A method and device for measuring copper grade in copper ore flotation process

Abstract

The present application provides a method and device for measuring the copper grade in the flotation process of a copper ore, the method comprising: obtaining measurement data, and performing data enhancement processing on the measurement data to reduce the dimension of unlabeled samples in the measurement data, and constructing a training set and a test set based on the unlabeled samples and labeled samples in the measurement data; using BiLSTM as a learner to bidirectionally explore value information from the time series data of the training set; for redundant variables in the process data, removing the redundant variables to obtain a robust BiLSTM model, and measuring the copper grade in the flotation process of the copper ore based on the robust BiLSTM model. The present application can overcome the problems of unlabeled samples, outliers and redundant variables in the measurement data of the existing copper ore flotation process, and realize the rapid and accurate estimation of the copper grade, thereby facilitating the control and optimization of the copper ore flotation process.

Description

Technical Field

[0001] The present application relates to the field of detection technology, and in particular to a method and device for measuring the copper grade in a copper ore flotation process. Background Art

[0002] Flotation is a commonly used mineral separation method in copper ore processing, which separates useful copper ore from waste rock through the action of bubbles. First, after the raw ore is crushed and ground in the grinding system, the copper ore particles obtained are fed into a flotation tank and mixed with water to form a slurry. Then, the slurry is treated with surfactants and foaming agents. Through the continuous stirring of the agitator, the supplied air and water produce bubbles under the action of the foaming agent, in which the valuable mineral powder is attached to the bubbles under the action of the surfactant. Then the bubbles float to the top of the slurry to form a foam layer, from which the valuable minerals are collected and concentrated. Finally, the concentrate is transferred to a dewatering system to obtain the finished product of the flotation process. The ore particles that cannot float are discharged from the bottom of the flotation tank as tailings. Because the copper grade of the product reflects the quality of the final product, it is regarded as a key performance indicator. Therefore, it is critical for this process to monitor and control this variable in real time. However, in field operations, the measurement of variables requires manual sampling and laboratory testing by skilled operators, and it takes 15 minutes to obtain each result, which is time-consuming and costly. Therefore, the development of an accurate soft-sensing sensor is crucial to provide a faster and more convenient way to monitor the target variable - the copper grade of the product.

[0003] However, the development of soft sensor has the following problems: (1) The process is quite long, resulting in a significant time delay between input and output; (2) The working environment of the process is very harsh, resulting in many outliers in the measurement data; (3) The target variable is obtained through laboratory measurements at intervals of about 15 minutes. Other process variables, such as ore feed rate, overflow fineness, and slurry pH, are measured at intervals of tens of seconds. There are many unlabeled samples in the historical data, which may contain a lot of valuable information. Therefore, it is necessary to develop a measurement method to solve this problem. Summary of the invention

[0004] In view of this, the embodiments of the present application provide a method, device, electronic device and storage medium for measuring copper grade in a copper ore flotation process, which can overcome the problems of unlabeled samples, outliers and redundant variables in the measurement data of the existing copper ore flotation process, and achieve fast and accurate estimation of copper grade, thereby facilitating the control and optimization of the copper ore flotation process.

[0005] The technical solution of the embodiment of the present application is implemented as follows:

[0006] In a first aspect, an embodiment of the present application provides a method for measuring copper grade in a copper ore flotation process, the method comprising:

[0007] Acquire measurement data, perform data enhancement processing on the measurement data to reduce the dimension of unlabeled samples in the measurement data, and construct a training set and a test set based on the unlabeled samples and labeled samples in the measurement data; wherein the measurement data is measurement data of a copper ore flotation process;

[0008] Using BiLSTM as a learner to bidirectionally explore value information from the time series data of the training set, wherein the BiLSTM includes a loss function, and the loss function is used to improve resistance to outliers;

[0009] The redundant variables in the process data are eliminated to obtain a robust BiLSTM model, and the copper grade of the copper ore flotation process is measured based on the robust BiLSTM model.

