Homogenizing furnace temperature control method, system and storage medium

Abstract

The present invention discloses a temperature control method, system and storage medium for a homogenizing furnace. The method includes: acquiring production data for a first production period and a second production period, using a temperature prediction model to perform temperature prediction on the production data of the two periods respectively, and obtaining a first temperature prediction result and a second temperature prediction result; collecting a thermal energy map for the second production period, and using an auxiliary prediction model to predict the target thermal energy map to obtain a predicted thermal energy map at the second moment; obtaining a temperature prediction result based on the first temperature prediction result and the second temperature prediction result, combined with the predicted thermal energy map, and controlling the temperature of the temperature compensation zone based on the temperature prediction result. The present application can prejudge the temperature of the temperature compensation zone of the homogenizing furnace in advance, realize the temperature prediction and temperature automatic control of the homogenizing furnace, improve the production efficiency and the production quality of aluminum alloys, and reduce the production cost. The present application is applied to the field of homogenizing furnace control technology.

Description

Technical Field

[0001] The present invention relates to the technical field of homogenizing furnace control, and in particular to a homogenizing furnace temperature control method, system and storage medium. Background Art

[0002] A homogenizing furnace is a device used for metal heat treatment, typically equipped with heating and holding zones. After casting, aluminum alloy products are prone to internal stress and compositional segregation. By controlling the temperature, a homogenizing furnace can alter the metallographic structure within the aluminum alloy, eliminating these casting stresses and compositional segregation to a certain extent, thereby improving the quality of the aluminum alloy product.

[0003] Currently, aluminum bars are typically stacked on a feed cart and then fed into a homogenizing furnace for heating. The interior of the homogenizing furnace is typically equipped with a heating zone and a holding zone. The holding zone, also known as the temperature compensation zone, heats the aluminum bars, while the holding zone compensates for and maintains the temperature, completing the aluminum alloy processing. The quality of aluminum alloy products depends largely on whether the temperatures in the heating and holding zones within the homogenizing furnace are within production requirements, particularly the temperature of the temperature compensation zone. The temperature compensation zone compensates for uneven temperature distribution across the aluminum alloy, addressing the issue of uneven temperature distribution. If the temperature in the temperature compensation zone fails to meet production requirements, the quality of the aluminum alloy will fall short of expectations. Therefore, real-time temperature control and prediction of the temperature in the homogenizing furnace's temperature compensation zone are essential.

[0004] However, existing homogenizing furnaces typically only have a temperature detection function. During manufacturing, workers typically manually adjust the temperature of the holding zone based on the detected temperature. This temperature control method relies heavily on workers' practical experience, resulting in high production and labor costs and reduced production efficiency. Therefore, how to predict the temperature of the heating and holding zones of a homogenizing furnace and control the temperature of the homogenizing furnace based on the predicted temperatures has become a pressing technical problem in the industry. Summary of the Invention

[0005] The purpose of the present invention is to provide a homogenizing furnace temperature control method, system and storage medium to solve one or more technical problems existing in the prior art and at least provide a beneficial option or create conditions.

[0006] The present invention solves the technical problem as follows: First, the present application provides a method for controlling the temperature of a homogenizing furnace, which is applied to a homogenizing furnace, wherein a heating zone and a temperature compensation zone are provided in the homogenizing furnace, comprising the following steps:

[0007] S100, obtaining first production data of a first production period, and performing temperature prediction on the first production data using a temperature prediction model to obtain a first temperature prediction result, wherein the first temperature prediction result includes temperature data at a second moment, the second moment occurring after the first production period;

[0008] S200, obtaining second production data of a second production period, and performing temperature prediction on the second production data using a temperature prediction model to obtain a second temperature prediction result, wherein the second production period is a period from the end of the first production period to the second moment, and the second temperature prediction result includes the temperature data at the second moment;

[0009] The heating time of the heating zone is composed of the first production period and the second production period, and the second moment is the time value when the temperature compensation zone starts to perform temperature compensation;

[0010] S300, collecting a thermal energy map of the second production period, and using an auxiliary prediction model to predict the target thermal energy map to obtain a predicted thermal energy map at the second moment;

[0011] S400 , obtaining a temperature prediction result based on the first temperature prediction result and the second temperature prediction result in combination with the predicted thermal energy map, and controlling the temperature of the temperature compensation area based on the temperature prediction result.

[0012] As a further improvement of the above technical solution, the temperature prediction model is a pre-trained neural network model, and the temperature prediction model is obtained by training the first sample after preprocessing; the input data of the first sample is: the energy consumption data, pressure data and temperature data of the heating zone in the historical production cycle and the airflow data of the combustion-supporting fan installed in the heating zone, and the output result of the first sample is: the temperature data at the second moment of the historical production cycle; the above data is collected by multiple sensors arranged in the heating zone and the temperature compensation zone.

