Thickener feeding amount prediction method, device and equipment and storage medium

Abstract

The invention relates to the technical field of data processing, and discloses a thickener feeding amount prediction method and device, equipment and a storage medium. The method comprises the steps that original beneficiation data are obtained and preprocessed, preprocessed beneficiation data are obtained, and the preprocessed beneficiation data comprise a training set, a verification set and a test set; constructing an intelligent sensing model by using the GRU network, and training the intelligent sensing model through the training set to obtain a trained intelligent sensing model; optimizing hyper-parameters of the trained intelligent sensing model through a verification set by adopting a particle swarm algorithm to obtain an optimized intelligent sensing model; and inputting the test set into the optimized intelligent sensing model, and outputting a predicted value of the feeding amount of the thickener through the optimized intelligent sensing model. According to the method, the online prediction problem of the ore feeding amount of the thickener is solved, a basis is provided for optimal control of the thickener, and the accuracy and practicability of the thickener are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly relates to a method, device, equipment and storage medium for predicting the feeding amount of a thickener. Background Art

[0002] Thickening and dewatering is one of the key processes in the beneficiation process, which has a direct impact on the quality of concentrate products. In actual production, the thickening and filtration production operation is based on the judgment of the operating state of the thickener, and the feeding amount of the thickener is very important. The feeding amount of the thickener is a non-linear and dynamic variable. Industrial field operations often rely on manual experience, which is blind, making it difficult to control the underflow concentration of the thickener and difficult to coordinate the production of thickening and filtration, directly affecting the quality of concentrate products. Therefore, to optimize the control of the thickener, it is necessary to detect the feeding amount of the thickener. Currently, the detection of ore quantity depends on off-line sampling or the installation of on-line analysis devices. However, due to the limitations of the industrial field environment and economic conditions, the current detection methods cannot meet the requirements of real-time and accurate measurement. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to overcome the deficiencies in the prior art and provide a method, device, equipment and storage medium for predicting the feeding amount of a thickener.

[0004] The present invention provides the following technical solutions:

[0005] In a first aspect, the present invention provides a method for predicting the feeding amount of a thickener, the method comprising:

[0006] Obtain original ore dressing data, preprocess the original ore dressing data to obtain preprocessed ore dressing data, and the preprocessed ore dressing data includes a training set, a validation set and a test set;

[0007] Use a GRU network to construct an intelligent perception model, and train the intelligent perception model through the training set to obtain a trained intelligent perception model;

[0008] Adopt a particle swarm algorithm and optimize the hyperparameters of the trained intelligent perception model through the validation set to obtain an optimized intelligent perception model;

[0009] Input the test set into the optimized intelligent perception model, and output the predicted value of the feeding amount of the thickener through the optimized intelligent perception model.

[0010] In an optional embodiment, the original ore dressing data includes at least one of the feeding amount of the semi-autogenous mill, the frequency of the grinding slag slurry pump, the current of the grinding slag slurry pump, the flow rate of the grinding hydrocyclone, the grade of the flotation raw ore, the grade of the flotation concentrate and the grade of the flotation tailings.

[0011] In an alternative embodiment, the preprocessing of the original ore dressing data to obtain the preprocessed ore dressing data includes:

[0012] Identifying missing values in the original ore dressing data and forward-filling the missing values to obtain the filled ore dressing data;

[0013] Detecting abnormal data in the filled ore dressing data and removing the abnormal data to obtain the ore dressing data after removal;

[0014] Performing standardization processing on the ore dressing data after removal to obtain the preprocessed ore dressing data.

[0015] In an alternative embodiment, the intelligent perception model includes an update gate and a reset gate. The training of the intelligent perception model using the training set to obtain the trained intelligent perception model includes:

[0016] Obtaining the hidden state at the previous moment, and using the training set at the current moment and the hidden state at the previous moment as the training set of the intelligent perception model at the current moment;

[0017] The reset gate performs a linear transformation on the training set at the current moment and the hidden state at the previous moment through the Sigmoid activation function to obtain the reset gate output;

[0018] The update gate performs a linear transformation on the training set at the current moment and the hidden state at the previous moment through the Sigmoid activation function to obtain the update gate output;

[0019] Calculating a candidate hidden state based on the training set at the current moment, the reset gate output, the hidden state at the previous moment, and the weight matrix;

[0020] Calculating the final hidden state at the current moment based on the update gate output, the candidate hidden state, and the hidden state at the previous moment, and performing inverse standardization processing on the final hidden state at the current moment to obtain the predicted value of the training set;

[0021] Obtaining the true value of the training set, and calculating the loss value of the training set using the loss function based on the true value and the predicted value of the training set;

[0022] Calculating the gradient of the loss function with respect to each parameter of the intelligent perception model based on the loss value of the training set;

[0023] Adjusting the parameters of the intelligent perception model according to the gradient of each parameter of the intelligent perception model to obtain the trained intelligent perception model.

