fluorite particle size (the median particle size corresponding to the proportion of 55% in the 0.1 - 0.5 mm range is 0.25 mm) is less than the historical median particle size (0.3 mm), the model identifies that more calcium active sites on the surface of fluorite in the ore are exposed and stronger calcium site coverage ability is required, so it outputs a ratio for increasing the proportion of modified sodium silicate (such as adjusting from the historical average of 3:1 to 3.2:0.8).

[0038] If the barite monomer dissociation degree (88%) of the ore to be processed is lower than the dissociation degree reference value (90%) of the historical data, the model judges that the barite is more highly intergrown with other minerals and the barium active sites on the surface are wrapped, and it is necessary to enhance the action on the barium sites to change its hydrophobicity, so it outputs a ratio for increasing the proportion of citric acid (such as adjusting from the historical average of 2.8:1.2 to 2.5:1.5). Such ratio adjustment logic is based on the correlation rules learned in the model training stage (such as for every 1% increase in calcite content, the proportion of modified sodium silicate increases by 0.1; for every 1% decrease in barite dissociation degree, the proportion of citric acid increases by 0.05), and the final recommended ratio is calculated by linearly combining the weight coefficients of each mineral property variable (such as the weight of calcite content is 0.4 and the weight of barite dissociation degree is 0.3).

[0039] Specifically, for the preparation method of the combined inhibitor for the high-calcium fluorite and barite symbiotic ore described in the present invention, step 5 includes: The raw materials of the combined inhibitor are modified water glass and citric acid, and the raw material ratio is determined based on the ratio of modified water glass and citric acid recommended by the correlation model in step 4; The raw materials are mixed by mechanical stirring, and the stirring speed is controlled at 100 - 150 revolutions per minute to uniformly disperse the modified water glass and citric acid; The stirring time is 15 - 20 minutes, and the mixing environment is at room temperature to fully mix the raw materials to form a homogeneous inhibitor system.

[0040] In the preparation method of the combined inhibitor for the high-calcium fluorite and barite symbiotic ore described in the present invention, steps 4 and 5 together constitute a closed-loop process of "model recommendation - raw material preparation". Through the ratio recommendation driven by real-time data and the standardized mixing process, the precise preparation of the combined inhibitor is achieved.

[0041] The input of real-time mineral characteristic data and model decision-making in step 4 are based on the historical statistical characteristics of the process mineralogy database. The acquisition and processing of real-time data are the same as those in step 1: the calcite content is obtained by detecting with an X-ray fluorescence spectrometer (XRF) (such as the calcite content of the current batch of ore is 19%), the fluorite particle size distribution is statistically analyzed by a mineral automatic analyzer (MLA) for the proportion in each particle size interval (such as the proportion in the 0.1 - 0.5 mm interval is 55%, and the corresponding median particle size is 0.25 mm), and the barite monomer dissociation degree is calculated by analyzing the intergrowth relationship with MLA (such as the intergrowth ratio is 12%, then the dissociation degree is 88%). These data are converted into a structured format through field mapping (such as "calcite_content = 19%", "fluorite_size = 55%", "barite_liberation = 88%"), which are exactly the same as the field names (such as "calcite_content") and data types (numeric type) of the historical data in the database.

[0042] When the model makes a decision, the historical statistical characteristics of the process mineralogy database are determined through a preprocessing step: the average threshold of calcite content is the average value of all historical samples in the database (e.g., 17%), the median particle size of fluorite is the particle size corresponding to the 50th percentile in historical samples (e.g., 0.3 mm), and the reference value of barite dissociation degree is the lowest dissociation degree when the flotation effect is optimal in historical samples (e.g., 90%). If the real-time calcite content (19%) is higher than the average threshold (17%) and the fluorite particle size (0.25 mm) is smaller than the median particle size (0.3 mm), the model identifies that the calcium active sites on the surface of fluorite in the ore are more fully exposed and the coverage ability of calcium sites needs to be enhanced. Therefore, the ratio of modified water glass is output to be increased (e.g., adjusted from the historical average of 3:1 to 3.2:0.8). If the real-time barite dissociation degree (88%) is lower than the reference value (90%), the model judges that the barium active sites on the surface of barite are wrapped and the role of barium sites needs to be strengthened. Therefore, the ratio of citric acid is output to be increased (e.g., adjusted from the historical average of 2.8:1.2 to 2.5:1.5).

