A soft-sensing method for indicators based on flotation mechanism and data hybrid drive

Abstract

This invention relates to the field of soft measurement of product quality, and more particularly to a soft measurement method for indicators based on flotation mechanism and data hybrid driving. This invention preprocesses data to construct a flotation production dataset, defines concentrate grade and tailings grade, and constructs a flotation grade mechanism model as the basic model for the flotation process, combining the mineral particle exchange relationship and mass balance relationship in the flotation process. The flotation system parameters are identified using this model, continuously acquiring the system identification parameters most suitable for the current flotation conditions. The flotation grade is measured under the new flotation conditions, and concentrate and tailings grade indicators are obtained. The recovery rate indicator is obtained based on an empirical function of the flotation process recovery rate. Therefore, this invention can provide guidance on key indicators for optimizing mineral flotation production, which is of great significance for ensuring flotation production indicators, improving flotation capacity and recovery rate, and has practical application value.

Description

Technical Field

[0001] The invention relates to the field of product quality soft measurement, and in particular to an indicator soft measurement method based on flotation mechanism and data hybrid drive. Background Art

[0002] The online soft measurement technology of key flotation indicators mainly relies on data-driven methods, combined with machine learning, deep learning and image processing technology to achieve real-time monitoring and optimization of the flotation process. The development and application of online soft measurement technology of key flotation indicators provides an effective means to improve the automation level of the flotation process, optimize operating conditions, reduce resource consumption and improve production efficiency.

[0003] For example, Chinese patent publication number: CN111222798A discloses a soft measurement method for key indicators of complex industrial processes, including the following steps: 1. Collect visual data and process parameter data, and obtain the values ​​of corresponding key indicators through subsequent laboratory tests; 2. Use the dilated convolution fusion neural network (DCFNN) proposed in the present invention to construct a video data spatiotemporal sequence feature extraction network model on video data; 3. Use the GRU network to construct a temporal feature extraction model on industrial parameter data; 4. Input the collected data and process parameters, perform dual-channel concurrent extraction of feature data with time series, and use a fully connected network with an attention mechanism to perform soft measurement of key indicators, and perform reverse model training based on the actual key indicator values ​​obtained from the test; 5. Use the trained soft measurement model to perform real-time estimation of key indicators that are difficult to monitor online.

[0004] However, there are still the following problems in the prior art:

[0005] The flotation process involves complex physical and chemical reactions and has unknown complex nonlinear dynamic characteristics. The internal mechanism of the flotation pool is complex, making it difficult to establish an accurate flotation index mechanism model. In addition, the historical flotation operating condition data set is limited, the data quality is difficult to guarantee, and it is difficult to establish a stable flotation data-driven model.

[0006] Concentrate grade and metal recovery rate are the two most important production evaluation indicators. The two key indicators of ore dressing grade and recovery rate are generally obtained through offline testing at industrial sites. They have the characteristics of long cycle, large lag and high labor intensity. They are difficult to synchronize with the real-time conditions of flotation and have multiple outlier characteristics. It is difficult to effectively guide the online optimization production of the flotation process and improve the flotation production indicators. Summary of the invention

[0007] To this end, the present invention provides an indicator soft measurement method based on flotation mechanism and data hybrid drive to overcome the problem in the prior art that it is difficult to establish an accurate flotation indicator mechanism model and difficult to establish a stable flotation data drive model.

[0008] To achieve the above object, the present invention provides an indicator soft sensing method based on flotation mechanism and data hybrid drive, which includes:

[0009] Step S1, preprocessing the initial sample set of flotation production conditions to construct a flotation production data set, wherein the preprocessing includes cleaning, interpolation and integration of the data in the initial sample set of flotation production conditions.

[0010] Step S2, defining the concentrate grade and tailings grade, and building a flotation grade mechanism model by combining the mineral particle exchange relationship and mass balance relationship in the flotation process;

[0011] Step S3, performing parameter identification on the flotation grade mechanism model based on a particle swarm optimization algorithm to continuously obtain system identification parameters that are most suitable for the current flotation working conditions;

[0012] Step S4, combining the optimal system identification parameters and the flotation grade mechanism model to perform soft measurement on the flotation grade under the current flotation working condition, and determining the soft measurement value of the concentrate grade and the soft measurement value of the tailings grade;

[0013] Step S5, obtaining a recovery rate index based on the flotation process recovery rate empirical function.