[0024] In an alternative embodiment, after obtaining the trained intelligent perception model, the following steps are further included:

[0025] Input the validation set into the trained intelligent perception model to obtain the predicted values of the validation set;

[0026] Obtain the true values of the validation set, and calculate the root mean square error and average relative error of the trained intelligent perception model based on the predicted values and true values of the test set;

[0027] Evaluate the performance of the trained intelligent perception model based on the root mean square error and the average relative error to obtain an evaluation result.

[0028] In an alternative embodiment, the calculating the root mean square error and average relative error of the trained intelligent perception model based on the predicted values and true values of the test set includes:

[0029] The formula for calculating the root mean square error of the trained intelligent perception model based on the predicted values and true values of the test set is:

[0030]

[0031] The formula for calculating the average relative error of the trained intelligent perception model based on the predicted values and true values of the test set is:

[0032]

[0033] In the formula, RMSE is the root mean square error, MRE is the average relative error, y i is the true value of the i-th sample of the validation set, pred i is the predicted value of the i-th sample of the validation set, and N is the number of samples in the validation set.

[0034] In an alternative embodiment, the optimizing the hyperparameters of the trained intelligent perception model by using the particle swarm algorithm through the validation set to obtain an optimized intelligent perception model includes:

[0035] S1: Randomly generate a group of particles, each particle including a position vector and a velocity vector;

[0036] S2: Construct a corresponding intelligent perception model for the position vector of each particle, and calculate the loss of each particle's intelligent perception model on the validation set to obtain the current fitness value of each particle;

[0037] S3: Obtain the historical best fitness value of each of the said particles. By comparing the current fitness value of each of the said particles with its historical best fitness value, update the individual optimal position of each of the said particles;

[0038] S4: Compare the current fitness values of all the said particles, determine the global optimal fitness value and its corresponding particle position, and update the global optimal position of the population;

[0039] S5: Adjust the velocity and position of each of the said particles according to the individual optimal position, the global optimal position and the current velocity of each of the said particles;

[0040] S6: Repeat S2 - S5, determine whether the preset number of iterations is reached. If so, converge and generate the optimized intelligent perception model.

[0041] In a second aspect, the present invention provides a thickener feed rate prediction device, and the device includes:

[0042] A preprocessing module, configured to obtain original ore dressing data, preprocess the original ore dressing data to obtain preprocessed ore dressing data, and the preprocessed ore dressing data includes a training set, a validation set and a test set;

[0043] A training module, configured to construct an intelligent perception model using a GRU network, and train the intelligent perception model through the training set to obtain a trained intelligent perception model;

[0044] An optimization module, configured to optimize the hyperparameters of the trained intelligent perception model by using a particle swarm algorithm and through the validation set to obtain an optimized intelligent perception model;

[0045] A prediction module, configured to input the test set into the optimized intelligent perception model, and output a thickener feed rate prediction value through the optimized intelligent perception model.

[0046] In a third aspect, an embodiment of the present disclosure provides a computer device, which includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the thickener feed rate prediction method described in the first aspect.

[0047] In a fourth aspect, an embodiment of the present disclosure provides a computer - readable storage medium, which stores a computer program, and when the computer program is executed by a processor, it implements the steps of the thickener feed rate prediction method described in the first aspect.