[0043] The preparation of the combined inhibitor in step 5 is carried out based on the recommended ratio in step 4. When weighing the raw materials, use an electronic balance (accuracy 0.1 g) to weigh the modified water glass (320 g) and citric acid (80 g) respectively according to the recommended ratio (e.g., 3.2:0.8) to achieve a mass error of no more than ±1%. The mixing equipment uses a laboratory mechanical stirrer (such as IKA RW 20 type), and the stirring paddle is an anchor-type blade (diameter 5 cm) to improve the mixing effect of high-viscosity raw materials. The stirring speed is set at 120 revolutions per minute (within the range of 100 - 150 revolutions per minute). This speed can not only avoid uneven mixing caused by too low a rotation speed but also prevent the introduction of bubbles caused by too high a rotation speed. The stirring time is controlled at 18 minutes (within the range of 15 - 20 minutes), and is accurately controlled by a timing device (such as a digital display timer) to make the raw materials fully contact. The mixing environment is at room temperature (25 ± 5 °C), without additional heating or cooling, reducing energy consumption while avoiding the influence of temperature changes on the properties of raw materials.

[0044] During the mixing process, the dispersion uniformity is judged by observing the state of the raw materials: in the initial stage (0 - 5 minutes), the modified water glass (viscous liquid) and citric acid (white powder) show a layered state; at 5 - 15 minutes, the paddle drives the liquid to form a vortex, and the citric acid powder gradually dissolves and disperses; after 15 minutes, the color and viscosity of the system tend to be consistent (such as a translucent uniform liquid), indicating that a homogeneous inhibitor system has been formed. The finally prepared inhibitor is transferred to a sealed container through a pipette, and the sample number (such as S20250515 - 02) and the preparation time (such as May 15, 2025, 14:30) are marked for subsequent flotation test verification.

[0045] Specifically, the preparation method of the combined inhibitor for the high-calcium fluorite and barite symbiotic ore described in the present invention further includes: Applying the combined inhibitor prepared in step 5 to the flotation test, and recording the floating sequence of fluorite and barite during the test, the calcium fluoride content, barium sulfate content and calcite residue content in the concentrate; Comparing and analyzing the flotation result data recorded in the test with the ratio recommended by the correlation model in step 4 to generate a feedback adjustment signal for adjusting the characteristic weight coefficient of the influence of mineral characteristics on the inhibitor ratio in the correlation model in step 3; After the correlation model is updated, based on the updated characteristic weight coefficient, optimize the inhibitor ratio recommendation logic for the subsequent ore to be processed, and improve the matching degree between the recommended ratio and the actual flotation effect.

[0046] In the preparation method of the combined inhibitor for the high-calcium fluorite and barite symbiotic ore described in the present invention, the flotation test verification and model feedback optimization link is the key to realizing the dynamic iteration of the inhibitor ratio. Through the comparative analysis of the actual flotation effect and the model recommended result, continuously improve the adaptability of the model to the complex ore characteristics.

[0047] The flotation test application of the combined inhibitor needs to be carried out under standardized conditions. Specifically, when implementing, use the laboratory XFD type single-tank flotation machine, set the pulp volume to 1L, control the pulp concentration at 30% (the ratio of solid mass to the total mass of the pulp), the aeration volume is 0.15 m³ / h, and the stirring speed is 1800 revolutions per minute. Add the combined inhibitor prepared in step 5 (such as the modified water glass and citric acid with a ratio of 3.2:0.8) to the pulp at a dosage of 100 g / t (add 100 g of inhibitor per ton of ore), and stir for 3 minutes to allow the inhibitor to fully act on the minerals. Then start aeration and record the floating sequence of fluorite and barite: Fluorite usually floats within 1-3 minutes due to its strong surface hydrophobicity, and barite has strong surface hydrophilicity due to the inhibitory effect and usually floats after 5-8 minutes; if barite floats too early (such as within 2 minutes), it indicates that the proportion of citric acid is insufficient and the effect on barium sites needs to be enhanced.