[0013] As a further improvement of the above technical solution, the step of preprocessing the first sample includes:

[0014] Dividing the first sample into a training set and a validation set, and using the training set as an initial candidate feature subset;

[0015] Performing several feature selections in parallel based on the initial candidate feature subset to obtain several feature sets, and performing weighted integration on the several feature sets to obtain the integration weight of the initial candidate feature subset;

[0016] Deleting sample production data with a value lower than the integration weight in the initial feature subset to generate a new candidate feature subset;

[0017] Based on the new candidate feature subset, a support vector machine classifier is trained in the training set, and samples of the validation set are classified by the trained support vector machine classifier;

[0018] Determine whether the dimension of the new candidate feature subset is less than or equal to three; if so, select the candidate feature subset with the highest accuracy as the sample data set; if not, return to the step of performing several feature selections in parallel based on the initial candidate feature subset.

[0019] As a further improvement to the above technical solution, performing weighted integration on the plurality of feature sets to obtain the integration weight of the initial candidate feature subset includes:

[0020] The class separability of the feature set is defined, and a weighted set of several feature sets is performed according to the class separability to obtain the integration weight; the integration weight satisfies:

[0021]

[0022] Among them, α is the integration weight, D(s k ) is the category separability of the feature set, s k is a feature set.

[0023] As a further improvement of the above technical solution, the auxiliary prediction model is a pre-trained neural network model, and the auxiliary prediction model is obtained by training the second sample after preprocessing. The input data of the second sample is: the thermal energy map of the second production period of the historical production cycle, and the output result of the second sample is: the thermal energy map of the second moment of the historical production cycle.

[0024] As a further improvement of the above technical solution, in S400, obtaining a temperature prediction result based on the first temperature prediction result and the second temperature prediction result in combination with the predicted thermal energy map includes:

[0025] S410, converting the predicted thermal energy map into feature vectors Y1, Y2, ..., Y i ; Y i represents the temperature characteristics corresponding to the i-th time frame in the second production period;

[0026] S420: Use the Pearson correlation coefficient to measure the correlation coefficient r1 between the temperature feature and the first temperature prediction result and the correlation coefficient r2 between the temperature feature and the second temperature prediction result, where:

[0027] is the sub-correlation coefficient, Y1 is the first temperature prediction result;

[0028] is the sub-correlation coefficient, i=1, 2, ..., n, Y2 is the second temperature prediction result; |r i 1 | and |r i 2 The closer | is to 1, the temperature characteristic Y i The smaller the gap with the temperature prediction result, the greater the correlation between the two;

[0029] S430, determining whether r1 is greater than a first correlation threshold, and determining whether the correlation coefficient r2 is greater than a second correlation threshold; if both r1 and r2 are greater than the corresponding correlation thresholds, executing S440; if both r1 and r2 are less than or equal to the corresponding correlation thresholds, correcting the first temperature prediction result and the second temperature prediction result using the eigenvector, and returning to S420; if r1 is greater than the first correlation threshold and r2 is less than or equal to the second correlation threshold, correcting the second temperature prediction result using the eigenvector, and returning to S420; if r1 is less than or equal to the first correlation threshold and r2 is greater than the second correlation threshold, correcting the first temperature prediction result using the eigenvector, and returning to S420;

[0030] S440: Obtain a first performance indicator of the temperature prediction model and a second performance indicator of the auxiliary prediction model, assign a first weight to the first temperature prediction result according to the first performance indicator and the second performance indicator, and assign a second weight to the second temperature prediction result;

[0031] S450: Perform weighted calculation on the first temperature prediction result and the second temperature prediction result according to the first weight and the second weight to obtain a temperature prediction result.

[0032] As a further improvement of the above technical solution, the step of correcting the first temperature prediction result by using the characteristic vector includes:

[0033] Filter out the sub-correlation coefficient r i 1 The first temperature prediction result is corrected using the following formula based on the eigenvectors that are greater than or equal to the first sub-correlation threshold:

[0034]

[0035] Among them, Y j is the filtered eigenvector, and its corresponding sub-correlation coefficient is r j 1 , j = 1, 2,…, m1. And m1 < n.

[0036] As a further improvement of the above technical solution, the step of correcting the second temperature prediction result by using the characteristic vector includes:

[0037] Filter out the sub-correlation coefficient r i 2 The second temperature prediction result is corrected using the following formula based on the eigenvectors that are greater than or equal to the second sub-correlation threshold:

[0038]

[0039] Among them, Y p is the filtered eigenvector, and its corresponding sub-correlation coefficient is r j 2 , p=1,2,…,m2.And m2<n.

[0040] In a second aspect, the present application provides a homogenizing furnace temperature control system, comprising:

[0041] a first prediction unit, configured to obtain first production data of a first production period, and perform temperature prediction on the first production data using a temperature prediction model to obtain a first temperature prediction result, wherein the first temperature prediction result includes temperature data at a second moment, the second moment occurring after the first production period;

[0042] The second prediction unit is used to obtain second production data of a second production period, and use the temperature prediction model to perform temperature prediction on the second production data to obtain a second temperature prediction result. The second production period is a time period from the end of the first production period to the second moment. The second temperature prediction result includes the temperature data at the second moment.