[0048] Advantages of the present application:

[0049] The method for predicting the feeding amount of a thickener provided by the embodiment of the present application obtains original ore dressing data, preprocesses the original ore dressing data to obtain preprocessed ore dressing data, where the preprocessed ore dressing data includes a training set, a validation set, and a test set; constructs an intelligent perception model using a GRU network, and trains the intelligent perception model through the training set to obtain a trained intelligent perception model; uses a particle swarm algorithm to optimize the hyperparameters of the trained intelligent perception model through the validation set to obtain an optimized intelligent perception model; inputs the test set into the optimized intelligent perception model, and outputs a predicted value of the thickener feeding amount through the optimized intelligent perception model. The present application solves the problem of online prediction of the ore feeding amount of the thickener, provides a basis for the optimized control of the thickener, and greatly improves the accuracy and practicability of the thickener.

[0050] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following specifically enumerates preferred embodiments and, in conjunction with the accompanying drawings, makes a detailed description as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts. In each drawing, similar components are numbered similarly.

[0052] Figure 1 Shows a flowchart of a method for predicting the feeding amount of a thickener provided by an embodiment of the present application;

[0053] Figure 2 Shows a flowchart of another method for predicting the feeding amount of a thickener provided by an embodiment of the present application;

[0054] Figure 3 Shows a flowchart of yet another method for predicting the feeding amount of a thickener provided by an embodiment of the present application;

[0055] Figure 4 Shows a schematic structural diagram of a device for predicting the feeding amount of a thickener provided by an embodiment of the present application;

[0056] Figure 5 Shows a schematic structural diagram of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] Embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are only used to explain the present invention and should not be construed as a limitation of the present invention.

[0058] It should be noted that the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "a plurality of" means two or more unless otherwise specifically defined.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs. The terms used in the specification of this template are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0060] Embodiment 1

[0061] As Figure 1 shown, it is a flowchart of a method for predicting the feed amount of a thickener in an embodiment of the present application. The method for predicting the feed amount of a thickener provided by the embodiment of the present application includes the following steps:

[0062] Step S110, obtain the original ore dressing data, preprocess the original ore dressing data to obtain the preprocessed ore dressing data, and the preprocessed ore dressing data includes a training set, a validation set, and a test set.

[0063] It can be understood that the ore dressing process is a continuous process, including processes such as crushing, grinding, flotation, and thickening and filtration. Each link is closely connected. The output of the previous step is usually the input of the next step, forming a tight process chain. From prior knowledge, it can be known that the feed amount of the thickener has a close relationship with the grinding and flotation processes. The grinding and flotation processes are relatively complex, with many devices and sensors. In this application, auxiliary variables with relatively strong correlation with the feed amount of the thickener in the grinding and flotation processes are selected, that is, the original ore dressing data, including but not limited to the feed amount of the semi-autogenous mill, the frequency of the grinding slag slurry pump, the current of the grinding slag slurry pump, the flow rate of the grinding cyclone, the grade of the original ore in flotation, the grade of the concentrate in flotation, and the grade of the tailings in flotation.

[0064] In this embodiment, the original mineral processing data can be detected in real time at the industrial site and saved in the database. The original mineral processing data is the actual operation data of the factory, and the data samples within the selected time period should cover various situations of normal operation of the process.

[0065] It should be noted that the original mineral processing data comes directly from the detection equipment on the industrial site, and it is inevitable that there are missing values ​​and random errors. If the original mineral processing data is directly used for subsequent model establishment and training without processing, the prediction accuracy of the model will be reduced. Therefore, data preprocessing is required:

[0066] (1) If the equipment or sensor data changes during sampling, the data will not be sampled normally. Therefore, it is necessary to identify the missing values ​​in the original mineral processing data and fill the missing values ​​forward, that is, fill the data at the missing value moment with the value of the previous moment to obtain the filled mineral processing data;

[0067] (2) Detecting abnormal data in the filled beneficiation data and removing the abnormal data. In this embodiment, a sliding median method can be used to detect and remove abnormal data. For example, a moving window with a length of 70 is selected, and the window is moved forward in sequence. The data in the window that differs from the local median by more than five times the local standard deviation is defined as an outlier, that is, abnormal data, and is removed to obtain the mineral processing data after removal;

[0068] (3) There are large differences in the values ​​and distributions of different auxiliary variables, and they cannot be directly used as the input of the model. In order to eliminate the influence of different dimensions of input variables, it is necessary to standardize the eliminated mineral processing data and convert them into data with a mean of 0 and a standard deviation of 1 to obtain the preprocessed mineral processing data. The standardization formula is as follows:

[0069]

[0070] In the formula, X i is the mineral processing data after preprocessing, x i is the beneficiation data after elimination, x max is the maximum value of the mineral processing data after elimination, x min It is the minimum value in the mineral processing data after elimination.

