Multi-task casting blank quality defect prediction method and device based on adaptive loss

Abstract

The invention provides a multi-task casting blank quality defect prediction method and device based on adaptive loss, and relates to the technical field of artificial intelligence. The method comprises the following steps: acquiring a continuous casting production sample data set; preprocessing the continuous casting production sample data set to obtain a preprocessed sample data set; an initial multi-task learning model based on TabNet is constructed; according to the preprocessed sample data set, training an initial multi-task learning model based on the TabNet to obtain a trained multi-task learning model based on the TabNet; acquiring continuous casting production data to be predicted; preprocessing the continuous casting production data to be predicted to obtain preprocessed continuous casting production data to be predicted; and inputting the preprocessed continuous casting production data to be predicted into the trained multi-task learning model based on the TabNet to obtain an internal quality prediction result of the continuous casting billet. By adopting the method, the performance of predicting the internal quality defects of the continuous casting billet can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a multi-task casting quality defect prediction method and device based on adaptive loss. Background Art

[0002] The development of the steel industry is closely related to a country's industrial level. China's steel industry has entered the era of intelligent manufacturing and is in urgent need of intelligent transformation and upgrading. In the entire process of steel production, the prediction of organizational properties is a crucial link, covering the quality prediction of continuous casting billets and the prediction of mechanical properties of rolled products.

[0003] Continuous casting plays a connecting role in the steel production process. The complexity of the continuous casting process makes the continuous casting billet prone to surface defects such as surface cracks, slag inclusions and pores, as well as internal defects such as center cracks, segregation, inclusions, looseness and shrinkage. The appearance of internal quality defects of continuous casting billets is mainly related to multiple factors. Uneven cooling may lead to internal stress concentration, which in turn causes cracks or other defects; secondly, improper casting temperature control will affect the fluidity of the metal liquid, resulting in problems such as shrinkage and pores. In addition, uneven alloy composition will also lead to differences in physical properties and increase the risk of defects. During the casting process, the metal liquid inhales gases such as water vapor and oxygen, which may form pores. Unreasonable mold design may affect the filling and exhaust of the casting, causing defects such as cold shut. Insufficient fluidity of the metal liquid, mold material and design problems, and improper heat treatment process are also potential causes of internal defects. Accurately predicting the quality of continuous casting billets can not only avoid destructive experiments, but also help technicians adjust the process in time, reduce the defective rate, ensure the smooth progress of subsequent rolling processes, and improve product quality. However, due to the nonlinearity, strong coupling characteristics and multiple disturbance factors of the continuous casting process, it is very challenging to accurately predict product quality. Therefore, the prediction of internal quality defects of continuous casting billets has always been an important research direction in the continuous casting production process.

[0004] Traditional methods mainly predict the quality of ingots by establishing a mechanism model of the continuous casting process, but there are problems such as low prediction accuracy, complex models and poor generalization ability. At present, the internal quality defect methods of continuous casting ingots have achieved good performance, but they are all for single defect prediction tasks and cannot predict multiple quality defects at the same time. However, in practical applications, due to the complexity and strong coupling of the production process, a continuous casting ingot may have multiple internal quality defects. Predicting one internal quality defect cannot achieve the purpose of detecting defective products. Therefore, how to achieve the simultaneous prediction of multiple internal quality defects, that is, multi-task prediction, and improve the system's detection of defective products is of great significance. The research on multi-task learning started early and has formed a relatively mature theoretical system and application examples. The multi-task learning framework first proposed by Caruana laid the foundation for subsequent research; in recent years, with the development of deep learning, many multi-task learning models based on neural networks have been proposed one after another, such as the network structure using shared hidden layers and adaptive task weight adjustment strategies, and multi-task learning models based on convolutional neural networks and recurrent neural networks for Chinese text classification and image recognition; in addition, for specific application scenarios, multi-task learning methods combined with domain knowledge can improve the performance of models in practical applications. At present, there is no multi-task learning model applied to the prediction of internal quality defects of continuous casting billets. Summary of the invention