[0048] The recorded data of the flotation test include the concentrate grade and impurity residue content. After the concentrate is collected by scraping foam, it is dehydrated and dried, and the calcium fluoride content (fluorite grade) and barium sulfate content (barite grade) are detected by an X-ray fluorescence spectrometer (XRF). For example, the test result of a certain test is 93% calcium fluoride and 89% barium sulfate. The calcite residue content is determined by chemical titration (such as EDTA complexometric titration of calcium content), and converted into the calcite mass percentage (such as 2%). All data are marked with the test number (such as T20250515-01) and time (such as 15:00 on May 15, 2025), and are jointly stored in the "Test Verification" table of the process mineralogy database with the ratio recommended by the correlation model in step 4 (such as 3.2:0.8).

[0049] The comparative analysis of the test data and the model-recommended ratio is achieved through deviation calculation. For example, the model-recommended ratio is modified sodium silicate:citric acid = 3.2:0.8. In the actual flotation effect, the grade of calcium fluoride is 93% (the target grade is 92%), and the residual amount of calcite is 2% (the target residual amount is 3%), indicating that this ratio effectively inhibits calcite. However, the grade of barite is 89% (the target grade is 90%), which is slightly lower, probably due to insufficient citric acid proportion. By calculating the deviation of each index (such as the grade deviation of barite = -1%) and combining with the mineral characteristics (such as the dissociation degree of barite is 88%), a feedback adjustment signal is generated: if the dissociation degree of barite and the grade deviation are negatively correlated (the lower the dissociation degree, the greater the grade deviation), then increase the characteristic weight coefficient of the dissociation degree of barite in the model (such as from 0.3 to 0.4) to strengthen the influence of this variable on the citric acid proportion.

[0050] The adjustment of the characteristic weights of the correlation model is achieved through the evaluation of the feature importance of the random forest algorithm. Each mineral characteristic variable (calcite content, fluorite particle size distribution, barite dissociation degree) in the model corresponds to a weight coefficient, reflecting its influence degree on the inhibitor ratio (such as the weight of calcite content 0.4 indicates that its influence on the proportion of modified sodium silicate accounts for 40%). The feedback adjustment signal updates the model by adjusting these weight coefficients: for example, if the correlation between the barite dissociation degree and the grade deviation is enhanced, its weight coefficient is increased from 0.3 to 0.4, and the model will pay more attention to the adjustment of the citric acid proportion by the barite dissociation degree in subsequent recommendations. The updated model re-evaluates the prediction error through the validation set (such as from 0.2:0.2 to 0.15:0.15), making the optimized recommendation logic more in line with the actual flotation effect.

[0051] Specifically, the preparation method of the combined inhibitor for high-calcium fluorite and barite symbiotic ore described in the present invention further includes: The modified starch assists in enhancing the inhibitory effect of the modified sodium silicate on fluorite by covering the calcium active sites on the surface of fluorite; The modified sodium silicate covers the calcium active sites on the surface of fluorite through chemical adsorption, reducing the affinity between fluorite and the collector, and keeping the surface of fluorite hydrophilic; Citric acid attaches to the barium active sites on the surface of barite through electrostatic or hydrogen bond interactions, changing the surface hydrophobicity of barite and regulating its binding ability with the collector; The three act synergistically in the homogeneous inhibitor system formed in step 5. By differentially inhibiting the adsorption ability of fluorite and barite to the collector, one of the minerals floats preferentially and the other remains in the pulp for selective separation.

[0052] In the combined inhibitor of high-calcium fluorite and barite symbiotic ore described in the present invention, modified starch, modified water glass and citric acid achieve precise regulation of the adsorption capacity of the two minerals through different action sites and a synergistic mechanism. The specific action mode and synergistic process need to be described in combination with the surface characteristics of the minerals and the physical and chemical properties of the inhibitor components.