[0071] The preprocessed mineral processing data is divided into training set, validation set and test set. The training set is used for model training, the validation set is used to evaluate model performance and hyperparameter tuning, and the test set is used to evaluate model performance.

[0072] The above steps ensure the integrity and accuracy of the input data by identifying and filling missing values, detecting and removing abnormal data, and reducing the impact of noise on model training. Through standardization, auxiliary variables with different dimensions are converted into a unified scale, enabling the model to more effectively learn the relationships between variables and providing an objective and accurate data basis for subsequent modeling and prediction.

[0073] Step S120: Construct an intelligent perception model using a GRU network and train the intelligent perception model with the training set to obtain a trained intelligent perception model.

[0074] Understandably, GRU (Gated Recurrent Unit) is a simplified variant of RNN (Recurrent Neural Network). Compared with LSTM (Long Short-Term Memory), GRU combines the input gate and forget gate of LSTM into an update gate and simplifies the storage unit to a hidden state, thereby reducing the number of parameters and computational complexity.

[0075] The intelligent perception model constructed using a GRU network is mainly composed of an update gate and a reset gate. Among them, the update gate determines how much information from the previous hidden state needs to be retained to the current moment, and the reset gate determines how much information from the previous hidden state needs to be forgotten.

[0076] Preferably, as Figure 2 shown, train the intelligent perception model with the preprocessed ore dressing data to obtain a trained intelligent perception model, including:

[0077] Step S121: Obtain the hidden state of the previous moment, and use the training set of the current moment and the hidden state of the previous moment as the training set of the current moment for the intelligent perception model;

[0078] Step S122: The reset gate performs a linear transformation on the training set of the current moment and the hidden state of the previous moment through a Sigmoid activation function to obtain a reset gate output;

[0079] Step S123: The update gate performs a linear transformation on the training set of the current moment and the hidden state of the previous moment through a Sigmoid activation function to obtain an update gate output;

[0080] Step S124: Calculate a candidate hidden state based on the training set of the current moment, the reset gate output, the hidden state of the previous moment, and a weight matrix;

[0081] Step S125: Calculate the final hidden state at the current moment based on the update gate output, the candidate hidden state, and the hidden state at the previous moment, and perform inverse normalization on the final hidden state at the current moment to obtain the predicted value of the training set;

[0082] Step S126: Obtain the true value of the training set, and use the loss function to calculate the loss value of the training set according to the true value and the predicted value of the training set;

[0083] Step S127: Calculate the gradient of the loss function with respect to each parameter of the intelligent perception model according to the loss value of the training set;

[0084] Step S128: Adjust the parameters of the intelligent perception model according to the gradient of each parameter of the intelligent perception model to obtain the trained intelligent perception model.

[0085] Understandably, the input of the intelligent perception model is the training set at the current moment and the hidden state at the previous moment. First, it passes through the reset gate, and the current moment's training set and the previous moment's hidden state are linearly transformed through the Sigmoid activation function to obtain the reset gate output (a value between 0 and 1, the closer the value is to 1, the more information is retained). The specific calculation formula is:

[0086] r t =σ(W r ·[h t-1 ,x t +b r )

[0087] In the formula, r t is the reset gate output, σ is the Sigmoid activation function, W r is the weight matrix of the reset gate, h t-1 is the hidden state at the previous moment, x t is the training set at the previous moment, and b r is the bias term of the reset gate.

[0088] Then it passes through the update gate, and the current moment's training set and the previous moment's hidden state are linearly transformed through the Sigmoid activation function to obtain the update gate output (a value between 0 and 1, the closer the value is to 0, the more information is forgotten). The specific calculation formula is:

[0089] z t =σ(W z ·[h t-1 ,x t +b z )

[0090] In the formula, z t is the update gate output, Wr To update the weight matrix of the gate, b r is the bias term for updating the gate.

[0091] Next, based on the training set at the current time, the output of the reset gate, the hidden state at the previous time, and the weight matrix, the candidate hidden state is calculated:

[0092]

[0093] In the formula, is the candidate hidden state, tanh is the hyperbolic tangent activation function, W is the weight matrix of the candidate hidden state, and b is the bias term of the candidate hidden state.