[0043] a third prediction unit, configured to collect the thermal energy map of the second production period, and predict the target thermal energy map using the auxiliary prediction model to obtain a predicted thermal energy map at the second moment;

[0044] A correction output unit is used to obtain a temperature prediction result based on the first temperature prediction result and the second temperature prediction result in combination with the predicted thermal energy map, and control the temperature of the temperature compensation area according to the temperature prediction result.

[0045] In a third aspect, the present application further provides a storage medium storing processor-executable instructions, wherein the processor-executable instructions are used to execute the homogenizing furnace temperature control method when executed by the processor.

[0046] The beneficial effects of the present invention are: providing a homogenizing furnace temperature control method, system and storage medium, predicting the temperature data of the temperature compensation zone through the production data of the heating zone, and numerically correcting the temperature data through the predicted thermal energy map, which not only can predict the temperature of the temperature compensation zone of the homogenizing furnace in advance, improve the accuracy of temperature prediction, accurately predict temperature, but also ensure that the temperature of the temperature compensation zone is within the production requirement range, ensure that the temperature distribution of the aluminum alloy is uniform, realize the temperature prediction and temperature automatic control of the homogenizing furnace, improve the production efficiency and production quality of the aluminum alloy and reduce the production cost.

[0047] Other features and advantages of the present application will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present application. The purposes and other advantages of the present application can be realized and obtained through the structures particularly pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A flow chart of the homogenizing furnace temperature control method provided in an embodiment of the present application;

[0049] Figure 2 A flowchart of the steps for pre-processing a first sample provided in an embodiment of the present application;

[0050] Figure 3 This is a structural diagram of the homogenizing furnace temperature control system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0052] The present application is further described below in conjunction with the accompanying drawings and specific embodiments. The described embodiments should not be considered as limiting the present application. All other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0053] In the following description, reference is made to “some embodiments”, which describes a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0055] Homogenizing furnaces are used for metal heat treatment. After casting, aluminum alloys are prone to internal stress and compositional segregation. Homogenizing furnaces, by controlling the temperature, can alter the metallographic structure within the aluminum alloy, eliminating these stresses and compositional segregation to a certain extent, thereby improving the quality of aluminum alloy products.

[0056] Currently, aluminum bars are typically stacked on a feed cart and then fed into a homogenizing furnace for heating. The furnace is typically equipped with a heating zone and a holding zone, also known as a temperature compensation zone. The heating zone heats the aluminum bars, while the holding zone compensates for and maintains the temperature, completing the aluminum alloy processing. The quality of aluminum alloy products depends largely on whether the temperatures in the heating and holding zones within the homogenizing furnace are within production requirements, particularly the temperature of the temperature compensation zone. The temperature compensation zone compensates for uneven temperature distribution across the aluminum alloy, addressing the issue of uneven temperature distribution. Failure to meet production requirements in the temperature compensation zone results in lower quality and substandard production. Therefore, real-time control and prediction of the homogenizing furnace temperature are essential.

[0057] However, existing homogenizing furnaces typically only have a temperature detection function. During manufacturing, workers typically manually adjust the temperature of the heating and holding zones based on the detected temperatures. This temperature control method relies heavily on workers' practical experience, resulting in high production and labor costs and reduced production efficiency. Therefore, how to predict the temperatures of the heating and holding zones of a homogenizing furnace and control the temperature of the homogenizing furnace based on the predicted temperatures has become a pressing technical problem in the industry.

[0058] In this regard, the embodiments of the present application provide a homogenizing furnace temperature control method, system and storage medium to solve the problems existing in the above-mentioned prior art. The heating zone in the homogenizing furnace is used for heating, and the temperature compensation is used to solve the problem of uneven temperature distribution on the aluminum alloy. If the temperature of the temperature compensation zone cannot meet the production requirements, the quality of the aluminum alloy will be low and will not meet the expected production requirements. In this regard, the present application predicts the temperature of the temperature compensation zone through the production data of the heating zone, which not only ensures that the temperature of the temperature compensation zone is within the production requirement range and ensures the uniform temperature distribution of the aluminum alloy, but also realizes production automation. During the homogenization process, there is no need to manually adjust the temperature in the homogenizing furnace, which improves production efficiency and reduces production costs, and can ensure the production quality of the aluminum alloy.

[0059] Reference Figure 1 The flow chart of the homogenizing furnace temperature control method shown is an embodiment of the present application, and the homogenizing furnace temperature control method will be described and elaborated below. The homogenizing furnace temperature control method may include but is not limited to the following steps.

[0060] S100 , obtaining first production data of a first production period, and performing temperature prediction on the first production data using a temperature prediction model to obtain a first temperature prediction result.

[0061] It should be noted that the first temperature prediction result includes temperature data at a second moment, and the second moment occurs after the first production period.

