Method, system and equipment for predicting end point of oxidation-reduction period of anode furnace and medium

Abstract

The invention provides a method, a system, equipment and a medium for predicting an end point of an oxidation-reduction period of an anode furnace, and belongs to the technical field of computer vision. The method comprises the following steps: acquiring sample images on the front and back surfaces of an anode copper observation sample, and recording relative acquisition time and time from the sample to an end point; labeling the sample image according to the acquisition relative time and the time from the sample to the end point, and generating to-be-detected image data; and inputting the to-be-detected image data into a pre-trained end point judgment model, outputting the classification of the oxidation or reduction stage, the time from the oxidation end point and the time from the reduction end point by using the end point judgment model, integrating the output data, and predicting the final time from the end point. According to the method, the visual features of the anode copper sample are captured through the image recognition technology, the time from the sample to the end point is analyzed and determined in combination with time data, the end point is predicted through the pre-trained deep learning model, and accurate prediction of the oxidation-reduction period end point of the anode furnace is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of computer vision, and more specifically, relates to a method, system, device and medium for predicting the end point of the oxidation-reduction period of an anode furnace. Background Art

[0002] Copper pyrometallurgical refining furnaces are mainly divided into reverberatory furnaces and rotary anode furnaces. Currently, most copper smelting plants use rotary anode furnaces. The pyrometallurgical refining of blister copper mainly includes four operation stages: feeding, oxidation, reduction, and casting. Among them, the oxidation and reduction stages are the core processes. The oxidation and reduction periods are complex processes involving chemical reactions, heat transfer, mass transfer, and fluid flow. The main purpose of refining is to further reduce the content of impurity elements such as sulfur and oxygen elements in blister copper, and provide qualified anode copper for subsequent electrolytic refining. Anodes with poor physical specifications will not only increase the pressure on the anode shaping unit of the electrolytic refining process, increase the workload of operators, but also deteriorate the electrolyte, bringing a series of problems to the electrolytic purification system. Therefore, in addition to ensuring the chemical composition of the anode copper, it is also crucial to ensure the physical specifications of the anode plate. Among them, accurate judgment of the oxidation and reduction end points during the pyrometallurgical refining process has an important impact on improving the physical specifications of the anode plate. Therefore, improving the accuracy of predicting the oxidation and reduction end points of anode copper is of great significance for reducing costs and increasing the quality of anode copper, improving operation safety, etc.

[0003] Currently, the methods for judging the oxidation and reduction end points of anode furnaces are mainly divided into three categories: manual experience method, chemical composition analysis method, and artificial intelligence-based method. The manual experience method mainly relies on copper smelting experts based on experience to make manual judgments according to the appearance characteristics of the anode copper observation samples. This method is too dependent on the experience and skill level of on-site operators. Inevitably, judgment errors will increase production costs and disrupt the production rhythm, resulting in a significant reduction in production efficiency. The chemical composition analysis method mainly judges the oxidation and reduction end points by analyzing the content of gases such as sulfur dioxide, oxygen, and carbon monoxide in the flue gas. This method is limited by sampling conditions, the accuracy of flue gas component detection, etc., and the accuracy of judging the oxidation and reduction end points is not high, and it is relatively mechanical and cannot adapt to complex and changeable production scenarios. The artificial intelligence-based method combines on-site work experience, sampling data, etc. with modern artificial intelligence technology to realize the intelligentization of anode copper refining. Specifically, a database of copper sample surface and cross-section images is established by collecting data, and during the prediction stage, the collected images are compared with the data in the database to judge whether it is the end point. If it is not the end point, the trained GRNN prediction model is called to predict the time of the sample from the end point. Such methods combine artificial intelligence technologies and ideas such as KNN to judge whether it is the end point online, and combine the deep learning network GRNN to judge the time of non-end point samples from the end point, but the artificial intelligence algorithms and models used are relatively backward and the effects are not ideal. Summary of the Invention

[0004] In view of the above problems, the purpose of the present invention is to provide a method, system, device and medium for predicting the end point of the oxidation-reduction period of an anode furnace. By using image recognition technology to capture the visual characteristics of anode copper samples, combined with time data analysis to determine the time of the sample from the end point, and using a pre-trained deep learning model for end point prediction, accurate prediction of the end point of the oxidation-reduction period of the anode furnace is realized, and the automation and intelligent level of the smelting process are significantly improved.

[0005] To achieve the above object, the present invention is realized through the following technical solutions: In a first aspect, an embodiment of the present application provides a method for predicting the end point of the oxidation-reduction period of an anode furnace, including: Collect sample images of the front and back of the anode copper observation sample, and record the relative acquisition time and the time of the sample from the end point; According to the relative acquisition time and the time of the sample from the end point, label the sample image to generate the image data to be tested; the label records the time of the sample from the end point; Input the image data to be tested into a pre-trained end point judgment model, and use one encoder and three multi-layer perceptrons in the end point judgment model to output the classification of the oxidation or reduction stage, the time from the oxidation end point, and the time from the reduction end point, and integrate the output data to predict the final time from the end point.