[0094] Furthermore, by combining the output of the update gate, the candidate hidden state, and the hidden state at the previous time, the final hidden state at the current time is calculated:

[0095]

[0096] In the formula, h t is the final hidden state at the current time, which contains all relevant information from the initial time to the current time and is used for subsequent tasks.

[0097] Finally, since the output of the model is the standardized result generated by the model and the standardized input data, the final hidden state at the current time needs to be de-normalized to obtain the predicted value of the training set.

[0098] Then, after obtaining the predicted value of the training set, it is necessary to use the loss function to quantify the difference between the predicted value and the true value to guide the optimization direction of the model. Specifically, the true value of the training set is obtained, and the loss value of the training set (such as cross-entropy loss and mean squared error loss) is calculated based on the true value and the predicted value of the training set. Then, through the chain rule, the gradient (derivative) of the loss function with respect to each parameter of the intelligent perception model is calculated to indicate the direction of parameter adjustment. Once the backpropagation is completed and the gradient of each parameter is obtained, the gradient descent algorithm can be used to update the parameters of the intelligent perception model.

[0099] Through the above process, the intelligent perception model is continuously trained to obtain the trained intelligent perception model.

[0100] It should be noted that in an alternative embodiment, after obtaining the trained intelligent perception model, the performance of the model can also be evaluated regularly using the validation set: input the validation set into the trained intelligent perception model to obtain the predicted values of the validation set; obtain the true values of the validation set, and calculate the performance metrics of the trained intelligent perception model based on the predicted values and true values of the test set, such as Root Mean Squared Error (RMSE) and Mean Relative Error (MRE). These metrics measure the differences and biases between the predicted values and the actual values of the model. The specific calculation formulas are as follows:

[0101]

[0102] In the formula, RMSE is the root mean squared error, MRE is the mean relative error, y i is the true value of the i-th sample in the validation set, and pred i is the predicted value of the i-th sample in the validation set, and N is the number of samples in the validation set.

[0103] Evaluate the performance of the trained intelligent perception model based on the root mean squared error and mean relative error to obtain the evaluation result. If the RMSE and MRE are low, it indicates that the prediction accuracy of the model is high and the performance is good; on the contrary, if these metrics are high, it indicates that the prediction accuracy of the model is low and further subsequent hyperparameter tuning processes are required. The above steps use a GRU network to construct an intelligent perception model that can capture the dynamic characteristics of time series data. Through continuous iterative training of the above steps, the model gradually learns the complex relationship between the input data and the feed rate of the thickener, thereby improving the performance and accuracy of the model.

[0104] Step S130: Use the particle swarm optimization algorithm and optimize the hyperparameters of the trained intelligent perception model through the validation set to obtain an optimized intelligent perception model.

[0105] In this embodiment, the particle swarm optimization (PSO) algorithm is used for hyperparameter optimization. The PSO algorithm is an optimization algorithm based on swarm intelligence that simulates the group behavior of bird flocks or fish schools and searches for the optimal solution through information sharing among individuals. It has the advantages of strong global search ability, simple implementation, and fast convergence speed, and is widely used in fields such as engineering optimization and machine learning.

[0106] Preferably, as Figure 3 shown, use the particle swarm optimization algorithm and optimize the hyperparameters of the trained intelligent perception model through the validation set to obtain an optimized intelligent perception model, including:

[0107] Step S1: Randomly generate a group of particles, where each particle includes a position vector and a velocity vector;

[0108] Step S2: Construct a corresponding intelligent perception model for the position vector of each particle, and calculate the loss of each particle's intelligent perception model on the validation set to obtain the current fitness value of each particle;

[0109] Step S3: Obtain the historical best fitness value of each particle, and update the individual optimal position of each particle by comparing the current fitness value of each particle with its historical best fitness value;

[0110] Step S4: Compare the current fitness values of all the particles, determine the global optimal fitness value and its corresponding particle position, and update the global optimal position of the group;

[0111] Step S5: Adjust the velocity and position of each particle according to the individual optimal position, the global optimal position, and the current velocity of each particle;

[0112] Step S6: Repeat the above steps S2 - S5, and determine whether the preset number of iterations is reached. If so, converge and generate the optimized intelligent perception model.