[0062] It should be noted that the temperature prediction model is a pre-trained neural network model, trained using the preprocessed first sample. The input data for the first sample includes: energy consumption, pressure, and temperature data for the heating zone during a historical production cycle, as well as airflow data for the combustion-supporting fan within the heating zone. The output of the first sample is: temperature data at the second moment of the historical production cycle. Optionally, the first sample and first production data are collected using multiple sensors located in the heating zone and temperature compensation zone.

[0063] Optionally, the temperature prediction model is a BP neural network model.

[0064] S200 , obtaining second production data of a second production period, and performing temperature prediction on the second production data using a temperature prediction model to obtain a second temperature prediction result.

[0065] It should be noted that the second production period is the time period from the end of the first production period to before the second moment, and the second temperature prediction result includes the temperature data at the second moment.

[0066] It should be noted that the heating time of the heating zone consists of the first production period and the second production period, and the second moment is defined as the time value when the temperature compensation zone starts to perform temperature compensation.

[0067] S300 , collecting a thermal energy map of the second production period, and using an auxiliary prediction model to predict the target thermal energy map to obtain a predicted thermal energy map at the second moment.

[0068] It should be noted that the auxiliary prediction model is a pre-trained neural network model, trained using the preprocessed second sample. The input data for the second sample is the thermal energy map for the second production period of the historical production cycle; the output of the second sample is the thermal energy map for the second moment of the historical production cycle. Optionally, the auxiliary prediction model is a YOLOv5-based neural network model.

[0069] S400 , obtaining a temperature prediction result based on the first temperature prediction result and the second temperature prediction result in combination with a predicted thermal energy map, and controlling the temperature of the temperature compensation zone based on the temperature prediction result.

[0070] In the present application, the temperature value at the second moment, i.e., the temperature value at which the temperature compensation zone begins to perform temperature compensation, is predicted respectively by using the data from the two stages of the heating time, and a thermal energy map at the second moment is obtained by predicting the thermal energy map of the stage close to the second moment. The thermal energy map can more comprehensively reflect the temperature data at the second moment, and the temperature value at the second moment is numerically corrected by the temperature characteristics mapped on the thermal energy map, thereby improving the accuracy of the predicted temperature data. Finally, the temperature of the temperature compensation zone of the homogenizing furnace is controlled by the temperature prediction data. The present application realizes the temperature prediction and automatic temperature control of the homogenizing furnace, can predict the temperature of the temperature compensation zone of the homogenizing furnace in advance, improves the accuracy of the temperature prediction, and can accurately perform temperature prediction.

[0071] Reference Figure 2 The flowchart of the steps for pre-processing the first sample shown is an embodiment of the present application, and the pre-processing of the first sample will be further described and elaborated below.

[0072] The accuracy of a temperature prediction model depends largely on the quality and quantity of the first sample. To ensure this accuracy, it is necessary to ensure that the features selected for the first sample meet the actual prediction requirements. In this field, swarm-inspired intelligent algorithms such as particle swarm optimization (PSO) or cuckoo search algorithms are commonly used for data feature selection. PSO algorithms offer fast early convergence, but suffer from weak search capabilities in the later stages of iteration.

[0073] To this end, this application performs feature selection on the first sample through a recursive feature elimination process (RFE) combined with a particle swarm algorithm to improve the ability of iterative post-search, ensure the stability and robustness of the feature selection method, and select the relationship between the energy consumption data, pressure data and temperature data of the heating zone in the historical production cycle and the airflow data of the combustion-supporting fan installed in the heating zone and the temperature data at the second moment.

[0074] The pre-processing of the first sample may include but is not limited to the following steps.

[0075] Step 1: Divide the first sample into a training set and a validation set, and use the training set as the initial candidate feature subset;

[0076] Step 2: Based on the initial candidate feature subset, several feature selections are performed in parallel to obtain several feature sets, and the several feature sets are weighted integrated to obtain the integration weight of the initial candidate feature subset.

[0077] In the above steps, the dataset is divided into a training set and a validation set. Parallel feature selection based on the initial candidate feature subset can objectively reflect the relative importance of features and is conducive to improving the quality of feature selection.

[0078] Furthermore, the feature selection is performed several times in parallel based on the initial candidate feature subset to obtain several feature sets including:

[0079] Step 2.1, determine the particle search space dimension, that is, the size of a single particle. The particle dimension is mapped to the first sample value of the feature selection. Initialize 2 according to the particle dimension. n Particles and various parameters in the PSO algorithm, set the number of small populations to a1, the maximum number of iterations to k max , fitness function and fitness threshold T f .

[0080] It should be noted that the fitness function is shown in the following formula:

[0081]

[0082] Among them, n is the number of initial candidate feature subsets, n i represents the first sample in the initial candidate feature subset, θ1 represents the distance from n i The number of the nearest similar samples, θ2 is the distance from n i The number of recent outlier samples; Representation based on particles Corresponding feature set sample n i Similar neighbor samples the distance between them; Particle-based Corresponding feature set sample n i and heterogeneous neighbor samples The distance between particles; n' is the distance between particles The number of features in the corresponding feature set.