[0010] In a second aspect, the embodiment of the present application further provides a device for measuring the copper grade in a copper ore flotation process, the device comprising:

[0011] A preprocessing module, used to obtain measurement data, perform data enhancement processing on the measurement data to reduce the dimension of unlabeled samples in the measurement data, and construct a training set and a test set based on the unlabeled samples and labeled samples in the measurement data; wherein the measurement data is measurement data of a copper ore flotation process;

[0012] A learning module, used for bidirectionally exploring value information from the time series data of the training set using BiLSTM as a learner, wherein the BiLSTM includes a loss function, and the loss function is used to improve resistance to outliers;

[0013] The measurement module is used to eliminate redundant variables in process data to obtain a robust BiLSTM model, and measure the copper grade of the copper ore flotation process based on the robust BiLSTM model.

[0014] In a third aspect, an embodiment of the present application further provides an electronic device, comprising: a processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium through the bus, and the processor executes the machine-readable instructions to execute the method for measuring copper grade in a copper ore flotation process as described in any one of the first aspects.

[0015] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for measuring the copper grade in a copper ore flotation process as described in any one of the first aspects is executed.

[0016] The embodiments of the present application have the following beneficial effects:

[0017] It overcomes the problems of unlabeled samples, outliers and redundant variables in the existing measurement data of copper ore flotation process, realizes fast and accurate estimation of copper grade, and thus helps to control and optimize the copper ore flotation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 is a flowchart of steps S101-S103 provided in an embodiment of the present application;

[0020] Figure 2 It is an algorithm framework diagram provided in the embodiment of the present application;

[0021] Figure 3 It is an overall flow chart provided by the embodiment of the present application;

[0022] Figure 4 It is a flotation process principle diagram provided in the embodiment of the present application;

[0023] Figure 5 It is a correlation matrix heat map provided in the embodiment of the present application;

[0024] Figure 6 is an outlier distribution diagram of the input variable 26 provided in an embodiment of the present application;

[0025] Figure 7 is an outlier distribution diagram of the target variable provided in the embodiment of the present application;

[0026] Figure 8 is a curve chart of the actual value and predicted value of copper grade provided in the embodiment of the present application;

[0027] Fig. 9 It is a schematic diagram of the structure of a device for measuring copper grade in a copper ore flotation process provided in an embodiment of the present application;

[0028] Fig.10It is a schematic diagram of the composition structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0029] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart under the guidance of the content of the present application, or remove one or more operations from the flowchart.

[0030] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0031] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0032] In the following description, the terms "first\second\third" involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first\second\third" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0033] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0035] See also Figure 1 , Figure 1 This is a flow chart of steps S101-S103 of the copper grade measurement method for copper ore flotation process provided in the embodiment of the present application, which will be combined with Figure 1 Steps S101-S103 are shown for explanation.

[0036] Step S101, obtaining measurement data, and performing data enhancement processing on the measurement data to reduce the dimension of unlabeled samples in the measurement data, and constructing a training set and a test set based on the unlabeled samples and labeled samples in the measurement data; wherein the measurement data is measurement data of a copper ore flotation process;

[0037] Step S102, using BiLSTM as a learner to bidirectionally explore value information from the time series data of the training set, wherein the BiLSTM includes a loss function, and the loss function is used to improve resistance to outliers;

[0038] Step S103: For redundant variables existing in the process data, the redundant variables are eliminated to obtain a robust BiLSTM model, and the copper grade of the copper ore flotation process is measured based on the robust BiLSTM model.

[0039] See also Figure 2 , Figure 2 It is an algorithm framework diagram provided in the embodiment of the present application, such as Figure 2 As shown, the embodiment of the present application is implemented by the following steps:

[0040] Step 1: Data enhancement is done through the t-SNE algorithm. This method reduces the dimension of unlabeled samples and concatenates them with labeled samples. The resulting data effectively captures the valuable information of unlabeled samples in the flotation process.

[0041] Step 2: BiLSTM is used as a basic learner to explore valuable information from time series data in both directions;

[0042] Step 3: A novel upper-bounded Huber loss function is proposed to improve the robustness of BiLSTM;

[0043] Step 4: Integrate the LASSO algorithm into BiLSTM to realize input variable selection.