[0005] In order to solve the technical problem that the existing technology cannot simultaneously predict the internal quality defects of multiple continuous casting billets, the embodiment of the present invention provides a multi-task billet quality defect prediction method and device based on adaptive loss. The technical solution is as follows:

[0006] On the one hand, a multi-task casting quality defect prediction method based on adaptive loss is provided, the method is implemented by a multi-task casting quality defect prediction device based on adaptive loss, and the method comprises:

[0007] S1. Obtain a continuous casting production sample data set;

[0008] S2, preprocessing the continuous casting production sample data set to obtain a preprocessed sample data set;

[0009] S3. Build an initial TabNet-based multi-task learning model;

[0010] S4. Training the initial TabNet-based multi-task learning model according to the preprocessed sample data set to obtain a trained TabNet-based multi-task learning model;

[0011] S5. Acquire the continuous casting production data to be predicted; preprocess the continuous casting production data to be predicted to obtain the preprocessed continuous casting production data to be predicted; input the preprocessed continuous casting production data to be predicted into the trained TabNet-based multi-task learning model to obtain the continuous casting billet internal quality prediction result.

[0012] Optionally, the step S2 of preprocessing the continuous casting production sample data set to obtain a preprocessed sample data set includes:

[0013] S21, using a data mining algorithm to remove missing values ​​and outliers in the continuous casting production sample data set to obtain a continuous casting production sample data set after the removal process;

[0014] S22. Use ONE-HOT coding to digitize the processed continuous casting production sample data set, normalize the digitized data, and obtain a preprocessed sample data set.

[0015] Optionally, the initial TabNet-based multi-task learning model includes:

[0016] Feature selection module, TabNet sharing layer, splitting layer, and adaptive loss module;

[0017] Wherein, the feature selection module is used to perform feature selection on input data;

[0018] Wherein, the TabNet shared layer is used to extract high-level features;

[0019] Wherein, the splitting layer includes two fully connected layers and an MLP classification head;

[0020] The adaptive loss module is used to balance the training progress of different tasks.

[0021] Optionally, the step S4 of training the initial TabNet-based multi-task learning model according to the preprocessed sample data set to obtain a trained TabNet-based multi-task learning model includes:

[0022] S41, inputting the preprocessed sample data set into the initial TabNet-based multi-task learning model, performing feature selection through a feature selection module, and obtaining a selected feature set;

[0023] S42, inputting the selected feature set into the TabNet shared layer to extract high-level features and obtain feature representation of the high-level features;

[0024] S43, inputting the feature representation of the high-level feature into the splitting layer, performing calculations through two fully connected layers and the MLP classification head, outputting the calculation results, performing task matching according to the calculation results, and obtaining the prediction results of the quality defects of the ingot;

[0025] S44. According to the prediction results of the ingot quality defects, the initial TabNet-based multi-task learning model is trained through an adaptive loss module to obtain a trained TabNet-based multi-task learning model.

[0026] Optionally, the step S41 of inputting the preprocessed sample data set into the initial TabNet-based multi-task learning model, performing feature selection through a feature selection module, and obtaining a selected feature set includes:

[0027] S411, inputting the preprocessed sample data set into the initial TabNet-based multi-task learning model, using a random forest algorithm for calculation, and obtaining a feature contribution vector;

[0028] S412, according to the feature contribution vector, setting the prior knowledge vector and the threshold through the mechanism model, and calculating the feature weight;

[0029] S413: Calculate based on feature weights and preprocessed sample data sets to obtain

[0030] Choose a good feature set.

