An intelligent real-time measurement and control method for the uniformity of three-dimensional temperature field distribution in the anode furnace refining process

Abstract

The present invention discloses an intelligent real-time measurement and control method for the uniformity of the three-dimensional temperature field distribution in the furnace of an anode furnace refining process, and belongs to the field of metallurgical technology. The present invention converts the collected high-temperature radiation energy signal in the three-dimensional furnace into a two-dimensional temperature image, and realizes the reconstruction of the three-dimensional combustion temperature field in the anode furnace refining process by establishing a multi-view generative adversarial network model; and calculates the temperature mixing deviation of the three-dimensional combustion temperature field, and finally inputs the temperature mixing deviation time series data set of the three-dimensional combustion temperature field and the temperature mixing deviation spatial distribution data set of the three-dimensional temperature field of different height sections into the random forest intelligent prediction model to obtain the time distribution law of the three-dimensional temperature field and the spatial distribution law of the three-dimensional temperature field. The method of the present invention evaluates the uniformity of the three-dimensional temperature field distribution in the anode furnace refining process, and accurately obtains the temporal and spatial distribution characteristics of the temperature field in the combustion process.

Description

Technical Field

[0001] The invention belongs to the technical field of metallurgy, and in particular relates to an intelligent real-time measurement and control method for the uniformity of the three-dimensional temperature field distribution in a furnace during an anode furnace refining process. Background Art

[0002] The molten pool smelting process is the mainstream non-ferrous pyrometallurgical smelting technology today, with the advantages of high smelting intensity, no smoke and high degree of self-heating smelting. As the core process of pyrometallurgy, it is mainly suitable for the smelting process of non-ferrous sulfide ores or oxide ores such as copper. Among them, the uniformity of the temperature field in the anode furnace refining combustion process is related to the reaction degree and speed of the refining process. The research on the uniformity control of the temperature field in the combustion process is conducive to reducing energy consumption and carbon emissions in the refining process. Based on the above, by developing new technologies such as the uniformity control of the temperature field in the anode furnace combustion process, the energy utilization efficiency of the non-ferrous metal smelting process can be improved, carbon emissions can be reduced, material resources and energy resources can be saved, and carbon dioxide and other emissions can be reduced. The research and development of energy-saving and carbon reduction technologies for non-ferrous metal molten pool smelting have important theoretical significance and application value.

[0003] The existing methods for detecting the uniformity of the temperature field mainly include: mean square deviation function method, coefficient of variation method, etc., but these methods have technical defects such as insufficient accuracy, low evaluation range and insufficient reliability. Summary of the invention

[0004] In view of the above-mentioned shortcomings of the prior art, the present invention provides an intelligent real-time measurement and control method for the uniformity of the three-dimensional temperature field distribution in the anode furnace refining process.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] An intelligent real-time measurement and control method for the uniformity of the three-dimensional temperature field distribution in the anode furnace refining process comprises the following steps:

[0007] S1, using a CCD camera as a two-dimensional radiation energy distribution sensor to receive high-temperature radiation energy signals in the furnace during the anode furnace refining process in real time;

[0008] S2. According to the radiation imaging model, the high-temperature radiation energy signal is converted into the temperature value of each point in the two-dimensional space to obtain a two-dimensional temperature image, and a multi-view generative adversarial network model is used to realize the reconstruction of the three-dimensional combustion temperature field of the anode furnace refining process;

[0009] S3. According to the distribution image of the three-dimensional combustion temperature field, the temperature mixing deviation inside the furnace at different times is calculated, and the time series data set composed of the temperature mixing deviation inside the furnace at different times is input as the prediction feature vector into the random forest intelligent prediction model to obtain the time distribution law of the uniformity of the three-dimensional temperature field in the anode furnace refining process;

[0010] S4. According to the distribution image of the three-dimensional combustion temperature field, the temperature data sets of all points on the horizontal cross-sections at different heights of the furnace are extracted respectively, and the temperature mixing deviation of the three-dimensional temperature field of the cross-sections at different heights during the anode furnace refining process is calculated. The spatial distribution data set composed of the temperature mixing deviation of the three-dimensional temperature field of the cross-sections at different heights is input into the random forest intelligent prediction model to obtain the spatial distribution law of the uniformity of the three-dimensional temperature field during the anode furnace refining process.

[0011] The three-dimensional temperature field distribution uniformity of the anode furnace refining process of the present invention includes the uniformity change of the temperature in the furnace with time during the anode furnace refining process and the uniformity change of the temperature in the furnace with position and space during the anode furnace refining process.