[0014] Furthermore, in step S2, the defined concentrate grade is expressed by formula (1), and the defined tailings grade is expressed by formula (2),

[0015]

[0016]

[0017] In formula (1), η C Indicates the concentrate grade, Q C Indicates the mineral particle flow rate to the concentrate pool, C MF It represents the mass of target mineral particles per unit volume of the foam layer, M C is the concentrate quality;

[0018] In formula (2), η T Indicates the tailings grade, Q T is the velocity of mineral particles flowing to the tailings pond, C MP Indicates the mass of target mineral particles per unit volume of the slurry layer, M T The mass of tailings.

[0019] Furthermore, in step S2, the flotation grade mechanism model is expressed by formula (3):

[0020]

[0021] In formula (3), ω1 represents the parameters to be identified in the first system, ω2 represents the parameters to be identified in the second system, α P represents the attachment and separation rate constant of mineral particles and bubbles in the slurry layer, α F represents the attachment and separation rate constant of mineral particles and bubbles in the foam layer, k R represents the rate constant of mineral particles falling from the foam layer to the slurry layer, r1 represents the first hyperparameter, r2 represents the second hyperparameter, V LP Indicates the volume of slurry in the slurry layer, V BP Indicates the volume of gas in the slurry layer, V LF Indicates the volume of slurry in the foam layer, Q A Indicates air velocity, Q C represents the mineral particle flow rate to the concentrate pool, Q T is the mineral particle flow rate to the tailings pond, M P is the ore quality, M C is the concentrate mass, M T is the tailings mass, η C (t) represents the concentrate grade of the flotation cell at the current moment, η T (t) represents the tailings grade of the flotation tank at the current moment, η C (t-1) represents the concentrate grade in the flotation cell at the previous moment, η T (t-1) represents the tailings grade in the flotation tank at the previous moment.

[0022] Furthermore, in step S3, the flotation system parameter identification of the flotation grade mechanism model includes:

[0023] Step S31, extracting sliding window data from the flotation production data set by a sliding window method, each sliding window data contains a set of input feature time series data;

[0024] Step S32, taking the sliding window data as a flotation grade mechanism model parameter identification data set;

[0025] Step S33, initializing the basic parameters of the particle swarm algorithm, including determining the number of particle swarms, the number of iterations, the particle dimension, the learning factor value, the inertia weight value range, the particle swarm position value range, the particle swarm velocity value range, and the initial position and initial velocity of the particle swarm, wherein each particle represents a set of identification parameters;

[0026] Step S34, the flotation grade mechanism model parameter identification data set is used as the input data of the flotation grade mechanism model, and the concentrate grade and tailings grade at the current particle position are solved according to the flotation grade mechanism model, and the fitness evaluation function F is used to evaluate the flotation grade mechanism model. fitness Calculating evaluation parameters for the solution result;

[0027] Step S35, repeating step S34, and determining the optimal system identification parameters under the minimum evaluation parameters through continuous iteration of the particle swarm algorithm;

[0028] Step S36, repeating steps S31-S35, and continuously acquiring the system identification parameters that are most suitable for the current flotation working conditions.

[0029] Furthermore, in step S3, the fitness evaluation function is calculated according to formula (4):

[0030]

[0031] In formula (4), Ω represents the system matrix parameter to be identified Ω = [ω1, ω2], represents the true value of the concentrate grade of the i-th group in the data set, represents the true value of tailings grade in the i-th group of the data set, It represents the predicted value of concentrate grade obtained by the flotation grade mechanism model after predicting the flotation conditions of the i-th group. It represents the predicted value of tailings grade after the flotation grade mechanism model predicts the i-th group of flotation conditions, N represents the number of sampling data, and λ represents the weight factor.