[0044] See also Figure 3 , Figure 3 This is the overall flow chart provided by the embodiment of the present application, combined with Figure 2 and Figure 3 , Step 1 is for the original data set including the labeled data set , valuable information will be extracted from unlabeled data. The specific execution process is as follows:

[0045] 1) Set the target dimension from [0, D]; in the t-SNE algorithm, "target dimension" refers to the target dimension for embedding high-dimensional data into a low-dimensional space. Specifically: the input of the t-SNE algorithm is a set of high-dimensional data points, the dimension of which is usually D. The "target dimension" is the dimension of the low-dimensional space to which t-SNE wants to project these high-dimensional data points, usually 2 or 3 dimensions. t-SNE will try to maintain the relative distance and structure between the original high-dimensional data points in this low-dimensional space.

[0046] 2) Set represents high-dimensional data, where , Represents a set of m-dimensional real vectors. Vector and The similarity between them is measured using conditional probability express.

[0047] ;

[0048] in, express The Gaussian variance of . Get probability distribution (For each data point, choose an appropriate Gaussian variance so that can correctly reflect the local similarity between data points). are three non-conflicting data points in high-dimensional data, where Based on the perplexity Prep ( ) to search, the perplexity can be expressed as a close data point The number of data points. The perplexity Prep ( ) is calculated as follows:

[0049] Prep ( )= ;

[0050] in yes The entropy of .

[0051] 3) Obtained conditional probability and , and The similarity between them is determined by the joint probability It is expressed as:

[0052] ; Where n represents the total number of data points in the data set;

[0053] 4) The variables extracted in the target dimension should be , and use t-SNE to obtain the probability of the data. Assume , and yes The data in = .Then, and The joint probability distribution of can be calculated as follows:

[0054] ;

[0055] 5) Kullback-Leibler (KL) divergence is calculated as follows:

[0056] ;

[0057] Among them, express The probability distribution of other corresponding data, other corresponding data refers to the probability distribution of a given data point For example, all other data points in the low-dimensional space. and The similarity between them is minimized and the KL divergence is updated by the gradient descent method. as follows:

[0058] ;

[0059] Momentum is introduced into gradient descent to avoid falling into local optimality and improve optimization efficiency. The update expression is as follows:

[0060] ;

[0061] in, represents the number of solution iterations, represents the learning rate, Represents iteration The momentum of the next time.

[0062] 6) When the prediction error does not decrease or when the maximum iteration is met, the loop is terminated. Finally, a sample containing labels is provided. And the samples after dimensionality reduction of unlabeled data ,Right now .

[0063] In this step, t-SNE technology provides an effective means to explore valuable information of unlabeled data, reduce the time and resources required for model training, and help prevent the occurrence of dimensionality disaster.

[0064] Step 2 selects BiLSTM as the basic learner, where the working mechanism of LSTM is shown in the following formula:

[0065] ;

[0066] ;

[0067] ;

[0068] ;

[0069] ;

[0070] ;

[0071] in, , and Represent the forget gate, input gate and output gate respectively; is the cell state; represents the candidate cell state, is a hidden state, represents the hidden state of the previous time step; Represents different weight matrices; Represents different bias vectors; and denote the Sigmoid function and the hyperbolic tangent function respectively; ⊙ denotes the Hadamard product; express The input value at the time.

[0072] The BiLSTM, which consists of two LSTMs, can fully capture and integrate the bidirectional information of the data. One LSTM is a conventional forward LSTM, and the other is a backward LSTM that learns the reverse order of the input. The output of the model is by combining the forward output and backward output Got it.

[0073] ;

[0074] ;

[0075] ;

[0076] in, represents element-wise summation, is the weight matrix of the output layer, Represents the bias vector.

[0077] Step 3 is to design a Huber loss with an upper limit for outliers in the data. The traditional BiLSTM model uses mean square error as the loss function. However, BiLSTM is very sensitive to outliers in the training data set. Therefore, in the embodiment of the present application, a Huber loss function with an upper limit is designed to improve the resistance to outliers of different degrees. The formula is as follows:

[0078] ;

[0079] in is a hyperparameter indicating the tolerance to outliers. After that, the loss no longer increases, thus ignoring the abnormal effect; is a threshold value used to determine when and in which section of the loss function the mean square error loss is used. The proposed loss is essentially a piecewise function with five intervals: 1) When When is the mean square error loss; 2) when When , it is a linear loss; 3) When , the loss is zero.