[0012] As a preferred embodiment of the present invention, the detection cameras of the CCD camera are arranged in layers and staggered on both sides of the furnace, so as to receive the high-temperature radiation energy signal of the flame inside the furnace in real time and in all directions.

[0013] More preferably, 12-16 detection cameras are installed on both sides of the furnace in 3-4 layers, with four cameras on each layer, arranged symmetrically, and the field of view of the detection cameras of the CCD cameras is 45 degrees downward to look down at the furnace space. Among them, since the furnace temperature is very high, the camera cannot be directly installed in the furnace to capture the flame image. For this reason, an optical periscope is selected to be inserted into the furnace, and the high-temperature resistant lens at the front end of the periscope is used to image the furnace flame and transmit it to the camera target surface through the periscope.

[0014] As a preferred embodiment of the present invention, the high temperature radiation energy signal is converted into a two-dimensional temperature image of the temperature distribution of each pixel point according to a radiation imaging model, characterized in that the radiation imaging model is:

[0015] (1);

[0016] in, represents the temperature distribution in the two-dimensional temperature image, E represents the high temperature radiation energy signal collected by the CCD camera, The coefficient matrix represents the correlation between the temperature distribution in the two-dimensional temperature image and the high temperature radiation energy signal.

[0017] As a preferred embodiment of the present invention, a multi-view generative adversarial network model (GAN) is used to realize a method for reconstructing the three-dimensional combustion temperature field of the anode furnace refining process:

[0018] The multi-view generative adversarial network model (GAN) can convert the radiation images collected by CCD industrial cameras at different angles in the anode furnace into two-dimensional temperature images according to the radiation imaging model, thereby obtaining the temperature distribution data in the anode furnace as the true value for model training. Finally, the generator creates a three-dimensional model or voxel grid, which is then transmitted to the discriminator to evaluate the authenticity of the model and feedback the improvement signal to the generator. The two continuously optimize the generator and discriminator through adversarial training to generate a more accurate three-dimensional combustion temperature field.

[0019] The present invention introduces temperature mixing deviation as a measurement index of the uniformity of the three-dimensional temperature field in the anode furnace refining process, which can more accurately reflect the uniformity of the temperature distribution in the anode furnace refining process, and the size of the temperature mixing deviation value reflects the uniformity of the temperature distribution in the anode furnace refining process: the smaller the temperature mixing deviation value, the more uniform the temperature distribution in the furnace during the anode furnace refining process, and the better the refining effect; the larger the temperature mixing deviation value, the greater the difference in temperature distribution in the furnace during the anode furnace refining process, resulting in uneven heating of the furnace, reduced refining efficiency, and shortened furnace life.

[0020] As a preferred embodiment of the present invention, the temperature mixing deviation calculation formula inside the furnace during the anode furnace refining process is:

[0021] (2);

[0022] in, represents the temperature mixing deviation, N represents all the temperature grid points inside the furnace at different times or all the temperature grid points on the horizontal cross-section of different furnace heights, n (n=1,2,3,4,…,N) is the nth temperature grid point in the reconstructed three-dimensional combustion temperature field of the anode furnace, It indicates the actual temperature of the nth temperature grid point inside the furnace or the furnace cross section during the anode furnace refining process. Represents the expected optimum combustion temperature set by the factory based on raw data and experience.

[0023] The invention realizes accurate measurement of the uniformity of the three-dimensional temperature field in the anode furnace refining process by calculating the temperature mixing deviation inside the furnace or on the furnace cross section during the actual anode furnace combustion process.

[0024] As a preferred embodiment of the present invention, in order to obtain the temporal and spatial variation law of the uniformity of the furnace temperature field, based on the reconstructed three-dimensional combustion temperature field inside the furnace, the temperature mixing deviation of the three-dimensional temperature field of the anode furnace at any time is calculated as a uniformity measurement index, thereby obtaining the law of the change of the three-dimensional temperature field of the anode furnace over time during the entire refining process. At the same time, based on the reconstructed three-dimensional temperature field inside the furnace, the temperature mixing deviation on the horizontal cross section of the furnace at any height can be calculated as a uniformity measurement index of the cross section, thereby obtaining the variation law of the temperature field uniformity at different heights of the entire anode furnace.

[0025] As a preferred embodiment of the present invention, the random forest intelligent prediction model is

[0026] (3);

[0027] in, Indicates the random forest intelligent prediction model for the first Sample data The predicted value of =1,2,3,4,5,……), M represents the number of decision trees in the random forest, is the mth decision tree for the input sample The predicted value of each decision tree The predicted value is calculated by the decision rule at the node of the tree. It is a time series dataset or a spatial distribution dataset.