[0032] Furthermore, in step S4, before the next sliding window data is updated, the equivalent working condition identification parameter [ω 1j ,ω 2j ] and the flotation grade mechanism model, determine the concentrate grade and tailings grade, and obtain the concentrate grade soft measurement value η Ce and tailings grade soft measurement value η Te .

[0033] Furthermore, in step S5, the flotation process recovery rate empirical function is used to calculate the recovery rate according to formula (5):

[0034]

[0035] In formula (5), γ represents the recovery rate, η P Indicates the ore grade, η Te represents the soft measurement value of tailings grade, η Ce Represents the soft measurement value of concentrate grade.

[0036] Compared with the existing technology, the method preprocesses the data, constructs the flotation production data set, defines the concentrate grade and tailings grade, and builds a flotation grade mechanism model based on the mineral particle exchange relationship and mass balance relationship in the flotation process. The model is used as the basic model of the flotation process, and the flotation system parameters are identified for the flotation grade mechanism model. The system identification parameters that are most suitable for the current flotation conditions are continuously obtained, the flotation grade under the current new flotation conditions is measured, the concentrate grade and tailings grade indicators are obtained, and the recovery rate indicator is obtained based on the empirical function of the flotation process recovery rate. Furthermore, the method can provide guidance on key indicators for the optimized production of mineral flotation, which is of great significance for ensuring flotation production indicators, improving flotation capacity and recovery rate, and has practical application value.

[0037] In particular, the present invention redefines the concentrate grade and tailings grade, calculates the exchange relationship between mineral particles in the foam layer and the pulp layer in the flotation cell based on the exchange relationship of mineral particles and the mass balance relationship in the flotation process, so as to obtain a simplified flotation grade mechanism model, and identifies the flotation system parameters of the flotation grade mechanism model based on the flotation production data set and the sliding window method, so as to continuously obtain the system identification parameters that are most suitable for the current flotation conditions. In actual situations, the flotation process includes complex physical and chemical reactions, has unknown complex nonlinear dynamic characteristics, and the internal mechanism of the flotation cell is complex, and it is difficult to establish an accurate flotation index mechanism model. Therefore, the present invention provides guidance on key indicators for the optimized production of mineral flotation, is of great significance for ensuring flotation production indicators, improving flotation capacity and recovery rate, and has practical application value.

[0038] In particular, the present invention measures the flotation grade under the current new flotation conditions based on the system identification parameters that are most suitable for the current flotation conditions and the flotation grade mechanism model to obtain concentrate grade and tailings grade indicators. In actual situations, the flotation historical condition data set is limited and has outlier characteristics, and the data quality is difficult to guarantee, and it is difficult to establish a stable flotation data-driven model. Therefore, the present invention provides guidance on key indicators for the optimized production of mineral flotation, which is of great significance for ensuring flotation production indicators and improving flotation capacity and recovery rate, and has practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 The flowchart of the indicator soft sensing method based on flotation mechanism and data mixing drive;

[0040] Figure 2 It is a diagram showing the fitness convergence result of the model identification parameters and the optimization result of the identification parameters of the embodiment of the invention;

[0041] Figure 3 A comparison diagram of soft-measurement waveforms of concentrate grade in an embodiment of the invention;

[0042] Figure 4This is a comparison diagram of soft-sensing waveforms of tailings grade according to an embodiment of the invention;

[0043] Figure 5 It is a comparison diagram of the recovery rate soft measurement waveform of the embodiment of the invention. DETAILED DESCRIPTION

[0044] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0045] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0046] See also Figures 1 to 5 As shown, Figure 1 This is a flow chart of the indicator soft sensing method based on flotation mechanism and data mixing drive. Figure 2 The figure is a diagram of the fitness convergence result of the model identification parameters and the optimization result of the identification parameters of the embodiment of the invention. Figure 3 This is a comparison diagram of the soft measurement waveform of the concentrate grade of the embodiment of the invention. Figure 4 This is a comparison diagram of the tailings grade soft measurement waveform of the embodiment of the invention. Figure 5 The recovery rate soft measurement waveform comparison diagram of the embodiment of the invention, the indicator soft measurement method based on flotation mechanism and data hybrid drive of the embodiment of the invention includes:

[0047] Step S1, preprocessing the initial sample set of flotation production conditions to construct a flotation production data set, wherein the preprocessing includes cleaning, interpolation and integration of the data in the initial sample set of flotation production conditions.