[0080] Compared with the original Huber loss, the Huber loss with upper bound assigns a zero value to each extreme outlier, thereby eliminating the influence of these outliers and obtaining a robust BiLSTM.

[0081] Step 4 introduces the LASSO algorithm to eliminate the redundant variables in the process data.

[0082] In the embodiment of the present application, the LASSO penalty term is incorporated into the robust BiLSTM to reduce the model complexity and avoid overfitting. First, the regularization operator Introduced into the input , then, Replace with , the optimization expression formula of the robust BiLSTM with LASSO is:

[0083] ;

[0084] in and Represents the upper bound Huber loss of BiLSTM; represents the penalty coefficient, represents the true value, Represents the set of all true values ​​in the entire training dataset.

[0085] Before executing the proposed algorithm, Select the best hyperparameters ,in yes The upper bound of , so that all input variables shrink to zero. Then, solve the above equation to get the optimal , and shrink the coefficients of redundant input variables to achieve variable selection and model optimization. Replace with , a compact and robust BiLSTM model is obtained.

[0086] In some embodiments, a flotation process of a copper ore concentrator is used, and the flotation process principle diagram is as follows: Figure 4 As shown, the data set is obtained from the automated system of the process, and consists of 860 labeled samples, 12040 unlabeled samples and 29 input variables, and the input variables are shown in Table 1. It should be noted that the measurement data in the embodiment of the present application is the measurement data of the copper ore flotation process (i.e., the candidate input variables of the flotation process in Table 1).

[0087] Table 1 Candidate input variables for copper grade flotation process

[0088]

[0089] See also Figure 5 , Figure 5 is a correlation matrix heat map provided in the embodiment of the present application, such as Figure 5 As shown in , where the rows and columns represent the correlation between the input variables and the target variable. It can be observed from the figure that some candidate input variables have significant correlations, indicating that there is redundancy between the input variables, which will reduce the performance of the model and increase the difficulty of variable selection. Therefore, it is necessary to develop an algorithm for variable selection problems that can effectively handle highly correlated redundant variables in nonlinear models.

[0090] In addition, outlier detection is performed using the Anomaly Detection Toolkit by selecting one input variable with the highest correlation with the target variable along with the target variable, see Figure 6 and Figure 7 , Figure 6 is an outlier distribution diagram of the input variable 26 provided in the embodiment of the present application, Figure 7 is an outlier distribution diagram of the target variable provided in the embodiment of the present application, Figure 6 and Figure 7The detection results are shown. Obviously, there are many outliers of varying degrees in both the input and output variables, which reduce the robustness of the model. Therefore, it is necessary to improve the resistance to model outliers to enhance its robustness.

[0091] In summary, the development of soft sensor for this process involves the following difficulties: (1) The process time is quite long, resulting in a significant time delay between input and output. (2) The working environment of the process is extremely harsh, resulting in many outliers in the measurement data. (3) The value of the target variable is obtained through laboratory tests, which are conducted approximately every 15 minutes. Other process variables, such as ore feed rate, overflow fineness, and slurry pH, are measured every tens of seconds. Therefore, there are many unlabeled samples in the historical data, which may contain a lot of valuable information. Therefore, it is necessary to develop a semi-supervised robust soft sensor method to solve these problems.

[0092] In the present embodiment, the first 80% of the data set is used as training data, and the remaining 20% ​​is used as test data. The performance of all algorithms is evaluated based on the root mean squared error (RMSE), mean absolute percentage error (MAPE), and correlation coefficient (COR).

[0093] ;

[0094] ;

[0095] ;

[0096] in, Represents the total number of samples in the test data set; and They are The measured and predicted values ​​of samples; express and The covariance of and They are and The variance of .

[0097] Table 2 gives the simulation results, which shows that the developed algorithm outperforms other algorithms in all indicators; Figure 8 is a curve chart of the actual value and predicted value of copper grade provided in the embodiment of the present application, Figure 8The predicted values ​​of the target variable using the proposed algorithm are presented, indicating that the developed algorithm is effective in dynamically following the changes in the target variable values.