[0006] In an optional embodiment, the collecting sample images of the front and back of the anode copper observation sample and recording the relative acquisition time and the time of the sample from the end point includes: Use an industrial camera to collect the visual characteristics of the front and back of the anode copper observation sample to generate a sample image; Record the furnace number, oxidation or reduction stage, time from the end point, and relative acquisition time of the sample, and store them in a database; The relative acquisition time is the time interval between the sample sampling moment and the shooting moment, and the time from the end point is the time interval between the sampling moment of the manually predicted end point sample and the sample sampling moment.

[0007] In an optional embodiment, the labeling the sample image according to the relative acquisition time and the time of the sample from the end point to generate the image data to be tested includes: If the sample image is the last sample image collected in the sampling stage, then mark the sample image as the end point sample; if the sample image is not the last sample image collected in the sampling stage, then mark the sample image as the process sample; For the end point sample, label the time of the sample from the end point as the label; For a process sample, based on the relative acquisition time and the sample's time to the end point, as well as the relative acquisition time and the sample's time to the end point of the corresponding end-point sample, calculate the reference sample's time to the end point, and determine the annotation time of the sample based on the sample's time to the end point, and use the annotation time as the label of the sample.

[0008] In an alternative embodiment, the calculating the reference sample's time to the end point based on the relative acquisition time and the sample's time to the end point, as well as the relative acquisition time and the sample's time to the end point of the corresponding end-point sample, and determining the annotation time of the sample based on the sample's time to the end point includes: In the same sampling stage, set the shooting time of the process sample as and set the shooting time of the end-point sample as Record the relative acquisition time of the end-point sample as and record the relative acquisition time of the process sample as Then the true sampling times 、 of the end-point sample and the process sample are respectively:

[0009]

[0010] Record the sample's time to the end point of the end-point sample as Then calculate the sampling moment of the true end-point sample through the following formula :

[0011] Based on the true acquisition time of the process sample and the acquisition time of the true end-point sample, calculate the reference sample's time to the end point through the following formula :

[0012] Obtain the sample's time to the end point of the process sample ; Judge Whether the difference from is less than a preset threshold; If so, then use as the annotation time of the sample; If not, then use as the annotation time of the sample.

[0013] In an alternative embodiment, the end-point judgment model includes: an encoder and three multi-layer perceptrons; The encoder uses an artificial intelligence model or a neural network as the base network; The encoder is used to read the image data to be measured, generate an intermediate tensor of 3×n, and decompose it into three output scalars, which are respectively used as the inputs of three multi-layer perceptrons. The three multi-layer perceptrons include a first multi-layer perceptron, a second multi-layer perceptron, and a third multi-layer perceptron. The first multi-layer perceptron is used to map the continuous output value of the sigmoid function to the output oxidation or reduction stage classification value c after using the sigmoid function and passing through a threshold process. The value range of c is {0, 1}, where 0 represents the oxidation stage and 1 represents the reduction stage. The second multi-layer perceptron is used to output the time t1 from the oxidation end point. The third multi-layer perceptron is used to output the time t2 from the reduction end point.

[0014] In an optional implementation manner, the end point judgment model further includes: an image enhancement module and a calculation module. The image enhancement module is used to perform enhancement processing on the image data to be measured by adopting an enhancement strategy and then input it into the encoder; the enhancement strategy includes but is not limited to image flipping, image occlusion, color contrast enhancement, and Gaussian noise processing. The calculation module is used to use the formula to calculate the final time from the end point according to the outputs of the three multi-layer perceptrons. .

[0015] In an optional implementation manner, the training process of the end point judgment model includes: continuously training the end point judgment model by using a preset training set and a loss function until the loss of the end point judgment model converges or the number of training rounds reaches a preset upper limit; respectively calculating the precision and recall of the oxidation-reduction stage classification, the precision and recall of the oxidation end point judgment, the precision and recall of the reduction end point judgment, and the MAE of the prediction results of the oxidation stage and the reduction stage, and evaluating the trained end point judgment model; optimizing and adjusting the hyperparameters of the end point judgment model according to the evaluation results; The loss function includes a classification loss , an oxidation end point regression loss and a reduction end point regression loss ; The classification loss function is defined as follows:

[0016] where N is the total number of samples, is the label value 0 or 1 of the i-th sample, is the predicted value of the end point judgment model for the i-th sample; Oxidation endpoint regression loss function Defined as:

[0017] Where N is the total number of samples, is the predicted value of the endpoint judgment model for the time from the ith sample to the oxidation endpoint, is the true value of the time from the ith sample to the oxidation endpoint, is a hyperparameter, is the loss function threshold; Restore endpoint regression loss is defined as:

[0018] Where N is the total number of samples, is the predicted value of the endpoint judgment model for the time from the i-th sample to the restoration endpoint, is the true value of the time at the end point of the i-th sample distance restoration, is a hyperparameter, is the loss function threshold.