[0048] Step S2, defining the concentrate grade and tailings grade, and building a flotation grade mechanism model by combining the mineral particle exchange relationship and mass balance relationship in the flotation process;

[0049] Step S3, performing parameter identification on the flotation grade mechanism model based on a particle swarm optimization algorithm to continuously obtain system identification parameters that are most suitable for the current flotation working conditions;

[0050] Step S4, combining the optimal system identification parameters and the flotation grade mechanism model to perform soft measurement on the flotation grade under the current flotation working condition, and determining the soft measurement value of the concentrate grade and the soft measurement value of the tailings grade;

[0051] Step S5, obtaining a recovery rate index based on the flotation process recovery rate empirical function.

[0052] Specifically, in step S1, the process of preprocessing the data includes:

[0053] The initial sample set of flotation production conditions is denoted as data_orgin, the sample set of flotation production conditions containing outliers is denoted as data_outlier, the initial sample set of flotation production conditions after outlier processing is denoted as data_non_outlier, the initial sample set of flotation production conditions containing missing values ​​is denoted as data_missing, the initial sample set of flotation production conditions after missing value interpolation is denoted as data_non_missing, the initial sample set of flotation production conditions without data integration is denoted as data_non_integrated, the initial sample set of flotation production conditions after data integration is denoted as data_integrated, and the flotation condition data set constructed after data preprocessing is denoted as datasets;

[0054] The flotation production measurement and control platform is used to obtain multi-source characteristic data of flotation conditions, including the input ore volume M obtained from sensors and weighing systems. P , inflation volume Q A , Slurry layer level height h P 、Concentrate amount M C and tailings volume M T , obtain the ore grade η through offline testing P 、Concentrate grade η C and tailings grade η T Therefore, the characteristic of the jth flotation condition of the i-th category in the initial sample set of flotation production conditions is x ij (i=1,2,…,8;j=1,2,…,122), in order to provide prior information of data distribution, the flotation condition characteristics are divided into Gaussian variables and non-Gaussian variables through distribution analysis;

[0055] Different outlier judgment methods are used for outliers of Gaussian variables and non-Gaussian variables:

[0056] If the i-th flotation condition characteristic is a Gaussian variable type, the 3σ principle is used to judge the abnormal value of the i-th flotation condition characteristic: including, obtaining K i The index value of the outlier I k (k=1,2,…,K i ), K i is the total number of outliers, and the mean of the flotation condition characteristics of the i-th type is calculated. and standard deviation s i , the lower boundary Upper Boundary So the upper and lower boundary conditions of the outlier are x ij (j=1,2,…,122)∈[b i min ,b imax ];

[0057] If the flotation condition characteristic of the i-th type is a non-Gaussian variable, the upper and lower quartiles of the box plot are used to determine the outliers: Get K i The index value of the outlier I k (k=1,2,…,K i ), sort the i-th category features in ascending order, record the first quartile as Q1, the third quartile as Q3, the interquartile range IQR = Q3-Q1, the lower boundary Upper Boundary The upper and lower boundary conditions of the outlier are x ij (j=1,2,…,122)∈[b i min ,b i max ],

[0058] Since outliers may contain abnormal information of flotation condition fluctuations, in order to retain the amount of information in the data preprocessing process, on the one hand, and to prevent the data from being too large or too small to conform to the flotation condition, the outlier boundary condition is used. The abnormal values ​​of the i-th type of flotation working condition characteristics are limited.

[0059] Until all the outliers of flotation operating condition characteristics of all categories have been processed, the initial sample set of flotation production conditions after outlier processing is data_non_outlier;

[0060] The jth missing value of the flotation condition feature of type a in the initial sample set data_missing of flotation production conditions containing missing values ​​in data_orgin is The corresponding index value is I l (k=1,2,…,L a ), L a Represents the total number of missing values.