[0098] Table 2 Simulation results of copper grade prediction using different algorithms

[0099]

[0100] In summary, the embodiments of the present application have the following beneficial effects:

[0101] It overcomes the problems of unlabeled samples, outliers and redundant variables in the existing measurement data of copper ore flotation process, realizes fast and accurate estimation of copper grade, and thus contributes to the control and optimization of copper ore flotation process.

[0102] Based on the same inventive concept, the embodiments of the present application also provide a device for measuring the copper grade in the copper ore flotation process, which corresponds to the method for measuring the copper grade in the copper ore flotation process in the first embodiment. Since the principle of solving the problem by the device in the embodiments of the present application is similar to the method for measuring the copper grade in the copper ore flotation process, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0103] like Fig. 9 As shown, Fig. 9 1 is a schematic diagram of the structure of a copper grade measuring device 900 for a copper ore flotation process provided in an embodiment of the present application. The copper grade measuring device 900 for a copper ore flotation process comprises:

[0104] A preprocessing module 901 is used to obtain measurement data, perform data enhancement processing on the measurement data to reduce the dimension of unlabeled samples in the measurement data, and construct a training set and a test set based on the unlabeled samples and labeled samples in the measurement data; wherein the measurement data is measurement data of a copper ore flotation process;

[0105] A learning module 902, configured to use BiLSTM as a learner to bidirectionally explore value information from the time series data of the training set, wherein the BiLSTM includes a loss function, and the loss function is used to improve resistance to outliers;

[0106] The measurement module 903 is used to eliminate redundant variables in the process data to obtain a robust BiLSTM model, and measure the copper grade of the copper ore flotation process based on the robust BiLSTM model.

[0107] Those skilled in the art should understand that Fig. 9The functions of the various units in the device 900 for measuring the copper grade in the copper ore flotation process shown can be understood by referring to the related description of the method for measuring the copper grade in the copper ore flotation process described above. Fig. 9 The functions of the various units in the copper grade measuring device 900 for copper ore flotation process shown can be implemented by a program running on a processor, or can also be implemented by a specific logic circuit.

[0108] In a possible implementation manner, the preprocessing module 901 performs data enhancement processing on the measurement data, including:

[0109] Set the target dimension from [0, D];

[0110] set up represents high-dimensional data, where , Represents a set of m-dimensional real vectors; vector and The similarity between them is measured using conditional probability express;

[0111] ;

[0112] in, express Gaussian variance of Get probability distribution , are three non-conflicting data points in high-dimensional data, where Based on the perplexity Prep ( ) to search, and the perplexity is expressed as the number of close data points The number of data points;

[0113] Prep ( )= ;

[0114] in yes Entropy of

[0115] In obtaining the conditional probability and back, and The similarity between them is determined by the joint probability It is expressed as:

[0116] ; Where n represents the total number of data points in the data set;

[0117] The variables extracted in the target dimension are , and use t-SNE to obtain the probability of the data; assuming , and yes The data in = , and The joint probability distribution of is calculated as follows:

[0118] ;

[0119] The Kullback-Leibler divergence is calculated as follows:

[0120] ;

[0121] Among them, express The probability distribution of other corresponding data, other corresponding data refers to the probability distribution of a given data point For example, all other data points in the low-dimensional space; in order to promote and The similarity between them is minimized and the KL divergence is updated by the gradient descent method; where the gradient for:

[0122] ;

[0123] Momentum is introduced into gradient descent to avoid falling into local optimality and improve optimization efficiency; The update expression is:

[0124] ;

[0125] in, represents the number of solution iterations, represents the learning rate, Represents iteration Momentum at the time;

[0126] When the prediction error does not decrease or when the maximum iteration is met, the loop is terminated. Finally, a sample containing labels is provided. And unlabeled samples after unlabeled data dimensionality reduction ,Right now .