[0083] In this step, the Euclidean distance between samples is used to examine the particle The class separability of the corresponding feature combination in the sample is measured by the Euclidean distance between the sample and its nearest non-class neighbor, and the similarity between the sample and its nearest class neighbor is measured by the Euclidean distance between the sample and its nearest class neighbor. The smaller the Euclidean distance between samples of the same class, the denser the distribution of samples of the same class. The larger the Euclidean distance between samples of the non-class, the more dispersed the distribution of samples of the non-class. This fitness function, which balances the class separability and the dimensionality of the feature combination, facilitates the selection of low-dimensional classification feature sets with large inter-class differences and high intra-class similarity.

[0084] Step 2.2, randomly initialize the particle swarm velocities v1 and v2, set the initial inertia weight w, acceleration constants c1 and c2, and give the solution space range [-a2, a2] to suppress the gradient anomaly.

[0085] It should be noted that Among them, n i represents the i-th first sample in the initial candidate feature subset, and n is the number of initial candidate feature subsets;

[0086] In step 2.3, the particle swarm is disrupted and randomly divided into a1 small populations.

[0087] Optionally, the particle swarm is disrupted by a Mini-Batch algorithm.

[0088] Step 2.4: Calculate the fitness of each ion as the individual historical optimal value, the optimal fitness of each small population as the global historical optimal value of the small population, and the optimal fitness of the entire particle swarm as the global historical optimal solution. Update the speed, position, and individual fitness function value of each particle, and compare the current fitness function value of the ion with the individual historical optimal value. If it is better, replace and update the individual historical best ion and individual historical best fitness.

[0089] Step 2.5, calculate the optimal fitness of the small population, and use its corresponding ion as the optimal solution of the small population, and calculate the optimal value of the large particle population, and compare it with the historical global optimal value. If it is better, update and replace the global historical optimal value and the historical optimal solution;

[0090] Step 2.6, when k>k max or g(p g )>T fWhen , the iteration stops, p g is the historical global optimal value, k is the number of iterations, and the optimal solutions of a1 small populations are used as the feature set; otherwise, return to the step of calculating the fitness of each ion as the individual historical optimal value, that is, step 2.4.

[0091] Furthermore, several feature sets are weighted integrated to obtain the integration weights of the initial candidate feature subsets, including:

[0092] Step 2.7, define the class separability of the feature set, and perform weighted aggregation on several feature sets based on the class separability to obtain the integration weight. The integration weight is shown in the following formula:

[0093]

[0094] Among them, α is the integration weight, D(s k ) is the category separability of the feature set, s k is a feature set.

[0095] In this step, the features of the feature combination are weighted by the class separability index. The integrated feature weights can reflect the frequency of feature selection and the relative difference between heterogeneous samples and homogeneous samples. Features with high weights contribute more to feature selection, while features with low weights contribute less to feature selection. This application performs secondary feature selection by integrating weights.

[0096] Step 3: Delete the sample production data with a value lower than the integration weight in the initial feature subset to generate a new candidate feature subset;

[0097] Step 4: Based on the new candidate feature subset, a support vector machine classifier is trained in the training set, and the samples of the validation set are classified by the trained support vector machine classifier.

[0098] It should be noted that the Support Vector Machine (SVM) classifier is a classification model derived based on the principle of structural risk minimization and using statistical learning theory.

[0099] Step 5: Determine whether the dimension of the new candidate feature subset is less than or equal to three; if so, select the candidate feature subset with the highest accuracy as the sample data set; if not, return to the step of performing several feature selections in parallel based on the initial candidate feature subset, i.e., step 2.

[0100] In one embodiment of the present application, the auxiliary prediction model is obtained by training the preprocessed second sample, and the step of preprocessing the second sample may include but is not limited to the following steps.

[0101] The high-frequency information and low-frequency information of the sample heat map are decomposed by a wavelet frequency division algorithm to obtain a first sub-band, a second sub-band, a third sub-band and a fourth sub-band of the characteristic map.

[0102] It should be noted that the first sub-band is a low-frequency image, representing horizontal and vertical low-frequency information; the second sub-band, the third sub-band and the fourth sub-band are high-frequency images, representing horizontal low-frequency and vertical high-frequency information, horizontal high-frequency and vertical low-frequency information, and horizontal high-frequency and vertical high-frequency information, respectively.

[0103] The first sub-band is processed by a histogram equalization method, and the second sub-band, the third sub-band and the fourth sub-band are subjected to denoising.

[0104] The processed first sub-band, second sub-band, third sub-band and fourth sub-band are subjected to image reconstruction and histogram equalization by using a wavelet reconstruction algorithm to obtain a preprocessed sample thermal energy map.