[0113] Understandably, first, a group of particles (i.e., candidate solutions) are randomly generated. Each particle has a position vector (representing a set of hyperparameters) and a velocity vector. Then, a corresponding GRU model is constructed for the position of each particle (i.e., a set of hyperparameters), and the model performance is evaluated on the validation set. The validation set loss is used as the fitness value. Next, the current fitness value of each particle is compared with its historical best fitness value. If the current value is better, its individual optimal position is updated. Also, the current fitness values of all particles are compared to find the global optimal fitness value and its corresponding particle position, and the global optimal position of the group is updated. According to the individual optimal and global optimal positions, and the current velocity of the particle, the velocity and position of the particle are adjusted. Repeat the above steps until the preset number of iterations is reached, and then generate the optimized intelligent perception model.

[0114] The above steps can automatically search and find the optimal combination of GRU model hyperparameters through the PSO algorithm, thereby improving the prediction accuracy and generalization ability of the model. This method avoids the blindness and inefficiency of manual hyperparameter tuning and is an effective means to realize the intelligent perception of the thickener feed rate.

[0115] Step S140: Input the test set into the optimized intelligent perception model, and output the predicted value of the thickener feed rate through the optimized intelligent perception model.

[0116] Understandably, the finally optimized intelligent perception model is used to predict each sample in the test set, and the corresponding predicted value of the thickener feed rate is obtained.

[0117] Preferably, according to the foregoing model evaluation process, the optimized intelligent perception model can also be used to perform a final performance evaluation on the test set to obtain the performance of the model in actual applications, and then deploy the final model to the actual production environment to provide users with an intelligent perception model for the thickener feed rate with more accurate prediction results and better performance.

[0118] The thickener feed rate prediction method provided by the embodiments of the present application obtains original ore dressing data, preprocesses the original ore dressing data to obtain preprocessed ore dressing data, and the preprocessed ore dressing data includes a training set, a validation set, and a test set; constructs an intelligent perception model using a GRU network, and trains the intelligent perception model through the training set to obtain a trained intelligent perception model; optimizes the hyperparameters of the trained intelligent perception model using a particle swarm algorithm and through the validation set to obtain an optimized intelligent perception model; inputs the test set into the optimized intelligent perception model, and outputs the predicted value of the thickener feed rate through the optimized intelligent perception model. The present application solves the problem of online prediction of the thickener feed ore volume, provides a basis for the optimized control of the thickener, and greatly improves the accuracy and practicality of the thickener.

[0119] Embodiment 2

[0120] As Figure 4 shown, it is a schematic structural diagram of a thickener feed rate prediction device 400 in the embodiments of the present application, and the device includes:

[0121] A preprocessing module 410, configured to obtain original ore dressing data, preprocess the original ore dressing data to obtain preprocessed ore dressing data, and the preprocessed ore dressing data includes a training set, a validation set, and a test set;

[0122] A training module 420, configured to construct an intelligent perception model using a GRU network, and train the intelligent perception model through the training set to obtain a trained intelligent perception model;

[0123] An optimization module 430, configured to optimize the hyperparameters of the trained intelligent perception model using a particle swarm algorithm and through the validation set to obtain an optimized intelligent perception model;

[0124] A prediction module 440, configured to input the test set into the optimized intelligent perception model, and output the predicted value of the thickener feed rate through the optimized intelligent perception model.

[0125] The thickener feed rate prediction device provided by the embodiment of the present application can implement each process of the thickener feed rate prediction method corresponding to Embodiment 1 and achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0126] The thickener feed rate prediction device provided by the embodiment of the present application solves the problem of online prediction of the ore feed amount of the thickener, provides a basis for the optimized control of the thickener, and greatly improves the accuracy and practicability of the thickener.

[0127] Embodiment 3

[0128] The embodiment of the present application also provides a computer device. Specifically, please refer to Figure 5 , Figure 5 which is the basic structural block diagram of the computer device in this embodiment.

[0129] The computer device 5 includes a memory 51, a processor 52, and a network interface 53 that are communicatively connected to each other through a system bus. It should be noted that only the computer device 5 with the memory 51, the processor 52, and the network interface 53 is shown in the figure, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be alternatively implemented. Among them, those skilled in the art of the present technology can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0130] The computer device can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The computer device can perform human-computer interaction with the user through a keyboard, a mouse, a remote control, a touchpad, or a voice control device, etc.