[0127] In a possible implementation manner, the preprocessing module 901 constructs a training set and a test set based on the unlabeled samples and the labeled samples in the measurement data, including:

[0128] Concatenate the unlabeled samples with the labeled samples to obtain a target set;

[0129] The target set is divided into the training set and the test set.

[0130] In one possible implementation, the working mechanism of LSTM is shown by the following formula:

[0131] ;

[0132] ;

[0133] ;

[0134] ;

[0135] ;

[0136] ;

[0137] in, , and Represent the forget gate, input gate and output gate respectively; is the cell state; represents the candidate cell state, is the hidden state, represents the hidden state of the previous time step; Represents different weight matrices; Represents different bias vectors; and denote the Sigmoid function and the hyperbolic tangent function respectively; ⊙ denotes the Hadamard product; express The input value at the moment.

[0138] The BiLSTM consists of two LSTMs, one of which is a regular forward LSTM and the other is a backward LSTM that learns the reverse order input. The output of the model is is by combining the forward output and backward output obtained;

[0139] ;

[0140] ;

[0141] ;

[0142] in, represents element-wise summation, is the weight matrix of the output layer, Represents the bias vector.

[0143] In a possible implementation, the BiLSTM uses Huber with an upper limit as a loss function to improve resistance to outliers. The formula is:

[0144] ;

[0145] in is a hyperparameter that indicates the tolerance to outliers. When the error reaches After that, the loss no longer increases, thus ignoring the abnormal effect; is a threshold value used to determine when and in which section of the loss function the mean square error loss is used; the loss is a piecewise function with five intervals. When is the mean square error loss; when When , it is a linear loss; when , the loss is zero.

[0146] In a possible implementation, the measurement module 903 removes redundant variables in the process data to obtain a robust BiLSTM model, including:

[0147] The LASSO algorithm is integrated into the BiLSTM to achieve the selection of input variables, thereby obtaining the robust BiLSTM model.

[0148] In a possible implementation, the measurement module 903 integrates the LASSO algorithm into the BiLSTM to implement the selection of input variables, thereby obtaining the robust BiLSTM model, including:

[0149] The regularization operator Introduced into the input , then, Replace with , the optimization expression formula of the robust BiLSTM with LASSO is:

[0150] ;

[0151] in, and Represents the upper bound Huber loss of BiLSTM; represents the penalty coefficient, represents the true value, Represents the set of all true values ​​in the entire training data set;

[0152] Before executing the algorithm, from the set Select the best hyperparameters ,in yes The upper bound of , so that all input variables shrink to zero; by solving the optimal , and shrink the coefficients of redundant input variables to achieve variable selection and model optimization; Replace with , and obtain a robust BiLSTM model.

[0153] The above-mentioned copper grade measurement device for the copper ore flotation process overcomes the problems of unlabeled samples, outliers and redundant variables in the existing measurement data of the copper ore flotation process, and realizes fast and accurate estimation of copper grade, thereby facilitating the control and optimization of the copper ore flotation process.

[0154] like Fig.10 As shown, Fig.10 The present invention provides a schematic diagram of the structure of an electronic device 1000, wherein the electronic device 1000 includes:

[0155] A processor 1001, a storage medium 1002 and a bus 1003, wherein the storage medium 1002 stores machine-readable instructions executable by the processor 1001. When the electronic device 1000 is running, the processor 1001 communicates with the storage medium 1002 via the bus 1003, and the processor 1001 executes the machine-readable instructions to execute the steps of the method for measuring the copper grade in the copper ore flotation process described in the embodiment of the present application.

[0156] In actual application, the components in the electronic device 1000 are coupled together via the bus 1003. It is understood that the bus 1003 is used to realize the connection and communication between these components. In addition to the data bus, the bus 1003 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, Fig.10 Various buses are labeled as bus 1003.

[0157] The electronic device overcomes the problems of unlabeled samples, outliers and redundant variables in the measurement data of the existing copper ore flotation process, and achieves a fast and accurate estimation of the copper grade, thereby facilitating the control and optimization of the copper ore flotation process.

[0158] The embodiment of the present application also provides a computer-readable storage medium, which stores executable instructions. When the executable instructions are executed by at least one processor 1001, the method for measuring copper grade in a copper ore flotation process described in the embodiment of the present application is implemented.