[0105] In this specific embodiment, histogram equalization refers to a transformation algorithm that makes the output image histogram approximately follow a uniform distribution. Typically, when grayscale histogram equalization is used to enhance image quality, the grayscale values ​​of the pixels in the entire image are statistically analyzed. Therefore, when transforming the entire image, both low-frequency and high-frequency information and noise-containing information are transformed simultaneously. This can modify or lose image details, amplify noise, and reduce image information entropy. While this method can achieve image enhancement to a certain extent, it can also result in the loss of some image information.

[0106] The low-frequency portion corresponds to the average grayscale level in an image, reflecting the smoothness of the image and approximating it. The high-frequency portion, with increasingly variable grayscale levels, corresponds to edges, details, and noise. Therefore, processing the low-frequency portion of an image does not lose or alter image details or edge information, nor does it alter the noise.

[0107] In the above steps, the image is transformed using wavelet transform to obtain the first sub-band, and histogram equalization is performed on it to adjust the grayscale value in the approximate image, enhance the smooth part of the image, and improve the contrast. In this process, the details in the original high-frequency information will not be changed or lost, and the noise will not be amplified. Then, wavelet reconstruction is performed on the low-frequency image and the high-frequency image, and the reconstructed image is subjected to histogram equalization to enhance the image quality. There are two histogram equalization processes in this process. The first one makes the grayscale distribution of the horizontal low-frequency and vertical low-frequency parts of the image uniform. The second one is the histogram equalization process after the first low-frequency information processing, which increases the contrast of the image, reduces the loss of image detail information caused by equalization, reduces the amplification of noise, and retains the details of the original image.

[0108] At the same time, after the original image is decomposed by wavelet, the high-frequency image contains the edges and detailed contours of the original image, but it also contains the noise in the original image. The high-frequency image is basically dominated by noise information. Therefore, by suppressing the noise in the high-frequency image (including sub-low frequency, sub-high frequency and high frequency), the purpose of eliminating noise to a certain extent has been achieved. In this way, the noise in the high-frequency image can be eliminated to a certain extent without affecting the approximate part of the low-frequency image.

[0109] In one embodiment of the present application, S400 will be further described and elaborated below. In S400, the correlation between the temperature characteristics mapped by the predicted thermal energy map and the two temperature prediction results is first calculated respectively, and different correction operations are performed on the temperature prediction results according to the correlation. After the temperature prediction results are corrected, the weights corresponding to the temperature prediction results are set according to the performance parameters of the prediction model, and the weighted calculation between the two temperature prediction results is completed by the weights, thereby obtaining the final temperature prediction results. S400 may include but is not limited to the following steps.

[0110] S410, converting the predicted thermal energy map into feature vectors Y1, Y2, ..., Y i .

[0111] Among them, Y i represents the temperature characteristics corresponding to the i-th time frame in the second production period.

[0112] S420 , using the Pearson correlation coefficient to measure the correlation coefficient r1 between the temperature feature and the first temperature prediction result and the correlation coefficient r2 between the temperature feature and the second temperature prediction result.

[0113] It should be noted that the correlation coefficient r1 satisfies:

[0114]

[0115] Among them, r i 1 is the sub-correlation coefficient, i=1, 2, ..., n, and Y1 is the first temperature prediction result.

[0116] It should be noted that the correlation coefficient r2 satisfies:

[0117]

[0118] Among them, r i 2 is the sub-correlation coefficient, i=1, 2, ..., n, and Y2 is the second temperature prediction result.

[0119] Among them, |r i1 | and |r i 2 The closer | is to 1, the temperature characteristic Y i The smaller the gap with the temperature prediction result, the greater the correlation between the two.

[0120] S430, determining whether r1 is greater than a first correlation threshold, and determining whether the correlation coefficient r2 is greater than a second correlation threshold;

[0121] If r1 and r2 are both greater than the corresponding correlation thresholds, it indicates that the first temperature prediction result and the second temperature prediction result are highly similar to the temperature features mapped on the thermal energy map, and there is no significant error between the first temperature prediction result and the second temperature prediction result, then S440 is executed;

[0122] If r1 and r2 are both less than or equal to the corresponding correlation thresholds, it indicates that the first temperature prediction result and the second temperature prediction result have low similarity with the temperature features mapped on the thermal energy map, and there is a large error between the first temperature prediction result and the second temperature prediction result. Then, the first temperature prediction result and the second temperature prediction result are corrected using the feature vector, and the process returns to S420.

[0123] If r1 is greater than the first correlation threshold and r2 is less than or equal to the second correlation threshold, it indicates that the second temperature prediction result has a low similarity to the temperature feature mapped on the thermal energy map, while the first temperature prediction result has a high similarity, then the second temperature prediction result is corrected using the feature vector, and the process returns to S420;

[0124] If r1 is less than or equal to the first correlation threshold and r2 is greater than the second correlation threshold, it indicates that the first temperature prediction result has a low similarity to the temperature feature mapped on the thermal energy map, while the second temperature prediction result has a high similarity, then the first temperature prediction result is corrected using the feature vector, and the process returns to S420;

[0125] S440 , obtaining a first performance indicator of the temperature prediction model and a second performance indicator of the auxiliary prediction model, assigning a first weight to the first temperature prediction result according to the first performance indicator and the second performance indicator, and assigning a second weight to the second temperature prediction result.