[0053] As an auxiliary inhibitory component, modified starch enhances the shielding of the calcium active sites on the fluorite surface through physical covering. Due to lattice fracture on the fluorite surface, a large number of calcium ions are exposed, forming positively charged active sites, which are easily combined with collectors (such as oleate ions) through electrostatic interaction. Modified starch (such as carboxymethyl starch) contains multiple hydroxyl groups (-OH) and carboxyl groups (-COOH) in its molecular chain, and can be adsorbed on the Ca²⁺ sites on the fluorite surface through hydrogen bonding to form a polymer film with a thickness of about 5 - 10 nm. Although the inhibitory ability of this film alone is limited (only reducing the collector adsorption amount by about 10%), it can form an overlapping coverage with modified water glass: Modified water glass (such as the modified product of sodium silicate) contains silicate ions, which can form a stable calcium silicate complex with Ca²⁺ through chemical adsorption, with a coverage thickness of about 2 - 5 nm; the polymer film of modified starch covers the surface of the calcium silicate layer, further hindering the contact between the collector and Ca²⁺, reducing the collector adsorption amount on the fluorite surface by about 30%, and significantly enhancing the inhibitory effect.

[0054] The core role of modified water glass is to reduce the affinity between fluorite and the collector through chemical adsorption. The binding energy between its silicate ions and Ca²⁺ on the fluorite surface is about -80 kJ / mol (calculated by density functional theory), which is much higher than the binding energy between the collector oleate ion (-COO⁻) and Ca²⁺ (about -50 kJ / mol), so it preferentially occupies the Ca²⁺ sites. After adsorption, the surface of fluorite changes from its original hydrophobicity (contact angle about 75°) to hydrophilicity (contact angle about 30°), and the surface energy increases by about 20 mN / m, resulting in its difficulty in attaching to bubbles and floating, and remaining in the pulp.

[0055] Citric acid regulates the hydrophobicity of the barite surface through electrostatic or hydrogen bonding. The barite surface is mainly composed of barium ions and is positively charged; the citric acid molecule contains three carboxyl groups (-COOH), which are negatively charged after dissociation (dissociation degree is about 80% at pH = 7), and can be adsorbed on the Ba²⁺ sites through electrostatic attraction. After adsorption, the hydrophobic carbon chain of the citric acid molecule faces outwards, increasing the contact angle of the barite surface from hydrophilicity (about 40°) to weak hydrophobicity (about 60°), and reducing the surface energy by about 15 mN / m. At this time, the binding ability between barite and the collector (such as alkyl sulfonate) is between that of fluorite (strongly hydrophilic) and un-inhibited minerals (strongly hydrophobic). During the flotation process, the floating time is later than that of un-inhibited minerals but earlier than that of strongly inhibited fluorite, achieving selective separation.

[0056] The synergistic effect of the three components depends on the homogeneous inhibitor system formed in Step 5. Modified sodium silicate (liquid) and citric acid (powder) are uniformly dispersed under mechanical stirring. Modified starch (powder) is added after pre-dissolution (such as stirring in water at 40 °C for 30 minutes to form a 1% solution). The three are mixed at the molecular level in the homogeneous system (dispersion particle size < 1 μm). When the inhibitor is added to the pulp, modified sodium silicate is preferentially adsorbed on the surface of fluorite (90% adsorption is completed within 2 minutes), followed by the coverage of modified starch (completed within 5 minutes), and citric acid is simultaneously adsorbed on the surface of barite (completed within 3 minutes). This sequential adsorption and spatial complementary coverage make the hydrophilicity of the fluorite surface significantly stronger than that of barite (the contact angle difference is about 30°). Finally, in the flotation process, barite floats up preferentially due to its weak hydrophobicity (5 - 8 minutes), and fluorite remains in the pulp due to its strong hydrophilicity (no obvious floating after 10 minutes), achieving the efficient separation of the two minerals.

[0057] The present invention solves the problem of selective separation of high-calcium fluorite and barite due to similar surface properties through the synergistic effect of the combined inhibitor components. Modified sodium silicate preferentially covers the calcium active sites on the surface of fluorite through chemical adsorption, reducing its affinity with the collector and keeping the surface of fluorite in a hydrophilic state; citric acid attaches to the barium active sites on the surface of barite through electrostatic or hydrogen bond interactions, regulating its hydrophobicity to a weak hydrophobic state; modified starch, as an auxiliary component, enhances the shielding effect of calcium sites on the surface of fluorite through physical coverage. The three components are sequentially adsorbed in the homogeneous inhibitor system, forming a surface difference with strong hydrophilicity of fluorite and weak hydrophobicity of barite, enabling barite to float up preferentially while fluorite remains in the pulp, achieving selective separation.