[0019] In a second aspect, the present application also provides an anode furnace redox period endpoint prediction system, comprising: A data acquisition module is used to collect sample images of the front and back sides of the anode copper observation sample, and record the acquisition relative time and the time when the sample is away from the end point; A data processing module is used to label the sample image according to the relative acquisition time and the time the sample is away from the end point, and generate image data to be tested; the label records the time the sample is away from the end point; The endpoint prediction module is used to input the image data to be tested into the pre-trained endpoint judgment model, and use an encoder and three multi-layer perceptrons in the endpoint judgment model to output the classification of the oxidation or reduction stage, the time to the oxidation endpoint, and the time to the reduction endpoint, integrate the output data, and predict the final time to the endpoint.

[0020] In a third aspect, an embodiment of the present application further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method for predicting the end point of the redox period of an anode furnace as described in any one of the above items are implemented.

[0021] In a fourth aspect, an embodiment of the present application further provides a storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the method for predicting the end point of the redox period of an anode furnace as described in any one of the above items are implemented.

[0022] As can be seen from the above technical solutions, the present invention has the following advantages: In the method for predicting the end point of the oxidation-reduction period of the anode furnace provided by this application, through the integration of image recognition technology, precise time data analysis, and an efficient pre-trained deep learning model, the precise prediction of the end point of the oxidation-reduction period of the anode furnace is achieved. This method not only significantly improves the automation level of the smelting process, reduces manual intervention, and lowers the subjectivity of operation, but also greatly improves production efficiency and ensures the stability and controllability of the smelting process. At the same time, this method also has excellent generalization ability and robustness, and can handle complex situations under different smelting conditions, providing strong technical support for the intelligent upgrade of the anode furnace smelting process.

[0023] This application generates image data to be measured by collecting the front and back images of the anode copper observation sample and combining the collected relative time and the time of the sample from the end point for label annotation. Using the pre-trained end point judgment model, it can comprehensively consider the visual features and time information of the sample, thereby improving the prediction accuracy of the end point of the oxidation-reduction period of the anode furnace.

[0024] This application realizes the automatic prediction of the end point of the oxidation-reduction period of the anode furnace, reducing manual intervention and the subjectivity of judgment. By introducing artificial intelligence models and neural networks, it can intelligently analyze the sample images and time data and output accurate prediction results.

[0025] By accurately predicting the end point of the oxidation-reduction period, this application can timely adjust the operation parameters during the smelting process, avoid over-oxidation or over-reduction, thereby improving production efficiency and product quality. At the same time, it can also reduce energy waste and environmental pollution and lower production costs.

[0026] When training the end point judgment model in this application, a variety of loss functions and evaluation indicators are adopted to comprehensively optimize the performance of the model. In addition, by enhancing the image data to be measured through the image enhancement module, the generalization ability and robustness of the model can be further improved.

[0027] The data collection and annotation process required by this application is relatively simple and easy to implement in the actual production environment. At the same time, the structure and training process of the end point judgment model are also scalable and can be adjusted and optimized according to actual needs. Description of the Drawings

[0028] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0029] Figure 1Schematic flow chart of the method for predicting the end point of the oxidation-reduction period of the anode furnace provided by this application.

[0030] Figure 2 Schematic structural diagram of the end point judgment model provided by this application.

[0031] Figure 3 Schematic structural diagram of the system for predicting the end point of the oxidation-reduction period of the anode furnace provided by this application.

[0032] Figure 4 Schematic structural diagram of the electronic device provided by this application. Detailed implementation manners

[0033] In the following, various embodiments of the present disclosure will be more comprehensively described in the specific steps of the method for predicting the end point of the oxidation-reduction period of the anode furnace. The present disclosure may have various embodiments and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments of the present disclosure to the specific embodiments disclosed herein, but the present disclosure should be understood to cover all adjustments, equivalents, and / or alternative solutions that fall within the spirit and scope of the various embodiments of the present disclosure.

[0034] Hereinafter, the term "comprising" or "may comprise" that may be used in various embodiments of the present disclosure indicates the presence of the disclosed functions, operations, or elements, and does not limit the addition of one or more functions, operations, or elements. In addition, as used in various embodiments of the present disclosure, the terms "comprising", "having" and their cognates are only intended to indicate specific features, numbers, steps, operations, elements, components, or combinations of the foregoing items, and should not be construed as precluding the existence or addition of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing items.

[0035] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0036] Please refer to Figure 1 The following is a method flow chart of a method for predicting the end point of the oxidation-reduction period of an anode furnace in a specific embodiment. The method includes: S1: Collect the sample images of the front and back of the anode copper observation sample, and record the relative acquisition time and the time of the sample from the end point.

[0037] In the specific implementation, an industrial camera is first used to collect the visual features of the front and back of the anode copper observation sample, generating a sample image. Then, the furnace number, oxidation or reduction stage, time to the end point, and relative acquisition time of the sample are recorded and stored in the database.