[0131] The memory 51 includes at least one type of readable storage medium, which includes flash memory, hard disk, multimedia card, card-type memory (such as SD or D-slot compatibility test memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 51 may be an internal storage unit of the computer device 5, such as the hard disk or memory of the computer device 5. In other embodiments, the memory 51 may also be an external storage device of the computer device 5, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc. equipped on the computer device 5. Of course, the memory 51 may also include both the internal storage unit and the external storage device of the computer device 5. In this embodiment, the memory 51 is generally used to store the operating system and various application software installed on the computer device 5, such as computer-readable instructions of the slot compatibility test method. In addition, the memory 51 may also be used to temporarily store various data that have been output or will be output.

[0132] In some embodiments, the processor 52 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other thickener feed rate prediction chips. The processor 52 is generally used to control the overall operation of the computer device 5. In this embodiment, the processor 52 is used to run the computer-readable instructions stored in the memory 51 or process data, such as running the computer-readable instructions of the slot compatibility test method.

[0133] The network interface 53 may include a wireless network interface or a wired network interface, and the network interface 53 is generally used to establish a communication connection between the computer device 5 and other electronic devices.

[0134] The computer device provided in this embodiment can execute the above-mentioned thickener feed rate prediction method. Here, the thickener feed rate prediction method may be the thickener feed rate prediction method of the above various embodiments.

[0135] Embodiment 4

[0136] This embodiment also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the thickener feed rate prediction method in the embodiment are implemented.

[0137] In this embodiment, the computer-readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the computer-readable storage medium may be an internal storage unit of a computer device, such as the hard disk or memory of the computer device. In other embodiments, the computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk equipped on the computer device, a Smart Media Card (SMC for short), a Secure Digital (SD for short) card, a Flash Card, etc. Of course, the computer-readable storage medium may also include both the internal storage unit and the external storage device of the computer device. In this embodiment, the computer-readable storage medium is generally used to store the operating system and various application software installed on the computer device. In addition, the computer-readable storage medium can also be used to temporarily store various data that have been output or will be output.

[0138] In several embodiments provided by the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely illustrative. For example, the flowcharts and structure diagrams in the accompanying drawings show the possible architectures, functions, and operations of devices, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in an alternative implementation, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the structure diagram and / or flowchart, and the combination of blocks in the structure diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0139] In addition, in each embodiment of the present invention, the various functional modules or units may be integrated together to form an independent part, or each module may exist alone, or two or more modules may be integrated to form an independent part.

[0140] When the above-described functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a smart phone, a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium can be a non-volatile storage medium or a volatile storage medium. For example, the storage medium can be: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, etc., which can store program codes of various kinds.

[0141] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, and all of them should be covered by the protection scope of the present invention.

Claims

1. A method for predicting thickener feed amount, characterized in that: The method comprises: Acquire original mineral processing data, preprocess the original mineral processing data to obtain preprocessed mineral processing data, wherein the preprocessed mineral processing data includes a training set, a verification set and a test set; Using the GRU network to build an intelligent perception model, and training the intelligent perception model through the training set to obtain a trained intelligent perception model; Using a particle swarm algorithm and the validation set to optimize the hyperparameters of the trained intelligent perception model, to obtain an optimized intelligent perception model; The test set is input into the optimized intelligent perception model, and the predicted value of the thickener feed amount is output through the optimized intelligent perception model.

2. The method for predicting thickener feed amount according to claim 1, characterized in that: The original mineral processing data includes at least one of the feed rate of the semi-autogenous mill, the frequency of the grinding slurry pump, the current of the grinding slurry pump, the flow rate of the grinding cyclone, the grade of the flotation ore, the grade of the flotation concentrate and the grade of the flotation tailings.

3. The method for predicting thickener feed amount according to claim 1, characterized in that: The preprocessing of the original mineral processing data to obtain the preprocessed mineral processing data includes: Identifying missing values ​​in the original mineral processing data, and forward filling the missing values ​​to obtain filled mineral processing data; Detecting abnormal data in the filled mineral processing data, and removing the abnormal data to obtain mineral processing data after removal; The eliminated mineral processing data is subjected to standardization processing to obtain the pre-processed mineral processing data.