[0028] As a preferred implementation scheme of the present invention, in step S3, the time series data set is divided into a training set and a test set, the training set is input into the random forest intelligent prediction model for training, and then the trained random forest intelligent prediction model is used to perform predictive analysis on the test set to obtain the temporal variation characteristics of the temperature uniformity in the furnace, that is, the temporal distribution law of the three-dimensional temperature field distribution in the furnace.

[0029] As a preferred implementation scheme of the present invention, in step S4, the spatial distribution data is divided into a training set and a test set, the training set is input into the random forest intelligent prediction model for training, and then the trained random forest intelligent prediction model is used to perform predictive analysis on the test set to obtain the variation characteristics of the temperature uniformity in the furnace with the height, that is, the spatial distribution law of the three-dimensional temperature field in the furnace.

[0030] The time series data set includes the temperature mixing deviation inside the furnace at each moment; the spatial distribution data set includes the temperature mixing deviation of the three-dimensional temperature field of different height sections in the furnace.

[0031] The present invention also claims protection for an intelligent real-time measurement and control system for the uniformity of the three-dimensional temperature field distribution in the furnace of the anode furnace refining process, comprising a communication-connected storage and an actuator, wherein the storage stores a program for realizing an intelligent real-time measurement and control method for the uniformity of the three-dimensional temperature field distribution in the furnace of the anode furnace refining process when the actuator executes the program.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows: on the one hand, the present invention adopts the reconstruction algorithm of the three-dimensional temperature field of the anode furnace refining process to monitor the furnace temperature information in real time, which solves the problem of non-visualization of the high-temperature environment and reduces the cost of manual operation. On the other hand, the present invention adopts the random forest intelligent prediction model to carry out real-time monitoring and analysis of the spatiotemporal series of temperature mixing deviations, and more accurately obtains the spatiotemporal distribution law of the furnace temperature field, which has better interpretability and robustness. In addition, the method of the present invention can be used not only to study the time series and spatial series of temperature fields, but also to study the time series of concentration fields; the method can be used not only to study the uniformity of temperature fields, but also to evaluate the uniformity of gas-liquid mixing. Therefore, the method of the present invention is non-contact, more intuitive, more reliable, and has strong applicability. In addition to the metallurgical industry, it can also be applied to many fields such as chemical industry and pharmaceutical industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 The present invention provides a flow chart of an intelligent real-time measurement and control method for the uniformity of the three-dimensional temperature field distribution in the anode furnace refining process.

[0034] Figure 2 The flowchart of the reconstruction method of three-dimensional combustion temperature field distribution image is shown in FIG.

[0035] Figure 3 Generate adversarial network model architecture for multi-view.

[0036] Figure 4 This is a framework diagram for analyzing the random forest intelligent prediction model of the present invention.

[0037] Figure 5 Comparison of the prediction results of the random forest intelligent prediction model of the intelligent real-time measurement and control method for the uniformity of the three-dimensional temperature field distribution in the anode furnace refining process described in the present invention and other models; (a) is the prediction result of the random forest intelligent prediction model of the present invention; (b) is the prediction result of the RBF intelligent prediction model; (c) is the prediction result of the SVM intelligent prediction model; (d) is the prediction result of the BP neural network intelligent prediction model. DETAILED DESCRIPTION

[0038] In order to better illustrate the purpose, technical solutions and advantages of the present invention, the present invention will be further described below in conjunction with specific embodiments.

[0039] Example 1

[0040] like Figure 1 As shown, an intelligent real-time measurement and control method for the uniformity of the three-dimensional temperature field distribution in the anode furnace refining process includes the following steps:

[0041] S1: 16 detection cameras are installed on both sides of the furnace in 4 layers, with 4 cameras on each layer, arranged symmetrically, and their field of view angles are all 45° downward to look down at the furnace space. Since the furnace temperature is very high, the camera cannot be directly installed in the furnace to capture the flame image. For this reason, a specially designed optical periscope is inserted into the furnace. The high-temperature resistant lens at the front end images the furnace flame and transmits it to the camera target surface through the periscope. Therefore, the CCD camera is used as a two-dimensional radiation energy distribution sensor to receive the high-temperature radiation energy signal in the furnace during the anode furnace refining process in real time.

[0042] S2: Reconstruct the three-dimensional combustion temperature field distribution image.