[0038] Among them, the relative acquisition time is the time interval between the sampling moment of the sample and the shooting moment, and the time to the end point is the time interval between the sampling moment of the end point sample predicted manually and the sampling moment of the sample.

[0039] Exemplarily, in this step, an industrial camera is used to collect the visual features of the front and back of the anode copper observation sample. The specific operation process of one sampling is as follows: The operator places the anode copper observation sample face up in the shooting area. The sampling system detects the presence of the anode copper observation sample in the shooting area and takes a picture for sampling. At the same time, the operator inputs the relevant data of the sample according to the prompts of the sampling system, and the relevant data of the sample will be saved in the local database. Among them, the relevant data of the sample includes the furnace number, oxidation / reduction stage, time to the end point, and relative acquisition time.

[0040] After the front side is collected, the operator flips the observation sample and places it face up in the shooting area to collect the back side. After the back side is collected, the operator takes the observation sample away from the shooting area. Thus, one sampling process ends.

[0041] After sampling is completed, this step also includes the data processing process, specifically including: Manually screen and eliminate unqualified samples from the collected front and back image data of the observation sample, such as incomplete, blurred, no observation sample, images of the same observation sample collected repeatedly, etc.

[0042] Data correction. Manually check the validity and authenticity of the time to the end point data in the sampling database, correct, modify or delete invalid or incorrect data. Incorrect data such as data with a large difference between the actual time to the end point and the manually judged time to the end point. At this time, the actual time to the end point is used to correct the manually judged time to the end point.

[0043] S2: According to the relative acquisition time and the time of the sample to the end point, label the sample image with a label, generating the data of the image to be measured; the label records the time of this sample to the end point.

[0044] In this step, due to the high complexity of the anode copper operation, there may be a long interval between the operator collecting the sample and sending it for shooting, that is to say, the sampling has relatively serious hysteresis. Therefore, during the sampling process, the relative acquisition time of the sample image (that is, the time interval between the sampling moment of this sample and the current moment) and the time of this sample to the end point judged according to experience need to be labeled. The specific data labeling process is as follows: 1. If the sample image is the last sample image collected during the sampling phase, mark this sample image as the end sample; the label of this sample is the manually predicted time to the end point. Note that this label may not be 0 (i.e., the true end sample was not collected during this sampling phase), and when the label is not 0, we call this sample a near-end sample.

[0045] 2. If the sample image is not the last sample image collected during the sampling phase, mark this sample image as a process sample. For a process sample, calculate the reference time to the end point of the sample based on the relative collection time and the time to the end point of the sample, as well as the relative collection time and the time to the end point of the corresponding end sample, and determine the annotation time of this sample based on the time to the end point of the sample, and use the annotation time as the label of this sample.

[0046] Specifically, in the same sampling phase, the shooting moment of the process sample is , the shooting moment of the end sample is , the relative collection time of the end sample is recorded as , the relative collection time of the process sample is recorded as , then the true sampling times , of the end sample and the process sample are respectively:

[0047]

[0048] Record the time to the end point of the end sample as , then calculate the sampling moment of the true end sample through the following formula :

[0049] Calculate the reference time to the end point of the sample according to the true collection time of the process sample and the collection time of the true end sample through the following formula :

[0050] At this time, first obtain the time to the end point of the process sample ; and judge whether the difference between is less than the preset threshold; If is not empty and valid, and the difference between is less than the preset threshold, then use as the annotation time of this sample; otherwise, use the calculated as the annotation time of this sample.

[0051] S3: Input the image data to be measured into the pre-trained endpoint judgment model. Use one encoder and three multi-layer perceptrons in the endpoint judgment model to output the classification of the oxidation or reduction stage, the time to the oxidation endpoint, and the time to the reduction endpoint. Integrate the output data to predict the final time to the endpoint.

[0052] It should be noted specifically that in this step, referring to Figure 2 it can be known that the endpoint judgment model includes: one encoder and three multi-layer perceptrons.

[0053] Among them, the encoder is an abstract concept. Its main function is to encode the input image and output an intermediate layer tensor. Any artificial intelligence model or neural network such as ResNet, VGG, CNN, or even SVM can be used as the base network of the encoder.

[0054] The encoder is used to read the image data to be measured, generate an intermediate tensor of 3×n, and decompose it into three output scalars, which are respectively used as the inputs of the three multi-layer perceptrons. In this model, the anode copper sample image is encoded by an Encoder, and the output after encoding is a tensor of 3×n, or it can be understood as the concatenation of the outputs of 3 n×1 vectors.

[0055] In this model, the tensor output by the encoder is followed by 3 MLPs (multi-layer perceptrons) to further decompose the tensor into 3 outputs. The three multi-layer perceptrons include the first multi-layer perceptron, the second multi-layer perceptron, and the third multi-layer perceptron. Among them, the first multi-layer perceptron is responsible for classifying the period (oxidation period / reduction period) of the sample, and the output value is a classification boolean value; the second multi-layer perceptron and the third multi-layer perceptron are the same type of MLP, which are responsible for regression prediction of the time of the sample to the endpoint, and the output value is a regression value.