[0058] Aiming at the problems of incompatible multi-source data formats and low integration efficiency in process mineralogy, the present invention realizes unified management through structured data processing and database construction. The tabular data output by the X-ray fluorescence spectrometer is converted into numerical database columns through field mapping. The graphic data output by the mineral automatic analyzer extracts key parameters (such as particle size interval values, dissociation percentage) through image recognition and stores them as numerical fields. After all data is appended with sample numbers and timestamp meta-information, a structured process mineralogy database is formed. This database eliminates the format differences between text tables and graphic statistics, providing a unified data input basis for subsequent correlation analysis.

[0059] Data-driven correlation model training and dynamic optimization further enhance the adaptability of the separation effect. Based on the historical data in the process mineralogy database, a random forest algorithm is used to train the correlation model to learn the correlation rules between mineral properties such as calcite content, fluorite particle size distribution, and barite monomer dissociation degree and the inhibitor ratio. After the real-time mineral property data of the ore to be processed is input into the model, a dynamically recommended ratio scheme is output. The flotation test results are fed back to the model, and the recommendation logic is optimized by adjusting the feature weight coefficients. This enables the inhibitor ratio to dynamically adapt to the characteristic differences of different batches of ore, ultimately improving the separation efficiency of high-calcium fluorite and barite.

[0060] Example 1: For a sample of a high-calcium fluorite and barite symbiotic ore, tabular data with a calcite content of 18%, a fluorite content of 45%, and a barite content of 22% is obtained by using an X-ray fluorescence spectrometer (model: PANalytical Axios). Through scanning and analysis with an automatic mineral analyzer (model: FEI MLA650), particle size distribution data for each mineral (the proportion in the 0.074 - 0.15 mm interval is 30%, the proportion in the 0.15 - 0.3 mm interval is 50%, and the proportion in the 0.3 - 0.6 mm interval is 20%) and intergrowth relationship data (the fluorite-barite intergrowth ratio is 25%, and the fluorite-calcite intergrowth ratio is 15%) are obtained. For the particle size distribution histogram and intergrowth relationship heat map output by the automatic mineral analyzer, OCR technology is used to extract the numerical values and corresponding proportions of each particle size interval in the histogram, and the intergrowth ratio percentage values (25%, 15%) in the heat map are identified through an image segmentation algorithm. All data is appended with the sample number S20250515-01 and the collection time of 10:00 on May 15, 2025, and stored in a MySQL database through field mapping to form a structured process mineralogy database containing fields such as "calcite content", "fluorite particle size distribution", and "barite monomer dissociation degree".

[0061] Example 2: Based on 100 groups of historical data in the process mineralogy database, a random forest algorithm is used to train the correlation model. The data is divided into a training set (70 groups) and a validation set (30 groups) in a 7:3 ratio. The training set includes calcite content (15% - 22%), fluorite particle size distribution (proportion in the 0.074 - 0.6 mm interval), barite monomer dissociation degree (85% - 95%), and the corresponding inhibitor ratio (modified water glass:citric acid = 2.5:1.5 to 3.5:0.5). The validation set is used to evaluate the model prediction error. During the training stage, the model learns the correlation rule that for every 1% increase in calcite content, the proportion of modified water glass increases by 0.1. During the validation stage, by adjusting the random forest tree depth (from 5 layers to 7 layers) and the feature sampling ratio (from 0.7 to 0.8), the model prediction error decreases from 0.3:0.3 to 0.2:0.2, and the stability is improved.

[0062] Example 3: Collect real-time mineral property data of the ore to be processed: The X-ray fluorescence spectrometer detects that the calcite content is 20% (higher than the historical average threshold of 17% in the database). The mineral automatic analyzer counts that the proportion of fluorite particle size distribution in the range of 0.1 - 0.5 mm is 55% (the median particle size is 0.25 mm, smaller than the historical median particle size of 0.3 mm), and the monomer dissociation degree of barite is 88% (lower than the historical reference value of 90%). Input this data into the trained correlation model. Based on the correlation law between calcite content and fluorite particle size, the model outputs the ratio of increasing the proportion of modified sodium silicate (3.2:0.8); based on the correlation law between barite dissociation degree and citric acid ratio, the model outputs the ratio of increasing the proportion of citric acid (2.5:1.5).