[0031] Optionally, the step S44 trains the initial TabNet-based multi-task learning model through an adaptive loss module according to the prediction result of the quality defects of the casting, including:

[0032] S441, initializing weights;

[0033] S442, calculating the total loss in a single round according to the initialization weight; calculating the gradient norm of a single task to the overall task according to the total loss in a single round; calculating the relative direction training speed of a single task according to the gradient norm of the single task to the overall task;

[0034] S443, calculating the average gradient norm of all tasks according to the gradient norm of a single task to the overall task;

[0035] S444, calculating the gradient loss in a single round according to the average gradient norm of all tasks in a single round, the gradient norm of a single task to the overall task, and the relative direction training speed of a single task;

[0036] S445, calculating the weights in a single round according to the gradient loss in a single round, obtaining the weights of the next round; back-propagating the update parameters according to the total loss in the single round, and updating the weights by normalizing the weights of the next round;

[0037] S446: Determine whether the current number of training rounds has reached the preset value and the total loss descent gradient is less than the set threshold. If so, stop training and output the trained model; if not, return to step S441 to continue training until the preset stop condition is met.

[0038] Optionally, the step S5 of inputting the preprocessed continuous casting production data to be predicted into the trained multi-task learning model based on TabNet to obtain the prediction result of the internal quality of the continuous casting billet includes:

[0039] S51, inputting the preprocessed continuous casting production data to be predicted into the trained multi-task learning model based on TabNet, calculating the probability of internal defect level in the data set, and obtaining the probability value of the defect level;

[0040] S52. Select the defect level with the highest probability value as the final ingot quality prediction result.

[0041] On the other hand, a multi-task casting quality defect prediction device based on adaptive loss is provided, and the device is applied to a multi-task casting quality defect prediction method based on adaptive loss, and the device comprises:

[0042] An acquisition unit, used for acquiring a continuous casting production sample data set;

[0043] A preprocessing unit, used for preprocessing the continuous casting production sample data set to obtain a preprocessed sample data set;

[0044] Construction unit, used to build the initial TabNet-based multi-task learning model;

[0045] A training unit, used to train the initial TabNet-based multi-task learning model according to the preprocessed sample data set to obtain a trained TabNet-based multi-task learning model;

[0046] The prediction unit is used to obtain the continuous casting production data to be predicted; preprocess the continuous casting production data to be predicted to obtain the preprocessed continuous casting production data to be predicted; input the preprocessed continuous casting production data to be predicted into the trained TabNet-based multi-task learning model to obtain the continuous casting billet internal quality prediction result.

[0047] Optionally, the preprocessing unit is used to:

[0048] Using a data mining algorithm to remove missing values ​​and outliers in the continuous casting production sample data set, and obtaining a continuous casting production sample data set after the removal process;

[0049] The processed continuous casting production sample data set is digitized using ONE-HOT coding, and the digitized data is normalized to obtain the preprocessed sample data set.

[0050] Optionally, the initial TabNet-based multi-task learning model includes:

[0051] Feature selection module, TabNet sharing layer, splitting layer, and adaptive loss module;

[0052] Wherein, the feature selection module is used to perform feature selection on input data;

[0053] Wherein, the TabNet shared layer is used to extract high-level features;

[0054] Wherein, the splitting layer includes two fully connected layers and an MLP classification head;

[0055] The adaptive loss module is used to balance the training progress of different tasks.

[0056] Optionally, the training unit is used to:

[0057] Inputting the preprocessed sample data set into the initial TabNet-based multi-task learning model, performing feature selection through a feature selection module, and obtaining a selected feature set;

[0058] The selected feature set is input into the TabNet shared layer to extract high-level features and obtain the feature representation of high-level features;

[0059] The feature representation of the high-level features is input into the splitting layer, and the calculation is performed through two fully connected layers and the MLP classification head. The calculation results are output, and task matching is performed based on the calculation results to obtain the prediction results of the quality defects of the casting;

[0060] According to the prediction results of ingot quality defects, the initial TabNet-based multi-task learning model is trained through the adaptive loss module to obtain a trained TabNet-based multi-task learning model.