[0056] The first multi-layer perceptron is used to map the continuous output value of the sigmoid function to the output classification value c of the oxidation or reduction stage after threshold processing using the sigmoid function. The value range of c is {0,1}, where 0 represents the oxidation stage and 1 represents the reduction stage. For example, the output of the first multi-layer perceptron is a single scalar value, which represents the classification of the oxidation / reduction stage, and the value range is {0,1}, where 0 represents the oxidation stage and 1 represents the reduction stage. The activation function of the first multi-layer perceptron is the sigmoid function and after threshold processing, it maps the continuous output value of the sigmoid to the discrete {0,1} value range. Here, the threshold is 0.5.

[0057] The second multi-layer perceptron is used to output the time t1 to the oxidation end point; the output of the second multi-layer perceptron is a single scalar regression value, which represents the time to the oxidation end point, and the value range is R. The second multi-layer perceptron is the linear regression layer.

[0058] The third multi-layer perceptron is used to output the time t2 to the reduction end point. The output of the third multi-layer perceptron is a single scalar regression value, which represents the time to the reduction end point, and the value range is R. The third multi-layer perceptron is also the linear regression layer.

[0059] In addition, the end point judgment model further includes: an image enhancement module and a calculation module; The image enhancement module is used to perform enhanced processing on the image data to be measured by using an enhancement strategy and then input it into the encoder; the enhancement strategy includes but is not limited to image flipping, image occlusion, color contrast enhancement, and Gaussian noise processing; The calculation module is used to use the formula to calculate the final time t to the end point according to the outputs of the three multi-layer perceptrons.

[0060] As an example, the specific implementation process of step S3 is as follows: The input is an image of an anode copper observation sample , represents the encoder, the first MLP is represented by , the second and third MLPs are represented by , , and the above inference process is as follows:

[0061]

[0062]

[0063] Among them, c is the classification output, and the value is 0 or 1. 0 represents the oxidation period, and 1 represents the reduction period; , are the times to the end points of the oxidation period and the reduction period respectively. , , are three output vectors decomposed from the intermediate tensor output by the encoder and used as inputs for subsequent different MLPs. Then the final output is a vector . The period can be directly determined by the c value, and the final time t to the end point can be obtained by the following formula:

[0064] It should be specifically noted that according to different usage scenarios, the usage mode of the end-point judgment model is divided into two modes: A. Manual classification mode, which is used in the case where the period of the anode copper observation sample is determined, that is, the period of the anode copper observation sample is clearly known. At this time, we can ignore the value and the irrelevant distance-to-end time value, and directly select the corresponding distance-to-end time according to the observation sample period. For example, when we know that the observation sample is in the oxidation period, we directly select the as the predicted value of the distance to the oxidation end time. When we know that the observation sample is in the reduction period, we directly select the as the predicted value of the distance to the reduction end time.

[0065] B. Automatic classification mode, which is used in the case where the period of the anode copper observation sample is unknown or undetermined. At this time, we can apply the above formula to calculate the predicted value of the final distance to the oxidation / reduction end time, and the period can be determined by the value.

[0066] In this embodiment, by using image recognition technology to accurately capture the visual features of the anode copper observation sample and combining detailed time data analysis, including accurate records of the relative time and the distance of the sample to the end time, this method cleverly uses a pre-trained deep learning model for complex data processing and pattern recognition. This comprehensive technical means not only significantly improves the prediction accuracy of the end point of the oxidation-reduction period of the anode furnace, but also greatly promotes the automation and intelligentization process of the smelting process. Compared with the traditional manual prediction method, this method not only reduces the subjectivity and error of human judgment, but also greatly improves the production efficiency, reduces the production cost, and at the same time plays a positive role in environmental protection and energy conservation. Therefore, this method has broad application prospects and profound social and economic value in the field of anode furnace smelting.

[0067] In an embodiment of the present invention, based on step S3, the following will give a possible embodiment to non-restrictively elaborate on its specific implementation scheme.

[0068] Based on the model architecture and specific functions of the end-point judgment model, the training process of the end-point judgment model includes: First, continuously train the end-point judgment model using a preset training set and a loss function until the loss of the end-point judgment model converges or the number of training rounds reaches the preset upper limit.

[0069] Then, the trained model is evaluated, and the evaluation metrics used are Precision, Recall, and Mean Absolute Error (MAE). According to the above three metrics, the Precision and Recall of the redox stage classification, the Precision and Recall of the oxidation endpoint judgment, the Precision and Recall of the reduction endpoint judgment, and the MAE of the prediction results of the oxidation stage and the reduction stage are calculated respectively.

[0070] Finally, according to the evaluation results, the hyperparameters of the endpoint judgment model are optimized and adjusted; During the training and optimization process of the model, the loss functions adopted include classification loss , oxidation endpoint regression loss and reduction endpoint regression loss .