[0061] Calculate the Spearman correlation coefficient between the flotation condition characteristics of the ath category and the flotation condition characteristics of the remaining categories. The Spearman correlation coefficient calculation formula is calculated according to the following formula:

[0062]

[0063] In the formula, R i represent The rank, Q i Represents x ij The rank of (i≠a), R i -Q i For x aj 、x ijThe rank difference of (i≠a), the value range of the correlation coefficient r is -1≤r≤1,

[0064] In order to select features and avoid multicollinearity, decorrelation processing is performed to reduce the computational cost, so the working condition feature category with correlation coefficient |r|≤0.5 is selected as the input feature of BPNN, and the data without missing values ​​in the sample set data_non_outlier is used for training to obtain the BPNN basic interpolation model Ma of class a feature, and the model is saved as a backup for subsequent interpolation.

[0065] Use the BPNN basic interpolation model to interpolate the jth missing value of the a-th type of flotation condition characteristics Interpolation is performed until the flotation condition characteristics with missing values ​​are processed, and the sample set data_non_missing after missing value processing is obtained. Steps S2-S5 are repeated to perform outlier judgment processing on the interpolated values, and the data set data_non_outlier is corrected.

[0066] In order to remove redundant or repeated flotation condition cases, the multi-source data are converted into flotation condition characteristics with unified standards. Therefore, data integration processing is performed on the sample set data_non_integrated containing non-unified standards in the acquired data_orgin to obtain the sample set data_integrated.

[0067] Combine all preprocessed sample sets and the initial sample set data_orgin to construct the flotation production dataset datasets, and the data preprocessing is completed.

[0068] In step S2, the defined concentrate grade is expressed by formula (1), and the defined tailings grade is expressed by formula (2),

[0069]

[0070]

[0071] In formula (1), η C Indicates the concentrate grade, Q C Indicates the mineral particle flow rate to the concentrate pool, C MF It represents the mass of target mineral particles per unit volume of the foam layer, M C is the concentrate quality;

[0072] In formula (2), η T Indicates the tailings grade, Q T is the velocity of mineral particles flowing to the tailings pond, C MP Indicates the mass of target mineral particles per unit volume of the slurry layer, M T The mass of tailings.

[0073] Specifically, in step S2, the flotation grade mechanism model is expressed by formula (3):

[0074]

[0075] In formula (3), ω1 represents the parameters to be identified in the first system, ω2 represents the parameters to be identified in the second system, α P represents the attachment and separation rate constant of mineral particles and bubbles in the slurry layer, α F represents the attachment and separation rate constant of mineral particles and bubbles in the foam layer, k R represents the rate constant of mineral particles falling from the foam layer to the pulp layer, r1 represents the film thickness around the bubble and is related to the average bubble diameter of the pulp layer, r2 represents the average gas content of the foam layer, r1 represents the first hyperparameter, r2 represents the second hyperparameter, V LP Indicates the volume of slurry in the slurry layer, V BP Indicates the volume of gas in the slurry layer, V LF Indicates the volume of slurry in the foam layer, Q A Indicates air velocity, Q C represents the mineral particle flow rate to the concentrate pool, Q T is the mineral particle flow rate to the tailings pond; η C is the concentrate grade of the flotation cell at the current moment, η T M is the tailings grade of the flotation tank at the current moment; P is the ore quality, M C is the concentrate mass, M T is the tailings mass, η C (t) represents the concentrate grade of the flotation cell at the current moment, η T (t) represents the tailings grade of the flotation tank at the current moment, η C (t-1) represents the concentrate grade in the flotation cell at the previous moment, η T (t-1) represents the tailings grade in the flotation tank at the previous moment.