[0061] Optionally, the preprocessed sample data set is input into the initial TabNet-based multi-task learning model, and feature selection is performed through a feature selection module to obtain a selected feature set, including:

[0062] Inputting the preprocessed sample data set into the initial TabNet-based multi-task learning model, using a random forest algorithm for calculation to obtain a feature contribution vector;

[0063] According to the feature contribution vector, the prior knowledge vector and threshold are set through the mechanism model to calculate the feature weight;

[0064] Calculate according to the feature weights and the preprocessed sample data set to obtain

[0065] Choose a good feature set.

[0066] Optionally, the training of the initial TabNet-based multi-task learning model through an adaptive loss module according to the prediction result of the quality defects of the casting includes:

[0067] Initialize weights;

[0068] According to the initialization weights, calculate the total loss in a single round; according to the total loss in a single round, calculate the gradient norm of a single task to the overall task; according to the gradient norm of a single task to the overall task, calculate the relative direction training speed of a single task;

[0069] According to the gradient norm of a single task to the overall task, calculate the average gradient norm of all tasks;

[0070] Calculate the gradient loss in a single round based on the average gradient norm of all tasks in a single round, the gradient norm of a single task to the overall task, and the relative direction training speed of a single task;

[0071] Calculate the weights in a single round according to the gradient loss in a single round, and obtain the weights for the next round; back-propagate the update parameters according to the total loss in a single round, and update the weights by normalizing the weights for the next round;

[0072] Determine whether the current number of training rounds reaches the preset value and the total loss descent gradient is less than the set threshold. If so, stop training and output the trained model; if not, return to step S441 to continue training until the preset stop condition is met.

[0073] Optionally, the prediction unit is used to:

[0074] Inputting the preprocessed continuous casting production data to be predicted into the trained multi-task learning model based on TabNet, calculating the probability of internal defect levels in the data set, and obtaining the probability value of the defect level;

[0075] The defect level with the highest probability value is selected as the final casting billet quality prediction result.

[0076] On the other hand, a multi-task ingot quality defect prediction device based on adaptive loss is provided, and the multi-task ingot quality defect prediction device based on adaptive loss comprises: a processor; a memory, wherein computer-readable instructions are stored on the memory, and when the computer-readable instructions are executed by the processor, any one of the multi-task ingot quality defect prediction methods based on adaptive loss is implemented.

[0077] On the other hand, a computer-readable storage medium is provided, in which at least one instruction is stored, and the at least one instruction is loaded and executed by a processor to implement any one of the above-mentioned multi-task ingot quality defect prediction methods based on adaptive loss.

[0078] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0079] The embodiment of the present invention first obtains a continuous casting production sample data set; preprocesses the continuous casting production sample data set to obtain a preprocessed sample data set; secondly, constructs an initial TabNet-based multi-task learning model; trains the initial TabNet-based multi-task learning model according to the preprocessed sample data set to obtain a trained TabNet-based multi-task learning model; obtains the continuous casting production data to be predicted; finally, preprocesses the continuous casting production data to be predicted to obtain the preprocessed continuous casting production data to be predicted; inputs the preprocessed continuous casting production data to be predicted into the trained TabNet-based multi-task learning model to obtain the continuous casting billet internal quality prediction result.

[0080] The embodiment of the present invention adopts a feature selection method based on prior knowledge, designs a feature selection module, combines prior knowledge with the feature contribution vector obtained by the random forest algorithm, guides feature selection, and makes feature selection more interpretable; the multi-task learning architecture based on TabNet can effectively integrate multiple related tasks, and utilize shared features and dynamic feature selection mechanisms, thereby reducing the training time of the model and improving the generalization ability of the model. An adaptive loss module is designed, which can automatically adjust the weights of different tasks in the total loss according to the training process, effectively optimize the training process of multi-task learning, improve the performance of the model on each task, and enhance the overall learning efficiency and robustness; a method for selecting the initial value of multi-task weights is proposed, and the correlation between the target variable and the input feature is used as the basis for allocating weights, so that the initial value of the weight is closer to the optimal solution, the training process is reduced, and the overfitting phenomenon is effectively avoided. The present invention can improve the performance of predicting internal quality defects of continuous casting billets. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0082] Figure 1 It is a schematic diagram of the overall structure of a multi-task casting quality defect prediction method based on adaptive loss provided by an embodiment of the present invention;