[0071] Classification loss : Since it is a classification of periods and the periods only include two periods, the oxidation period and the reduction period, this problem can be defined as a binary classification problem. The loss function uses the most classic Binary Cross Entropy (BCELoss) in binary classification problems. The classification loss function is defined as follows:

[0072] where N is the total number of samples, is the label value 0 or 1 of the i-th sample, is the predicted value of the endpoint judgment model for the i-th sample, that is, the probability value that the model predicts the label value of the -th sample as 1.

[0073] Oxidation endpoint regression loss : Referring to the Mean Absolute Error loss, the oxidation endpoint regression loss function is defined as:

[0074] where N is the total number of samples, is the predicted value of the endpoint judgment model for the time of the i-th sample from the oxidation endpoint, is the true value of the time of the i-th sample from the oxidation endpoint, is a hyperparameter, a very small real number to prevent the denominator of the formula from being 0, is the loss function threshold, set to 0.5.

[0075] Similarly, the definition of the reduction endpoint regression loss is:

[0076] Wherein, N is the total number of samples, is the predicted value of the end-point judgment model for the reduction end time of the i-th sample, is the true value of the reduction end time of the i-th sample, is a hyperparameter, which is a very small real number to prevent the denominator of the formula from being zero, is the loss function threshold, which is set to 0.5.

[0077] As Figure 3 shown, the following is an embodiment of the end-point prediction system for the oxidation-reduction period of the anode furnace provided by the present disclosure. This system and the end-point prediction method for the oxidation-reduction period of the anode furnace in the above embodiments belong to the same inventive concept. For the details not described in detail in the embodiment of the end-point prediction system for the oxidation-reduction period of the anode furnace, reference can be made to the embodiment of the end-point prediction method for the oxidation-reduction period of the anode furnace.

[0078] An end-point prediction system for the oxidation-reduction period of an anode furnace includes: A data acquisition module, configured to acquire the sample images of the front and back of the anode copper observation sample, and record the acquisition relative time and the sample distance to the end time.

[0079] A data processing module, configured to label the sample images according to the acquisition relative time and the sample distance to the end time, and generate the to-be-tested image data; the label records the sample distance to the end time.

[0080] An end-point prediction module, configured to input the to-be-tested image data into a pre-trained end-point judgment model, and use an encoder and three multi-layer perceptrons in the end-point judgment model to output the classification of the oxidation or reduction stage, the time to the oxidation end point, and the time to the reduction end point, and integrate the output data to predict the final distance to the end time.

[0081] The end-point prediction system for the oxidation-reduction period of the anode furnace provided by this embodiment combines image recognition technology with time data analysis, and uses a pre-trained deep learning model to achieve accurate prediction of the end point of the oxidation-reduction period of the anode furnace. This system not only significantly improves the prediction accuracy, but also promotes the automation and intelligence of the smelting process, effectively reduces the error of manual judgment, improves the production efficiency, and reduces the cost. At the same time, this system also shows a positive impact on environmental protection and energy conservation, bringing significant improvements and benefits to the anode furnace smelting field, and having important application value.

[0082] Figure 4 It is a schematic diagram of the hardware structure of an electronic device for implementing various embodiments of the present invention.

[0083] The method for predicting the end point of the oxidation-reduction period of the anode furnace provided by the embodiments of the present application can be applied to electronic devices. Those skilled in the art can understand that the structure of the electronic devices involved in the embodiments of the present invention does not constitute a limitation on the electronic devices. The electronic devices may include more or fewer components than those shown in the figures, or combine some components, or have different component arrangements. In the embodiments of the present invention, the electronic devices include, but are not limited to, laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic devices may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of the present application described herein and / or claimed.

[0084] The electronic device may include a processor, an external memory interface, an internal memory, a universal serial bus (USB) interface, a charging management module, a power management module, a battery, a wireless communication module, an audio module, a speaker, a microphone, a sensor module, keys, a camera, a display screen, and a SIM card interface, etc.

[0085] The processor may include one or more processing units. For example, the processor may include a central processing unit (CPU), etc., an application processor (AP), a modem processor, a graphics processing unit (GPU), an image signal processor (ISP), a controller, a memory, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural-network processing unit (NPU), etc. Among them, different processing units may be independent devices or integrated in one or more processors.

[0086] Among them, the processor may be the nerve center and command center of the electronic device. The controller may generate operation control signals according to the instruction operation code and timing signals to complete the control of fetching instructions and executing instructions.

[0087] A memory can also be set in the processor for storing instructions and data. In some embodiments, the memory in the processor is a cache memory. This memory can save the instructions or data that the processor has just used or recycled. If the processor needs to use the instruction or data again, it can directly call it from this memory. This avoids repeated accesses, reduces the waiting time of the processor, and thus improves the system efficiency.

[0088] The external memory interface can be used to connect an external memory card, such as a MicroSD card, to implement the storage capacity expansion of the electronic device. The external memory card communicates with the processor through the external memory interface to implement the data storage function. For example, files such as music and videos are saved in the external memory card.