[0159] In some embodiments, the storage medium can be a magnetic random access memory (FRAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a flash memory, a magnetic surface storage, an optical disk, or a compact disc read-only memory (CD-ROM); it can also be various devices including one or any combination of the above memories.

[0160] In some embodiments, executable instructions may be in the form of a program, software, software module, script or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a stand-alone program or as a module, component, subroutine or other unit suitable for use in a computing environment.

[0161] As an example, executable instructions may, but do not necessarily, correspond to a file in a file system, may be stored as part of a file that stores other programs or data, for example, in one or more scripts in a HyperText Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple coordinated files (for example, files storing one or more modules, subroutines, or code portions).

[0162] By way of example, executable instructions may be deployed to be executed on one computing device, or on multiple computing devices located at one site, or on multiple computing devices distributed across multiple sites and interconnected by a communication network.

[0163] The computer-readable storage medium overcomes the problems of unlabeled samples, outliers and redundant variables in the measurement data of the existing copper ore flotation process, and achieves a fast and accurate estimation of the copper grade, thereby facilitating the control and optimization of the copper ore flotation process.

[0164] In the several embodiments provided in the present application, it should be understood that the disclosed methods and electronic devices can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0165] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0166] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0167] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a platform server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard drives, ROM, RAM, magnetic disks, or optical disks.

[0168] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for measuring copper grade in a copper ore flotation process, characterized in that: The method comprises: Acquire measurement data, perform data enhancement processing on the measurement data to reduce the dimension of unlabeled samples in the measurement data, and construct a training set and a test set based on the unlabeled samples and labeled samples in the measurement data; wherein the measurement data is measurement data of a copper ore flotation process; Using BiLSTM as a learner to bidirectionally explore value information from the time series data of the training set, wherein the BiLSTM includes a loss function, and the loss function is used to improve resistance to outliers; For redundant variables existing in process data, redundant variables are eliminated to obtain a robust BiLSTM model, and the copper grade of the copper ore flotation process is measured based on the robust BiLSTM model; The performing data enhancement processing on the measurement data includes: Set the target dimension from [0, D]; set up represents high-dimensional data, where , Represents a set of m-dimensional real vectors; vector and The similarity between them is measured using conditional probability express; ； in, express Gaussian variance of Get probability distribution , are three non-conflicting data points in high-dimensional data, where Based on the perplexity Prep ( ) to search, and the perplexity is expressed as the number of close data points The number of data points; Prep ( )= ; in yes Entropy of In obtaining the conditional probability and back, and The similarity between them is determined by the joint probability It is expressed as: ; Where n represents the total number of data points in the data set; The variables extracted in the target dimension are , and use t-SNE to obtain the probability of the data; assuming , and yes The data in , where the set of variables extracted in the target dimension = , and The joint probability distribution of is calculated as follows: ； The Kullback-Leibler divergence is calculated as follows: ； in, express The probability distribution of other corresponding data, other corresponding data refers to the probability distribution of a given data point For example, all other data points in the low-dimensional space; in order to promote and The similarity between them is minimized and the KL divergence is updated by the gradient descent method; where the gradient for: ； Momentum is introduced in gradient descent; The update expression is: ； in, represents the number of solution iterations, represents the learning rate, Represents iteration Momentum at the time; When the prediction error does not decrease or when the maximum iteration is met, the loop is terminated; finally, a sample containing labels is provided. And unlabeled samples after unlabeled data dimensionality reduction ,Right now .

2. The method according to claim 1, characterized in that The constructing of a training set and a test set based on the unlabeled samples and the labeled samples in the measurement data includes: Concatenate the unlabeled samples with the labeled samples to obtain a target set; The target set is divided into the training set and the test set.

3. The method according to claim 1, characterized in that The working mechanism of LSTM is shown in the following formula: ； ； ； ； ； ； in, , and Represent the forget gate, input gate and output gate respectively; is the cell state; represents the candidate cell state, is a hidden state, represents the hidden state of the previous time step; Represents different weight matrices; Represents different bias vectors; and denote the Sigmoid function and the hyperbolic tangent function respectively; ⊙ denotes the Hadamard product; express The input value at the time. The BiLSTM consists of two LSTMs, one of which is a forward LSTM and the other is a backward LSTM that learns the reverse order input; the output of the model is by combining the forward output and backward output obtained; ； ； ； in, represents element-wise summation, is the weight matrix of the output layer, Represents the bias vector.