[0063] Example 4: Weigh 320 g of modified sodium silicate and 80 g of citric acid according to the recommended ratio of 3.2:0.8. Use an IKA RW 20 mechanical stirrer (anchor impeller, diameter 5 cm) to stir at 120 revolutions per minute at room temperature (25°C) for 18 minutes. In the first 5 minutes of stirring, the viscous modified sodium silicate and citric acid powder are in a layered state; from 5 to 15 minutes, the impeller drives the liquid to form a vortex, and the citric acid powder gradually dissolves and disperses; after 15 minutes, the color and viscosity of the system tend to be consistent, forming a homogeneous translucent liquid. After detection, the dispersed particle size is less than 1 μm, and the preparation of the combined inhibitor is completed.

[0064] Example 5: Add the prepared combined inhibitor (3.2:0.8) at a dosage of 100 g / t to an XFD single-cell flotation machine (pulp volume 1 L, concentration 30%, aeration rate 0.15 m³ / h, stirring speed 1800 revolutions per minute). Test process record: Fluorite does not show obvious floating after 10 minutes, and barite floats up between 5 - 8 minutes; the calcium fluoride content in the concentrate is 93% (target 92%), the barium sulfate content is 89% (target 90%), and the calcite residue is 2% (target 3%). Comparing the recommended ratio with the test results, the grade deviation of barite is -1%. Combining with the barite dissociation degree of 88%, a feedback adjustment signal is generated, and the characteristic weight of barite dissociation degree in the model is adjusted from 0.3 to 0.4. After the update, the prediction error of the model for subsequent samples is reduced to 0.15:0.15, and the matching degree between the recommended ratio and the actual flotation effect is improved.

[0065] Example 6: Analysis of the action mechanism of the combined inhibitor: Modified sodium silicate (silicate ions) covers the Ca²⁺ sites on the surface of fluorite through chemical adsorption (binding energy -80 kJ / mol), reducing the contact angle of fluorite from 75° to 30° and enhancing the surface hydrophilicity; citric acid (negatively charged after carboxyl dissociation) adsorbs on the Ba²⁺ sites on the surface of barite through electrostatic interaction, increasing the contact angle of barite from 40° to 60° and making the surface weakly hydrophobic; modified starch (carboxymethyl starch) covers the surface of fluorite through hydrogen bonding, forming a polymer film above the calcium silicate layer and further reducing the collector adsorption amount by about 30%. The three are adsorbed sequentially in a homogeneous system (modified sodium silicate for 2 minutes, modified starch for 5 minutes, and citric acid for 3 minutes), forming a surface difference between strongly hydrophilic fluorite (contact angle 30°) and weakly hydrophobic barite (contact angle 60°), realizing the selective separation of preferential flotation of barite and retention of fluorite in the pulp.

Claims

1. A preparation method of a combined inhibitor for a high-calcium fluorite and barite symbiotic ore, characterized in that, Including: Step 1: Collect process mineralogy data of high-calcium fluorite and barite symbiotic ore. The process mineralogy data includes element content data output by an X-ray fluorescence spectrometer, particle size distribution data and monomer dissociation degree associated relationship data output by a mineral automatic analyzer, and corresponding data of inhibitor components, concentrate grade and recovery rate in historical flotation tests; Step 2: Convert the process mineralogy data into a structured format. The tabular data in the process mineralogy data is stored as database columns through field mapping, and the graphic data extracts key parameters through image recognition and stores them as numerical fields, with additional sample numbers and collection timestamp meta-information added to form a process mineralogy database; Step 3: Based on the historical data of mineral characteristics and inhibitor components in the process mineralogy database, use a machine learning algorithm to train an association model. The mineral characteristic data includes calcite content, fluorite particle size distribution, and barite monomer dissociation degree; Step 4: Receive real-time mineral characteristic data of the ore to be processed, input the real-time mineral characteristic data of the ore to be processed into the association model, and obtain the recommended inhibitor component ratio; Step 5: Prepare a combined inhibitor by mixing raw materials according to the recommended ratio, record the flotation test results, and feedback the test results to the association model to update the model parameters. The flotation test results include concentrate grade, recovery rate, and mineral floating order data.