[0089] The internal memory can be used to store computer-executable program codes, and the computer-executable program codes include instructions. The processor executes various functional applications and data processing of the electronic device by running the instructions stored in the internal memory. The internal memory can include a program storage area and a data storage area. The internal memory can include a high-speed random access memory and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, a universal flash storage (UFS), etc.

[0090] The wireless communication function of the electronic device can be implemented through an antenna, a wireless communication module, a modulation and demodulation processor, a baseband processor, etc.

[0091] The wireless communication module can provide wireless communication solutions applied to the electronic device, including wireless local area networks (WLANs) (such as wireless fidelity (Wi-Fi) networks), Bluetooth (BT), global navigation satellite systems (GNSSs), frequency modulation (FM), near field communication (NFC), infrared technology (IR), etc.

[0092] The electronic device can implement audio functions, etc. through an audio module, a speaker, a receiver, a microphone, a headphone interface, an application processor, etc.

[0093] The electronic device can implement a shooting function through an ISP, a camera, a video codec, a GPU, a display screen, an application processor, etc.

[0094] An electronic device can implement a display function through a GPU, a display screen, an application processor, etc.

[0095] The GPU is a microprocessor for image processing, which connects the display screen and the application processor. The GPU is used to perform mathematical and geometric calculations for graphics rendering. The processor may include one or more GPUs, which execute program instructions to generate or change display information.

[0096] The display screen is used to display images, videos, etc. The display screen includes a display panel.

[0097] The above electronic device realizes the deep fusion of the image recognition technology and time data of the endpoint prediction method of the oxidation-reduction period of the anode furnace in this application, and combines a pre-trained deep learning model for accurate prediction, achieving the beneficial effects of significantly improving the accuracy of the endpoint prediction of the oxidation-reduction period of the anode furnace, promoting the automation and intelligence of the smelting process, reducing production costs, and enhancing environmental friendliness.

[0098] In the storage medium provided by this application, there is a program product capable of realizing the endpoint prediction method of the oxidation-reduction period of the anode furnace.

[0099] The endpoint prediction method of the oxidation-reduction period of the anode furnace includes: collecting sample images of the front and back of the anode copper observation sample, and recording the relative collection time and the time of the sample from the endpoint; according to the relative collection time and the time of the sample from the endpoint, labeling the sample image to generate the image data to be tested; the label records the time of the sample from the endpoint; inputting the image data to be tested into the pre-trained endpoint judgment model, and using an encoder and three multi-layer perceptrons in the endpoint judgment model to output the classification of the oxidation or reduction stage, the time from the oxidation endpoint, and the time from the reduction endpoint, and integrating the output data to predict the final time from the endpoint.

[0100] In some possible implementation manners, the endpoint prediction method of the oxidation-reduction period of the anode furnace in the present disclosure may be implemented in the form of a program product, which includes program code. When the program product runs on a terminal device, the program code is used to enable the terminal device to execute the steps according to various exemplary embodiments of the present disclosure described in the "Exemplary Method" section above of this specification.

[0101] The storage medium of the present disclosure may adopt any combination of one or more readable media. The readable media may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0102] The foregoing description of the disclosed embodiments enables those skilled in the art to practice or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the end point of the oxidation-reduction period of a reverberatory furnace, characterized in that, Including: Collecting sample images of the front and back of the anode copper observation sample, and recording the relative acquisition time and the time of the sample from the end point; Labeling the sample images according to the relative acquisition time and the time of the sample from the end point to generate the image data to be measured; the label records the time of the sample from the end point; Inputting the image data to be measured into a pre-trained end point judgment model, and using one encoder and three multi-layer perceptrons in the end point judgment model to output the classification of the oxidation or reduction stage, the time from the oxidation end point, and the time from the reduction end point, and integrating the output data to predict the final time from the end point.

2. The method for predicting the end point of the oxidation-reduction period of the anode furnace according to claim 1, wherein The collecting of the sample images of the front and back of the anode copper observation sample and recording the relative acquisition time and the time of the sample from the end point includes: Using an industrial camera to collect the visual features of the front and back of the anode copper observation sample to generate sample images; Recording the furnace number, oxidation or reduction stage, time from the end point, relative acquisition time of the sample, and storing them in a database; The relative acquisition time is the time interval between the sample sampling moment and the shooting moment, and the time from the end point is the time interval between the artificial predicted end point sample sampling moment and the sample sampling moment.

3. The method for predicting the end point of the oxidation-reduction period of the anode furnace according to claim 2, wherein The labeling of the sample images according to the relative acquisition time and the time of the sample from the end point to generate the image data to be measured includes: If the sample image is the last sample image collected in the sampling stage, then mark this sample image as the end point sample; if the sample image is not the last sample image collected in the sampling stage, then mark this sample image as the process sample; For the end point sample, label the time of the sample from the end point as the label; For the process sample, calculate the reference time of the sample from the end point according to the relative acquisition time and the time of the sample from the end point, and the relative acquisition time and the time of the corresponding end point sample from the end point, and determine the labeling time of this sample based on the time of the sample from the end point, and use the labeling time as the label of this sample.