[0083] Figure 2 It is a flow chart of a multi-task casting quality defect prediction method based on adaptive loss provided by an embodiment of the present invention;

[0084] Figure 3 is a schematic diagram of feature selection provided by an embodiment of the present invention;

[0085] Figure 4 It is a schematic diagram of the TabNet structure provided by an embodiment of the present invention;

[0086] Figure 5 is a schematic diagram of weight updating provided by an embodiment of the present invention;

[0087] Figure 6 It is a block diagram of a multi-task casting quality defect prediction device based on adaptive loss provided by an embodiment of the present invention;

[0088] Figure 7 It is a structural schematic diagram of a multi-task casting quality defect prediction device based on adaptive loss provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0089] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0090] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0091] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0092] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0093] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0094] The embodiment of the present invention provides a method for predicting quality defects of a multi-task casting billet based on adaptive loss, which can be implemented by a multi-task casting billet quality defect prediction device based on adaptive loss, and the multi-task casting billet quality defect prediction device based on adaptive loss can be a terminal or a server. Figure 1 FIG. 1 is a schematic diagram of the overall structure of a multi-task casting quality defect prediction method based on adaptive loss provided by an embodiment of the present invention; FIG. Figure 2 The flowchart of the multi-task casting quality defect prediction method based on adaptive loss is shown, and the processing flow of the method may include the following steps:

[0095] S1. Obtain a continuous casting production sample data set.

[0096] S2. Preprocess the continuous casting production sample data set to obtain a preprocessed sample data set.

[0097] Optionally, the specific implementation process of S2 may include S21-S22:

[0098] S21. Using a data mining algorithm, remove missing values ​​and outliers in the continuous casting production sample data set to obtain a continuous casting production sample data set after the removal process;

[0099] The data mining algorithms may include: decision tree, random forest algorithm, support vector machine, K-nearest neighbor algorithm and principal component analysis algorithm, etc., and any one of them may be selected.

[0100] In a feasible implementation, a box plot is used to detect outliers in the continuous casting production sample data set and delete them; a small portion of missing data is filled using a group filling method, and a portion with a large number of missing data is deleted.

[0101] Among them, the box plot and group filling method are conventional technical means, and the present invention will not be further elaborated.

[0102] S22. Use ONE-HOT coding to digitize the processed continuous casting production sample data set, normalize the digitized data, and obtain a preprocessed sample data set.

[0103] Among them, normalizing the data after digitization can remove the interference of dimension on the prediction of the multi-task learning model based on TabNet.

[0104] Among them, ONE-HOT coding is a conventional technical means and will not be further elaborated in the present invention.

[0105] S3. Build an initial TabNet-based multi-task learning model.

[0106] Among them, TabNet's adaptive attention mechanism can efficiently capture feature interactions, extract potential features of data, provide feature importance evaluation, and enhance the interpretability of TabNet-based multi-task learning models. TabNet processes sparse data quickly and accurately, and can scale well on large-scale data sets. TabNet's fewer hyperparameter requirements make model tuning simple and suitable for building multi-task learning architectures.

[0107] Optionally, the initial TabNet-based multi-task learning model includes:

[0108] Feature selection module, TabNet sharing layer, splitting layer, and adaptive loss module;

[0109] Among them, the feature selection module is used to select features of input data;

[0110] Among them, the TabNet shared layer is used to extract high-level features;

[0111] Among them, the split layer includes two fully connected layers and an MLP classification head;

[0112] Among them, the adaptive loss module is used to balance the training progress of different tasks.

[0113] S4. Train the initial TabNet-based multi-task learning model according to the preprocessed sample data set to obtain a trained TabNet-based multi-task learning model.

[0114] The ratio of training set to test set is 8:2.