[0126] Optionally, the performance indicator includes at least one of accuracy, precision, recall rate, and F-value.

[0127] S450 , performing weighted calculation on the first temperature prediction result and the second temperature prediction result according to the first weight and the second weight to obtain a temperature prediction result.

[0128] Furthermore, the step of correcting the first temperature prediction result using the characteristic vector includes:

[0129] Filter out the sub-correlation coefficient ri 1 The first temperature prediction result is corrected using the following formula based on the eigenvectors that are greater than or equal to the first sub-correlation threshold:

[0130]

[0131] Among them, Y j is the filtered eigenvector, and its corresponding sub-correlation coefficient is r j 1 , j = 1, 2,…, m1. And m1 < n.

[0132] Furthermore, the step of correcting the second temperature prediction result by using the characteristic vector includes:

[0133] The step of correcting the second temperature prediction result by using the characteristic vector includes:

[0134] Filter out the sub-correlation coefficient r i 2 The second temperature prediction result is corrected using the following formula based on the eigenvectors that are greater than or equal to the second sub-correlation threshold:

[0135]

[0136] Among them, Y p is the filtered eigenvector, and its corresponding sub-correlation coefficient is r j 2 , p=1,2,…,m2.And m2<n.

[0137] In an optional embodiment of the present application, controlling the temperature of the temperature compensation zone according to the temperature prediction result includes: presetting a first temperature threshold, and when the temperature prediction data is greater than the first temperature threshold, controlling the temperature of the temperature compensation zone to be lowered; when the temperature prediction data is less than or equal to the first temperature threshold, controlling the temperature of the temperature compensation zone to be increased.

[0138] Based on the above embodiment, refer to Figure 3 , the embodiment of the present application further provides a homogenizing furnace temperature control system, comprising:

[0139] The first prediction unit is configured to obtain first production data in a first production period, perform temperature prediction on the first production data using a temperature prediction model, and obtain a first temperature prediction result.

[0140] The first temperature prediction result includes temperature data at a second moment, and the second moment occurs after the first production period.

[0141] The second prediction unit is used to obtain second production data in a second production period, perform temperature prediction on the second production data using a temperature prediction model, and obtain a second temperature prediction result.

[0142] The second production period is a time period from the end of the first production period to before the second moment, and the second temperature prediction result includes temperature data at the second moment.

[0143] The third prediction unit is used to collect the thermal energy map of the second production period, and use the auxiliary prediction model to predict the target thermal energy map to obtain the predicted thermal energy map at the second moment.

[0144] The correction output unit is used to obtain a temperature prediction result based on the first temperature prediction result and the second temperature prediction result in combination with the predicted thermal energy map, and control the temperature of the temperature compensation zone according to the temperature prediction result.

[0145] In addition, an embodiment of the present application further provides a storage medium storing processor-executable instructions, which are used to execute a homogenizing furnace temperature control method when executed by the processor.

[0146] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can, for example, be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0147] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0148] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0149] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0150] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0151] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk.

[0152] The step numbers in the above method embodiment are only provided for the convenience of explanation and do not limit the order of the steps. The execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

Claims

1. A method for controlling the temperature of a homogenizing furnace, applied to a homogenizing furnace, wherein a heating zone and a temperature compensation zone are provided in the homogenizing furnace, characterized in that: The steps include: S100, obtaining first production data of a first production period, and performing temperature prediction on the first production data using a temperature prediction model to obtain a first temperature prediction result, wherein the first temperature prediction result includes temperature data at a second moment, the second moment occurring after the first production period; S200, obtaining second production data of a second production period, and performing temperature prediction on the second production data using a temperature prediction model to obtain a second temperature prediction result, wherein the second production period is a period from the end of the first production period to the second moment, and the second temperature prediction result includes the temperature data at the second moment; The heating time of the heating zone is composed of the first production period and the second production period, and the second moment is the time value when the temperature compensation zone starts to perform temperature compensation; S300, collecting a thermal energy map of the second production period, and using an auxiliary prediction model to predict the target thermal energy map to obtain a predicted thermal energy map at the second moment; S400 , obtaining a temperature prediction result based on the first temperature prediction result and the second temperature prediction result in combination with the predicted thermal energy map, and controlling the temperature of the temperature compensation area based on the temperature prediction result.

2. A homogenizing furnace temperature control method according to claim 1, characterized in that: The temperature prediction model is a pre-trained neural network model, and the temperature prediction model is obtained by training the first sample after preprocessing; the input data of the first sample are: the energy consumption data, pressure data and temperature data of the heating zone in the historical production cycle and the airflow data of the combustion-supporting fan installed in the heating zone, and the output result of the first sample is: the temperature data at the second moment of the historical production cycle; the data in the corresponding areas are collected by multiple sensors arranged in the heating zone and the temperature compensation zone.