[0043] S2-1: Figure 2 As shown, according to the radiation imaging model, the high temperature radiation energy signal is converted into the temperature distribution of each pixel in the two-dimensional temperature image to form a two-dimensional temperature image. The principle of the radiation imaging model is:

[0044] (1);

[0045] in, represents the temperature distribution in the two-dimensional temperature image, E represents the high temperature radiation energy signal collected by the CCD camera, The coefficient matrix represents the temperature distribution in the two-dimensional temperature image and the high temperature radiation energy signal.

[0046] S2-2: Using a multi-view generative adversarial network model (GAN) to reconstruct the three-dimensional combustion temperature field of the anode furnace refining process:

[0047] like Figure 3 As shown in the figure, first, the two-dimensional temperature matrix at different angles in the furnace is extracted from the two-dimensional temperature image to obtain the temperature distribution data inside the anode furnace; then the temperature distribution data set inside the anode furnace is used as the training truth value of the multi-view generative adversarial network model (GAN), and the obtained multi-angle two-dimensional temperature image is preprocessed by standardization, registration and alignment; finally, a generative network is designed to accept multi-view two-dimensional images as input. The goal of the generative network is to generate a three-dimensional voxel grid, and the generator and discriminator are trained using an adversarial loss function. The generator learns to generate real temperature data, while the discriminator learns to distinguish between generated data and real data. The model ensures that the generator can effectively fuse information from different perspectives through multi-view fusion to generate a more accurate and consistent three-dimensional combustion temperature field.

[0048] S3. The temperature mixing deviation of the three-dimensional combustion temperature field is used as an indicator to quantify the uniformity of the three-dimensional temperature field distribution in the furnace during the anode furnace refining process and to analyze the spatiotemporal characteristics of the temperature field; the temperature mixing deviation of the three-dimensional combustion temperature field includes the temperature mixing deviation inside the furnace at each moment and the temperature mixing deviation of the three-dimensional temperature field in sections at different heights in the furnace.

[0049] S3-1: Based on the three-dimensional temperature field of the anode furnace reconstructed by the multi-view generative adversarial network (GAN) model, the temperature mixing deviation inside the furnace at a certain moment in the three-dimensional temperature field can be calculated:

[0050] (2);

[0051] in, represents the temperature mixing deviation, N represents all the temperature grid points in the reconstructed three-dimensional combustion temperature field of the anode furnace or the temperature field section (i.e. all the temperature points in the temperature field space), n (1, 2, 3, 4, …, N) represents the nth temperature grid point in the reconstructed three-dimensional combustion temperature field of the anode furnace, It indicates the actual temperature of the nth temperature grid point inside the furnace or the furnace cross section during the anode furnace refining process. It means that according to the original data and the expected optimal combustion temperature set by the experience factory, the temperature mixing deviation inside the furnace or on the furnace cross section during the actual anode furnace combustion process is calculated to achieve an accurate measurement of the uniformity of the three-dimensional temperature field in the anode furnace refining process.

[0052] S3-2: Figure 4 As shown, the temperature mixing deviation of the three-dimensional temperature field of the anode furnace at a certain moment calculated in steps S1 and S2 is used as a uniformity data, and then the temperature distribution uniformity time series data set of the anode furnace refining process is collected according to the refining time. The time series data set is used as a prediction feature vector, and the training set and the test set are divided. The training set is input into the random forest intelligent prediction model for training, and then the trained random forest intelligent prediction model is used to predict and analyze the test set to obtain the time-varying characteristics of the temperature uniformity in the furnace, that is, the time distribution law of the three-dimensional temperature field distribution in the furnace.

[0053] S3-3: If Figure 4As shown in the figure, in order to obtain the spatial distribution uniformity of the furnace temperature field, the three-dimensional combustion temperature field of the anode furnace reconstructed by the multi-view generative adversarial network (GAN) model is used to collect the temperature values ​​of all points on the horizontal section at any height of the furnace at a certain moment and calculate its temperature mixing deviation as the temperature distribution uniformity data of the horizontal section of the furnace. The actual height of the anode furnace is taken as the standard to calculate the uniformity data set under the spatial sequence of the anode furnace refining process. The spatial distribution uniformity data is divided into a training set and a test set. The training set is input into the random forest intelligent prediction model for training, and then the trained random forest intelligent prediction model is used to predict and analyze the test set to obtain the characteristics of the change of temperature uniformity in the furnace with height, that is, the spatial distribution law of the three-dimensional temperature field in the furnace.

[0054] The random forest intelligent prediction model is an algorithm based on ensemble learning, which performs regression tasks by constructing multiple decision trees and integrating the prediction results of the decision trees. In a random forest, each decision tree is trained independently on a randomly selected subsample, and the prediction results of multiple decision trees are averaged or weighted averaged to obtain the final regression result.