[0115] In a feasible implementation manner, the embodiment of the present invention adopts the Pytorch1.13.1 learning framework, uses the Adam optimizer, and sets the initial learning rate to 0.02.

[0116] Optionally, the specific implementation process of S4 may include S41-S44:

[0117] S41, inputting the preprocessed sample data set into the initial TabNet-based multi-task learning model, performing feature selection through the feature selection module, and obtaining a selected feature set;

[0118] Among them, Figure 3 The figure is a schematic diagram of feature selection provided by an embodiment of the present invention; input features are selected through a feature selection module.

[0119] Optionally, the specific implementation process of S41 may include S411-S413:

[0120] S411, input the preprocessed sample data set into the initial TabNet-based multi-task learning model, use the random forest algorithm to perform calculations, and obtain a feature contribution vector;

[0121] Among them, the parameters of random forest are searched using grid search method, and the optimal results are retained.

[0122] Among them, the random forest algorithm is a conventional technical means and will not be further elaborated in the present invention.

[0123] S412, according to the feature contribution vector, setting the prior knowledge vector and the threshold through the mechanism model, and calculating the feature weight;

[0124] S413: Calculate based on feature weights and preprocessed sample data sets to obtain

[0125] Choose a good feature set.

[0126] In a feasible implementation, the process of obtaining the selected feature set can be expressed by the following formula (1) and formula (2):

[0127] (1)

[0128] (2)

[0129] Among them, FW represents feature weight; FI represents feature contribution vector; PK represents prior knowledge vector; It indicates the set threshold value, and the threshold value set in the present invention is 15. Input indicates the sample data set after preprocessing; Output indicates the selected feature set.

[0130] In a feasible implementation, prior knowledge is combined with feature contribution vectors to jointly guide feature selection, which can reduce noise and redundant features and improve the interpretability of feature selection.

[0131] S42, inputting the selected feature set into the TabNet shared layer to extract high-level features and obtain feature representation of the high-level features;

[0132] In a feasible implementation, a TabNet sharing layer is set, and a splitting layer of the number of tasks is set after the TabNet sharing layer. Figure 4 : It is a schematic diagram of the TabNet structure provided by an embodiment of the present invention; TabNet first encodes the input tabular data into a feature vector; the feature vector is sequentially input into a plurality of stacked Blocks for processing; wherein, the number of Blocks is set to 3-5, and the present invention sets the number of Blocks to 3; wherein, the output of the previous Block is used as the input of the next Block, and the structure of each Block is the same, and the output of the last Block is input into the MLP classification head to obtain the classification result; wherein, in each Block, the input feature first passes through the mask layer, and the mask layer receives the information of the previous Block and automatically adjusts the selected feature; the selected feature passes through the feature converter, segmentation and activation once, and the subsequent Block continues to calculate, wherein the segmented part of the information is sent to the mask layer of the next Block through an attention converter for feature selection. wherein, the splitting layer uses 2 full connections, matches the number of neurons with the task, predicts the quality of the continuous casting billet, and obtains the prediction results of different quality defects and defect levels.

[0133] S43, inputting the feature representation of the high-level feature into the splitting layer, performing calculations through two fully connected layers and the MLP classification head, outputting the calculation results, performing task matching according to the calculation results, and obtaining the prediction results of the quality defects of the ingot;

[0134] S44. According to the prediction results of the ingot quality defects, the initial TabNet-based multi-task learning model is trained through an adaptive loss module to obtain a trained TabNet-based multi-task learning model.

[0135] Among them, during the training process, the adaptive loss module will update the weights at any time according to the gradient of the loss decrease of different tasks.

[0136] In a feasible implementation, the adaptive loss module dynamically adjusts the weight of a single task loss in the total loss, thereby ensuring that the model adaptively focuses on the current poorly performing task during training, thereby promoting the improvement of overall performance. The adaptive loss module is not only suitable for multi-task learning scenarios, but also for other deep learning tasks that require dynamic adjustment of loss weights.