4. The method for predicting thickener feed amount according to claim 1, characterized in that: The intelligent perception model includes an update gate and a reset gate, and the intelligent perception model is trained by the training set to obtain a trained intelligent perception model, including: Obtaining the hidden state at the previous moment, and using the training set at the current moment and the hidden state at the previous moment as the training set of the intelligent perception model at the current moment; The reset gate performs a linear transformation on the training set at the current moment and the hidden state at the previous moment through a Sigmoid activation function to obtain a reset gate output; The update gate performs a linear transformation on the training set at the current moment and the hidden state at the previous moment through a Sigmoid activation function to obtain an update gate output; Calculate candidate hidden states based on the training set at the current moment, the reset gate output, the hidden state at the previous moment, and the weight matrix; Based on the update gate output, the candidate hidden state and the hidden state at the previous moment, the final hidden state at the current moment is calculated, and the final hidden state at the current moment is denormalized to obtain the predicted value of the training set; Obtaining the true value of the training set, and calculating the loss value of the training set according to the true value and the predicted value of the training set using a loss function; Calculate the gradient of the loss function to each parameter of the intelligent perception model according to the loss value of the training set; The parameters of the intelligent perception model are adjusted according to the gradient of each parameter of the intelligent perception model to obtain the trained intelligent perception model.

5. The method for predicting thickener feed amount according to claim 4, characterized in that: After obtaining the trained intelligent perception model, the method further includes: Inputting the verification set into the trained intelligent perception model to obtain a predicted value of the verification set; Obtaining the true value of the validation set, and calculating the root mean square error and mean relative error of the trained intelligent perception model according to the predicted value and the true value of the test set; The performance of the trained intelligent perception model is evaluated according to the root mean square error and the mean relative error to obtain an evaluation result.

6. The method for predicting thickener feed amount according to claim 5, characterized in that: The step of calculating the root mean square error and the mean relative error of the trained intelligent perception model according to the predicted value and the true value of the test set includes: According to the predicted value and the true value of the test set, the formula for calculating the root mean square error of the trained intelligent perception model is: According to the predicted value and the true value of the test set, the formula for calculating the average relative error of the trained intelligent perception model is: Where RMSE is the root mean square error, MRE is the mean relative error, and y i is the true value of the i-th sample in the validation set, pred i is the predicted value of the i-th sample in the validation set, and N is the number of samples in the validation set.

7. The method for predicting thickener feed amount according to claim 1, characterized in that: The method adopts the particle swarm algorithm and optimizes the hyper parameters of the trained intelligent perception model through the validation set to obtain the optimized intelligent perception model, including: S1: Randomly generate a group of particles, each particle includes a position vector and a velocity vector; S2: constructing a corresponding intelligent perception model for the position vector of each particle, and calculating the loss of the intelligent perception model of each particle on the validation set to obtain the current fitness value of each particle; S3: obtaining the historical best fitness value of each particle, and updating the individual optimal position of each particle by comparing the current fitness value of each particle with its historical best fitness value; S4: comparing the current fitness values ​​of all the particles, determining the global optimal fitness value and its corresponding particle position, and updating the optimal position of the group; S5: adjusting the speed and position of each particle according to the individual optimal position, the group optimal position and the current speed of each particle; S6: Repeat S2 to S5 to determine whether the preset number of iterations is reached. If so, converge to generate the optimized intelligent perception model.

8. A thickener feed amount prediction device, characterized in that: The device comprises: A preprocessing module is used to obtain original mineral processing data, preprocess the original mineral processing data, and obtain preprocessed mineral processing data, wherein the preprocessed mineral processing data includes a training set, a verification set, and a test set; A training module, used to construct an intelligent perception model using a GRU network, and train the intelligent perception model using the training set to obtain a trained intelligent perception model; An optimization module, used for optimizing the hyperparameters of the trained intelligent perception model by using a particle swarm algorithm and the verification set to obtain an optimized intelligent perception model; A prediction module is used to input the test set into the optimized intelligent perception model, and output a predicted value of the thickener feed amount through the optimized intelligent perception model.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the thickener feed amount prediction method according to any one of claims 1 to 7 when executing the computer program.

10. 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 steps of the thickener feed amount prediction method according to any one of claims 1 to 7 are implemented.

Images