[0076] Specifically, in step S3, the flotation system parameter identification of the flotation grade mechanism model includes:

[0077] Step S31, extracting sliding window data from the flotation production data set by a sliding window method, each sliding window data contains a set of input feature time series data;

[0078] Step S32, taking the sliding window data as a flotation grade mechanism model parameter identification data set;

[0079] Step S33, initializing the basic parameters of the particle swarm algorithm, including determining the number of particle swarms, the number of iterations, the particle dimension, the learning factor value, the inertia weight value range, the particle swarm position value range, the particle swarm velocity value range, and the initial position and initial velocity of the particle swarm, wherein each particle represents a set of identification parameters;

[0080] Step S34, the flotation grade mechanism model parameter identification data set is used as the input data of the flotation grade mechanism model, and the concentrate grade and tailings grade at the current particle position are solved according to the flotation grade mechanism model, and the fitness evaluation function F is used to evaluate the flotation grade mechanism model. fitness Calculating evaluation parameters for the solution result;

[0081] Step S35, repeating step S34, and determining the optimal system identification parameters under the minimum evaluation parameters through continuous iteration of the particle swarm algorithm;

[0082] Step S36, repeating steps S31-S35, and continuously acquiring the system identification parameters that are most suitable for the current flotation working conditions.

[0083] Specifically, in step S3, the fitness evaluation function is calculated according to formula (4):

[0084]

[0085] In formula (4), Ω represents the system matrix parameter to be identified Ω = [ω1, ω2], represents the true value of the concentrate grade of the i-th group in the data set, represents the true value of tailings grade in the i-th group of the data set, It represents the predicted value of concentrate grade obtained by the flotation grade mechanism model after predicting the flotation conditions of the i-th group. It represents the predicted value of tailings grade after the flotation grade mechanism model predicts the i-th group of flotation conditions, N represents the number of sampling data, and λ represents the weight factor.

[0086] Specifically, in step S4, before the next sliding window data is updated, the equivalent working condition identification parameter [ω 1j ,ω 2j ] and the flotation grade mechanism model, determine the concentrate grade and tailings grade, and obtain the concentrate grade soft measurement value η Ce and tailings grade soft measurement value η Te .

[0087] Specifically, in step S5, the flotation process recovery rate empirical function is calculated according to formula (5):

[0088]

[0089] In formula (5), γ represents the recovery rate, η P Indicates the ore grade, η Te represents the soft measurement value of tailings grade, η Ce Represents the soft measurement value of concentrate grade.

[0090] In the embodiment, an experiment is conducted using production data from a copper flotation site. The experimental data are shown in Table 1.

[0091]

[0092] The root mean square error of soft measurement of concentrate grade is 0.88863%, the root mean square error of soft measurement of tailings grade is 0.02971%, and the root mean square error of soft measurement of recovery rate is 2.752%. The prediction accuracy of the prediction model under clustering stratified sampling of flotation conditions is the same as that of the prediction model under non-stratification, and the root mean square error is not much different.

[0093] By taking 122 sets of flotation condition historical data, the first 60% of the data set is used to construct 73 sets of training sets, and the last 40% of the data set is used to construct 49 sets of test sets. The model prediction comparison experimental results are shown in Table 2.

[0094]

[0095] The prediction performance of the indicator prediction model of the present application for concentrate grade and recovery rate is better than that of SVM, BPNN and LSTM.

[0096] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A soft-sensing method for indicators based on flotation mechanism and data hybrid drive, characterized in that: include, Step S1, preprocessing the initial sample set of flotation production conditions to construct a flotation production data set, wherein the preprocessing includes cleaning, interpolation and integration of the data in the initial sample set of flotation production conditions; Step S2, defining the concentrate grade and tailings grade, and building a flotation grade mechanism model by combining the mineral particle exchange relationship and mass balance relationship in the flotation process; Step S3, performing parameter identification on the flotation grade mechanism model based on a particle swarm optimization algorithm to continuously obtain system identification parameters that are most suitable for the current flotation working conditions; Step S4, combining the optimal system identification parameters and the flotation grade mechanism model to perform soft measurement on the flotation grade under the current flotation working condition, and determining the soft measurement value of the concentrate grade and the soft measurement value of the tailings grade; Step S5, obtaining a recovery rate index based on an empirical function of the recovery rate of the flotation process; In step S2, the flotation grade mechanism model is expressed by formula (3): In formula (3), ω1 represents the parameters to be identified in the first system, ω2 represents the parameters to be identified in the second system, α P represents the attachment and separation rate constant of mineral particles and bubbles in the slurry layer, α F represents the attachment and separation rate constant of mineral particles and bubbles in the foam layer, k R represents the rate constant of mineral particles falling from the foam layer to the slurry layer, r1 represents the first hyperparameter, r2 represents the second hyperparameter, V LP Indicates the volume of slurry in the slurry layer, V BP Indicates the volume of gas in the slurry layer, V LF Indicates the volume of slurry in the foam layer, Q A Indicates air velocity, Q C represents the mineral particle flow rate to the concentrate pool, Q T is the mineral particle flow rate to the tailings pond, M P is the ore quality, M C is the concentrate mass, M T is the mass of tailings, η C (t) represents the concentrate grade of the flotation cell at the current moment, η p (t) represents the ore grade of the flotation cell at the current moment, η T (t) represents the tailings grade of the flotation tank at the current moment, η C (t-1) represents the concentrate grade in the flotation cell at the previous moment, η T (t-1) represents the tailings grade in the flotation tank at the previous moment.