4. The method for predicting the end point of the oxidation-reduction period of the anode furnace according to claim 3, wherein The calculating of the reference time of the sample from the end point according to the relative acquisition time and the time of the sample from the end point, and the relative acquisition time and the time of the corresponding end point sample from the end point, and determining the labeling time of this sample based on the time of the sample from the end point includes: During the same sampling stage, the shooting time of the process sample is , the shooting time of the end sample is , the relative acquisition time of the end sample is recorded as , the relative acquisition time of the process sample is recorded as , then the true sampling times of the end sample and the process sample 、 are respectively: Record the time of the sample at the end point as the time to the end point , then calculate the sampling time of the true end-point sample through the following formula : Based on the actual collection time of the process sample and the collection time of the actual end-point sample, calculate the reference sample time to the end point through the following formula : Obtain the time to the end point for the sample of this process ; Determine Whether the difference from is less than a preset threshold value; If so, then take as the marked time of this sample; Otherwise, use as the marked time of this sample.

5. The method for predicting the end point of the oxidation-reduction period of the anode furnace according to claim 4, characterized in that, The end point judgment model includes: one encoder and three multi-layer perceptrons; The encoder uses an artificial intelligence model or a neural network as the base network; The encoder is used to read the image data to be measured, generate an intermediate tensor of 3×n, and decompose it into three output scalars as the inputs of the three multi-layer perceptrons respectively; The three multi-layer perceptrons include the first multi-layer perceptron, the second multi-layer perceptron, and the third multi-layer perceptron; The first multi-layer perceptron is used to map the continuous output value of the sigmoid function to the output oxidation or reduction stage classification value c after using the sigmoid function and passing through the threshold processing, and the value range of c is {0,1}, where 0 represents the oxidation stage and 1 represents the reduction stage; The second multi-layer perceptron is used to output the time t1 from the oxidation end point; The third multi-layer perceptron is used to output the time t2 from the reduction end point.

6. The method for predicting the end point of the oxidation-reduction period of the anode furnace according to claim 5, wherein The end point judgment model further includes: an image enhancement module and a calculation module; The image enhancement module is used to input the encoder after enhancing and processing the image data to be measured by using an enhancement strategy; the enhancement strategy includes, but is not limited to, image flipping, image occlusion, color contrast enhancement, and Gaussian noise processing; The calculation module is used to utilize the formula to calculate the final distance-to-end time based on the outputs of three multi-layer perceptrons .

7. The method for predicting the end point of the oxidation-reduction period of the anode furnace according to claim 6, wherein, The training process of the endpoint judgment model includes: Continuously training the endpoint judgment model by using a preset training set and a loss function until the loss of the endpoint judgment model converges or the number of training rounds reaches a preset upper limit; Calculating the precision and recall of the oxidation-reduction stage classification, the precision and recall of the oxidation endpoint judgment, the precision and recall of the reduction endpoint judgment, and the MAE of the prediction results of the oxidation stage and the reduction stage respectively, and evaluating the trained endpoint judgment model; According to the evaluation results, optimizing and adjusting the hyperparameters of the endpoint judgment model; The loss function includes a classification loss , an oxidation endpoint regression loss and a reduction endpoint regression loss ; Classification loss function is defined as follows: where N is the total number of samples, is the label value 0 or 1 of the i-th sample, is the predicted value of the end-point judgment model for the i-th sample; Oxidation endpoint regression loss function It is defined as: where N is the total number of samples, is the predicted value of the end-point judgment model for the oxidation end-point time of the i-th sample, is the true value of the oxidation end-point time of the i-th sample, is a hyperparameter, is the loss function threshold; Restoring end regression loss is defined as: where N is the total number of samples, is the predicted value of the end point judgment model for the reduction end time of the i-th sample, is the true value of the reduction end time of the i-th sample, is a hyperparameter, is the loss function threshold.

8. An endpoint prediction system for the oxidation-reduction period of an anode furnace, characterized in that, The system adopts the anode furnace oxidation-reduction period endpoint prediction method according to any one of claims 1 to 7; The system includes: A data acquisition module, which is used to acquire the sample images of the front and back of the anode copper observation sample, and record the acquisition relative time and the sample distance from the endpoint time; A data processing module, which is used to label the sample images according to the acquisition relative time and the sample distance from the endpoint time, and generate the image data to be measured; the label records the sample distance from the endpoint time; An endpoint prediction module, which is used to input the image data to be measured into a pre-trained endpoint judgment model, and use one encoder and three multi-layer perceptrons in the endpoint judgment model to output the classification of the oxidation or reduction stage, the time from the oxidation endpoint, and the time from the reduction endpoint, and integrate the output data to predict the final time from the endpoint.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the anode furnace oxidation-reduction period endpoint prediction method according to any one of claims 1 to 7.

10. A storage medium, on which a computer program is stored, characterized in that, When the computer program is executed by the processor, it implements the steps of the anode furnace oxidation-reduction period endpoint prediction method according to any one of claims 1 to 7.

Images