[0055] The random forest intelligent prediction model is

[0056] (3);

[0057] in, Indicates the random forest intelligent prediction model for the first Sample data The predicted value of =1,2,3,4,5,……), M represents the number of decision trees in the random forest, is the mth decision tree for the input sample The predicted value of each decision tree The predicted value is calculated by the decision rule at the node of the tree. It is a time series dataset or a spatial distribution dataset.

[0058] In order to verify the accuracy and robustness of the model prediction, the present invention uses the random forest intelligent prediction model, radial basis neural network (RBF) intelligent prediction model, BP neural network intelligent prediction model and support vector machine (SVM) intelligent prediction model to predict the top blowing furnace slag temperature data time series of Jinchuan Group. The prediction results of the random forest intelligent prediction model are compared with those of other models. Figure 5As shown in the figure, the random forest intelligent prediction model has the smallest prediction error, with the mean square error and root mean square error being only 0.12 and 0.23 respectively, and the prediction accuracy being 99.8%, which is 0.4%, 1.2% and 1.6% higher than the prediction accuracy of the RBF intelligent prediction model, the BP neural network intelligent prediction model and the SVM intelligent prediction model respectively.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the essence and scope of the technical solution of the present invention.

Claims

1. An intelligent real-time measurement and control method for the uniformity of the three-dimensional temperature field distribution in the anode furnace refining process, characterized in that: The steps include: S1, using a CCD camera as a two-dimensional radiation energy distribution sensor to receive high-temperature radiation energy signals in the furnace during the anode furnace refining process in real time; S2. According to the radiation imaging model, the high-temperature radiation energy signal is converted into the temperature value of each point in the two-dimensional space to obtain a two-dimensional temperature image, and a multi-view generative adversarial network model is used to realize the reconstruction of the three-dimensional temperature field of the anode furnace refining process; The radiation imaging model is: （1）； in, represents the temperature distribution in the two-dimensional temperature image, E represents the high temperature radiation energy signal collected by the CCD camera, The coefficient matrix represents the correlation between the temperature distribution in the two-dimensional temperature image and the high temperature radiation energy signal; S3. According to the distribution image of the three-dimensional temperature field, the temperature mixing deviation inside the furnace at different times is calculated, and the time series data set composed of the temperature mixing deviation inside the furnace at different times is input as the prediction feature vector into the random forest intelligent prediction model to obtain the time distribution law of the uniformity of the three-dimensional temperature field in the anode furnace refining process; S4. According to the distribution image of the three-dimensional temperature field, the temperature data sets of all points on the horizontal cross-sections at different heights of the furnace are extracted respectively, and the temperature mixed deviation of the three-dimensional temperature field of the cross-sections at different heights during the anode furnace refining process is calculated. The spatial distribution data set composed of the temperature mixed deviation of the three-dimensional temperature field of the cross-sections at different heights is input into the random forest intelligent prediction model to obtain the spatial distribution law of the uniformity of the three-dimensional temperature field during the anode furnace refining process; The temperature mixing deviation calculation formula is: （2）； in, represents the temperature mixing deviation, N represents all the temperature grid points inside the furnace at different times or all the temperature grid points on the horizontal cross-section of different furnace heights, n is the nth temperature grid point in the reconstructed three-dimensional temperature field of the anode furnace, The actual temperature of the nth temperature grid point inside the furnace or the furnace cross section during the anode furnace refining process, Indicates the expected optimal combustion temperature set; The mathematical expression of the random forest intelligent prediction model is: （3）； in, Indicates the random forest intelligent prediction model for the input temperature mixed deviation dataset time series. Sample data The predicted value of =1,2,3,4,5,…, M represents the number of decision trees in the random forest, is the mth decision tree for the input sample The predicted value of It is a time series dataset or a spatial distribution dataset.

2. The intelligent real-time measurement and control method for the uniformity of the three-dimensional temperature field distribution in the anode furnace refining process according to claim 1 is characterized in that: In the step S1, the detection cameras of the CCD camera are arranged in layers and staggered on both sides of the anode furnace to collect high-temperature radiation energy signals in real time.

3. The intelligent real-time measurement and control method for the uniformity of the three-dimensional temperature field distribution in the anode furnace refining process according to claim 2 is characterized in that: In step S1, 12-16 detection cameras are installed on both sides of the furnace in 3-4 layers, with four cameras on each layer, arranged symmetrically, and the field of view of the detection cameras is 45 degrees downward to overlook the furnace space.

Images