2. The indicator soft sensing method based on flotation mechanism and data hybrid drive according to claim 1 is characterized in that: In step S2, the defined concentrate grade is expressed by formula (1), and the defined tailings grade is expressed by formula (2), In formula (1), η C Indicates the concentrate grade, C MF Indicates the mass of target mineral particles per unit volume of the foam layer; In formula (2), η T Indicates the tailings grade, C MP It indicates the mass of target mineral particles per unit volume of the slurry layer.

3. The indicator soft sensing method based on flotation mechanism and data hybrid drive according to claim 1 is characterized in that: In the step S3, the flotation system parameter identification of the flotation grade mechanism model includes: Step S31, extracting sliding window data from the flotation production data set by a sliding window method, each sliding window data contains a set of input feature time series data; Step S32, taking the sliding window data as a flotation grade mechanism model parameter identification data set; Step S33, initializing the basic parameters of the particle swarm algorithm, including determining the number of particle swarms, the number of iterations, the particle dimension, the learning factor value, the inertia weight value range, the particle swarm position value range, the particle swarm velocity value range, and the initial position and initial velocity of the particle swarm, wherein each particle represents a set of identification parameters; Step S34, taking the flotation grade mechanism model parameter identification data set as input data of the flotation grade mechanism model, solving the concentrate grade and tailings grade at the current particle position according to the flotation grade mechanism model, and calculating the evaluation parameters for the solution result according to the fitness evaluation function; Step S35, repeating step S34, and determining the optimal system identification parameters under the minimum evaluation parameters through continuous iteration of the particle swarm algorithm; Step S36, repeating steps S31-S35, and continuously acquiring the system identification parameters that are most suitable for the current flotation working conditions.

4. The indicator soft sensing method based on flotation mechanism and data hybrid drive according to claim 3 is characterized in that: In step S3, the fitness evaluation function is calculated according to formula (4): In formula (4), Ω represents the system matrix parameter to be identified Ω = [ω1, ω2], represents the true value of the concentrate grade of the i-th group in the data set, represents the true value of tailings grade in the i-th group of the data set, It represents the predicted value of concentrate grade obtained by the flotation grade mechanism model after predicting the flotation conditions of the i-th group. It represents the predicted value of tailings grade after the flotation grade mechanism model predicts the i-th group of flotation conditions, N represents the number of sampling data, and λ represents the weight factor.

5. The indicator soft sensing method based on flotation mechanism and data hybrid drive according to claim 1 is characterized in that: In step S4, before the next sliding window data is updated, the equivalent working condition identification parameter [ω 1j ,ω 2j ] and the flotation grade mechanism model, determine the concentrate grade and tailings grade, and obtain the concentrate grade soft measurement value η Ce and tailings grade soft measurement value η Te .

6. The indicator soft sensing method based on flotation mechanism and data hybrid drive according to claim 1 is characterized in that: In step S5, the flotation process recovery rate empirical function is used to calculate the recovery rate according to formula (5): In formula (5), γ represents the recovery rate, η P Indicates the ore grade, η Te represents the soft measurement value of tailings grade, η Ce Represents the soft measurement value of concentrate grade.

Images