3. A homogenizing furnace temperature control method according to claim 2, characterized in that: The step of preprocessing the first sample comprises: Dividing the first sample into a training set and a validation set, and using the training set as an initial candidate feature subset; Performing several feature selections in parallel based on the initial candidate feature subset to obtain several feature sets, and performing weighted integration on the several feature sets to obtain the integration weight of the initial candidate feature subset; Deleting sample production data with a value lower than the integration weight in the initial feature subset to generate a new candidate feature subset; Based on the new candidate feature subset, a support vector machine classifier is trained in the training set, and samples of the validation set are classified by the trained support vector machine classifier; Determine whether the dimension of the new candidate feature subset is less than or equal to three; if so, select the candidate feature subset with the highest accuracy as the sample data set; if not, return to the step of performing several feature selections in parallel based on the initial candidate feature subset.

4. A homogenizing furnace temperature control method according to claim 3, characterized in that: The weighted integration of the plurality of feature sets to obtain the integration weight of the initial candidate feature subset includes: The class separability of the feature set is defined, and a weighted set of several feature sets is performed according to the class separability to obtain the integration weight; the integration weight satisfies: Among them, α is the integration weight, D(s k ) is the category separability of the feature set, s k is a feature set.

5. A homogenizing furnace temperature control method according to claim 1, characterized in that: The auxiliary prediction model is a pre-trained neural network model, which is obtained by training the preprocessed second sample. The input data of the second sample is: the thermal energy map of the second production period of the historical production cycle, and the output result of the second sample is: the thermal energy map of the second moment of the historical production cycle.

6. A homogenizing furnace temperature control method according to claim 1, characterized in that: In S400, obtaining a temperature prediction result based on the first temperature prediction result and the second temperature prediction result in combination with the predicted thermal energy map includes: S410, converting the predicted thermal energy map into feature vectors Y1, Y2, ..., Y i ; Y i represents the temperature characteristics corresponding to the i-th time frame in the second production period; S420: Use the Pearson correlation coefficient to measure the correlation coefficient r1 between the temperature feature and the first temperature prediction result and the correlation coefficient r2 between the temperature feature and the second temperature prediction result, where: is the sub-correlation coefficient, Y1 is the first temperature prediction result; is the sub-correlation coefficient, i=1, 2, ..., n, Y2 is the second temperature prediction result; |r i 1 | and |r i 2 The closer | is to 1, the temperature characteristic Y i The smaller the gap with the temperature prediction result, the greater the correlation between the two; S430, determining whether r1 is greater than a first correlation threshold, and determining whether the correlation coefficient r2 is greater than a second correlation threshold; if both r1 and r2 are greater than the corresponding correlation thresholds, executing S440; if both r1 and r2 are less than or equal to the corresponding correlation thresholds, correcting the first temperature prediction result and the second temperature prediction result using the eigenvector, and returning to S420; if r1 is greater than the first correlation threshold and r2 is less than or equal to the second correlation threshold, correcting the second temperature prediction result using the eigenvector, and returning to S420; if r1 is less than or equal to the first correlation threshold and r2 is greater than the second correlation threshold, correcting the first temperature prediction result using the eigenvector, and returning to S420; S440: Obtain a first performance indicator of the temperature prediction model and a second performance indicator of the auxiliary prediction model, assign a first weight to the first temperature prediction result according to the first performance indicator and the second performance indicator, and assign a second weight to the second temperature prediction result; S450: Perform weighted calculation on the first temperature prediction result and the second temperature prediction result according to the first weight and the second weight to obtain a temperature prediction result.

7. A homogenizing furnace temperature control method according to claim 6, characterized in that: The step of correcting the first temperature prediction result by using the characteristic vector includes: Filter out the sub-correlation coefficient r i 1 The first temperature prediction result is corrected using the following formula based on the eigenvectors that are greater than or equal to the first sub-correlation threshold: Among them, Y j is the filtered eigenvector, and its corresponding sub-correlation coefficient is r j 1 , j = 1, 2,…, m1 and m1 < n.

8. A homogenizing furnace temperature control method according to claim 6, characterized in that: The step of correcting the second temperature prediction result by using the characteristic vector includes: Filter out the sub-correlation coefficient r i 2 The second temperature prediction result is corrected using the following formula based on the eigenvectors that are greater than or equal to the second sub-correlation threshold: Among them, Y p is the filtered eigenvector, and its corresponding sub-correlation coefficient is r j 2 , p=1,2,…,m2 and m2<n.

9. A storage medium storing instructions executable by a processor, characterized in that: The processor-executable instructions, when executed by the processor, are used to execute the homogenizing furnace temperature control method according to any one of claims 1 to 8.

Images