2. The preparation method of the combined inhibitor for the coexisting ore of high-calcium fluorite and barite according to claim 1, characterized in that, The said Step 1 includes: Calcite, fluorite and barite content distribution data output by an X-ray fluorescence spectrometer; Proportion data of each mineral particle size interval output by a mineral automatic analyzer and associated ratio data of fluorite with barite and calcite; Particle size interval numerical values and dissociation degree percentage parameters extracted from the graphic data output by a mineral automatic analyzer through image recognition.

3. The preparation method of the combined inhibitor for the high-calcium fluorite and barite symbiotic ore according to claim 2, wherein, The said Step 3 includes: Use the random forest algorithm to train the association model, and the input variables are calcite content, fluorite particle size distribution, and barite monomer dissociation degree data in the process mineralogy database; Divide the historical data in the process mineralogy database into a training set and a validation set according to a ratio of 7:

3. The training set is used for the association model to learn the association law between mineral characteristics and inhibitor ratio; The validation set is used to evaluate the prediction error of the association model, and optimize the model stability by adjusting hyperparameters such as the depth of the tree and the feature sampling ratio.

4. The preparation method of the combined inhibitor for high-calcium fluorite and barite symbiotic ore according to claim 3, wherein, The said Step 4 includes: The real-time mineral characteristic data of the ore to be processed includes calcite content, fluorite particle size distribution interval, and barite monomer dissociation degree; Input the real-time mineral characteristic data into the association model trained in Step 3. If the calcite content is higher than the average threshold of the historical data in the process mineralogy database and the fluorite particle size is smaller than the median particle size of the historical data, the association model outputs a ratio to increase the proportion of modified sodium silicate; If the barite monomer dissociation degree is lower than the dissociation degree reference value of the historical data, the association model outputs a ratio to increase the proportion of citric acid.

5. The preparation method of the combined inhibitor for the high-calcium fluorite and barite symbiotic ore according to claim 4, wherein, The said Step 5 includes: The raw materials of the combined inhibitor are modified sodium silicate and citric acid, and the raw material ratio is determined based on the ratio of modified sodium silicate and citric acid recommended by the association model in Step 4; Mix the raw materials by mechanical stirring, and control the stirring speed at 100 - 150 revolutions per minute to make the modified sodium silicate and citric acid evenly dispersed; The stirring time is 15 - 20 minutes, and the mixing environment is at room temperature to fully mix the raw materials to form a homogeneous inhibitor system.

6. The preparation method of the combined inhibitor for high-calcium fluorite and barite symbiotic ore according to claim 5, characterized in that, It also includes: Applying the combined inhibitor prepared in step 5 to the flotation test, and recording the floating sequence of fluorite and barite, the calcium fluoride content, barium sulfate content and calcite residue in the concentrate during the test; Comparing and analyzing the flotation result data recorded in the test with the ratio recommended by the correlation model in step 4 to generate a feedback adjustment signal for adjusting the characteristic weight coefficient of the influence of mineral characteristics on the inhibitor ratio in the correlation model in step 3; After the correlation model is updated, optimize the inhibitor ratio recommendation logic for the subsequent ore to be processed based on the updated characteristic weight coefficient to improve the matching degree between the recommended ratio and the actual flotation effect.

7. The preparation method of the combined inhibitor for the coexisting ore of high-calcium fluorite and barite according to claim 6, characterized in that, It also includes: The modified starch covers the calcium active sites on the surface of fluorite to assist in enhancing the inhibitory effect of the modified water glass on fluorite; The modified water glass covers the calcium active sites on the surface of fluorite through chemical adsorption to reduce the affinity between fluorite and the collector and keep the surface of fluorite hydrophilic; Citric acid attaches to the barium active sites on the surface of barite through electrostatic or hydrogen bond interactions to change the hydrophobicity of the barite surface.

Images