4. The method according to claim 3, characterized in that The BiLSTM uses the Huber upper limit as the loss function to improve resistance to outliers. The formula is: ； in is a hyperparameter that indicates the tolerance to outliers. When the error reaches After that, the loss no longer increases, thus ignoring the abnormal effect; is a threshold value used to determine when and in which section of the loss function the mean square error loss is used; the loss is a piecewise function with five intervals. When is the mean square error loss; when When , it is a linear loss; when , the loss is zero.

5. The method according to claim 1, characterized in that The method removes redundant variables in the process data to obtain a robust BiLSTM model, including: The LASSO algorithm is integrated into the BiLSTM to achieve the selection of input variables, thereby obtaining the robust BiLSTM model.

6. The method according to claim 5, characterized in that The LASSO algorithm is integrated into the BiLSTM to select input variables, thereby obtaining the robust BiLSTM model, including: The regularization operator Introduced into the input , then, Replace with , the optimization expression formula of the robust BiLSTM with LASSO is: ； in, and Represents the upper bound Huber loss of BiLSTM; represents the penalty coefficient, represents the true value, Represents the set of all true values ​​in the entire training data set; Before executing the algorithm, from the set Select the best hyperparameters ,in yes The upper bound of , so that all input variables shrink to zero; by solving the optimal , and shrink the coefficients of redundant input variables to achieve variable selection and model optimization; Replace with , and obtain a robust BiLSTM model.

7. A device for measuring copper grade in copper ore flotation process, characterized in that: The device comprises: A preprocessing module is used to obtain measurement data and perform data enhancement processing on the measurement data to reduce the dimension of unlabeled samples in the measurement data, and construct a training set and a test set based on the unlabeled samples and labeled samples in the measurement data; wherein the measurement data is measurement data of a copper ore flotation process; the data enhancement processing on the measurement data includes: Set the target dimension from [0, D]; set up represents high-dimensional data, where , Represents a set of m-dimensional real vectors; vector and The similarity between them is measured using conditional probability express; ； in, express Gaussian variance of Get probability distribution , are three non-conflicting data points in high-dimensional data, where Based on the perplexity Prep ( ) to search, and the perplexity is expressed as the number of close data points The number of data points; Prep ( )= ; in yes Entropy of In obtaining the conditional probability and back, and The similarity between them is determined by the joint probability It is expressed as: ; Where n represents the total number of data points in the data set; The variables extracted in the target dimension are , and use t-SNE to obtain the probability of the data; assuming , and yes The data in , where the set of variables extracted in the target dimension = , and The joint probability distribution of is calculated as follows: ； The Kullback-Leibler divergence is calculated as follows: ； in, express The probability distribution of other corresponding data, other corresponding data refers to the probability distribution of a given data point For example, all other data points in the low-dimensional space; in order to promote and The similarity between them is minimized and the KL divergence is updated by the gradient descent method; where the gradient for: ； Momentum is introduced in gradient descent; The update expression is: ； in, represents the number of solution iterations, represents the learning rate, Represents iteration Momentum at the time; When the prediction error does not decrease or when the maximum iteration is met, the loop is terminated; finally, a sample containing labels is provided. And unlabeled samples after unlabeled data dimensionality reduction ,Right now ; A learning module, used for bidirectionally exploring value information from the time series data of the training set using BiLSTM as a learner, wherein the BiLSTM includes a loss function, and the loss function is used to improve resistance to outliers; The measurement module is used to eliminate redundant variables in process data to obtain a robust BiLSTM model, and measure the copper grade of the copper ore flotation process based on the robust BiLSTM model.

8. An electronic device, characterized in that: include: A processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to execute the method for measuring copper grade in a copper ore flotation process as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for measuring the copper grade in a copper ore flotation process according to any one of claims 1 to 6 is executed.

Images