[0137] Optionally, the specific implementation process of S44 may include S441-S446:

[0138] S441, initializing weights;

[0139] In a feasible implementation, a Pearson correlation analysis is performed on the input features and the output features to obtain the correlation between each single target variable and the input features, and the initial weight is determined according to the weighting of the average correlation, wherein the initialization weight can be expressed by the following formula (3):

[0140] (3)

[0141] in, represents the sum of the absolute values ​​of the correlation between the target variable of the i-th task and the input feature, and n represents the number of single tasks included in multi-task learning. Represents the initialization weights.

[0142] Among them, Pearson correlation analysis is a technical means mastered by those skilled in the art, and the present invention will not be further elaborated here.

[0143] Among them, by analyzing the relationship between tasks, optimizing the setting of initial weights, improving the convergence speed and overall performance of the model, and through correlation analysis, providing appropriate initial weights for different tasks, promoting the effective implementation of the training process.

[0144] S442, calculating the total loss in a single round according to the initialization weight; calculating the gradient norm of a single task to the overall task according to the total loss in a single round; calculating the relative direction training speed of a single task according to the gradient norm of the single task to the overall task;

[0145] In a feasible implementation, in the tth round, the total loss is calculated and can be expressed by the following formula (4):

[0146] (4)

[0147] in, represents the total loss in round t; represents the loss of the i-th task; It represents the weight of the loss of the i-th task in the total loss.

[0148] In a feasible implementation, the gradient norm of the i-th task to the whole can be expressed by the following formula (5):

[0149] (5)

[0150] in, Represents the gradient norm of the i-th task to the whole.

[0151] In a feasible implementation, the relative direction training speed of a single task can be expressed by the following formula (6):

[0152] (6)

[0153] in, represents the loss ratio of task i at step t, Represents the average loss ratio of all tasks in t steps; represents the relative reverse training speed of task i; represents the initial loss of the i-th task.

[0154] S443, calculating the average gradient norm of all tasks according to the gradient norm of a single task to the overall task;

[0155] In a feasible implementation, the average gradient norm of all tasks can be expressed by the following formula (7):

[0156] (7)

[0157] in, represents the average gradient norm of all tasks in round t.

[0158] S444, calculating the gradient loss in a single round according to the average gradient norm of all tasks in a single round, the gradient norm of a single task to the overall task, and the relative direction training speed of a single task;

[0159] In a feasible implementation, the gradient loss in a single round can be expressed by the following formula (8):

[0160] (8)

[0161] in, Represents the gradient loss of the tth round; represents an additional hyperparameter, which is set according to different tasks. Usually, a larger value is selected to force a stronger training rate balance. Set to 1.5.

[0162] S445, calculating the weights in a single round according to the gradient loss in a single round, obtaining the weights of the next round; back-propagating the update parameters according to the total loss in the single round, and updating the weights by normalizing the weights of the next round;

[0163] like Figure 5 It is a weight update schematic diagram provided by an embodiment of the present invention. In a feasible implementation, the partial derivative of the gradient loss of the tth round with respect to the weight of the i-th task in the tth round is calculated to obtain the weight of the i-th task in the t+1th round.

[0164] In a feasible implementation, the weight of the next round is normalized and can be expressed by the following formula (9):

[0165] (9)

[0166] S446: Determine whether the current number of training rounds has reached the preset value and the total loss descent gradient is less than the set threshold. If so, stop training and output the trained model; if not, return to step S441 to continue training until the preset stop condition is met.

[0167] S5. Acquire the continuous casting production data to be predicted; preprocess the continuous casting production data to be predicted to obtain the preprocessed continuous casting production data to be predicted; input the preprocessed continuous casting production data to be predicted into the trained TabNet-based multi-task learning model to obtain the prediction result of the internal quality of the continuous casting billet.

[0168] The continuous casting production data to be predicted is the actual production data of a steel plant during the continuous casting process, and the output features include: the chemical composition of the continuous casting billet, the process parameters and equipment information during the continuous casting process. The two labels of the center porosity and shrinkage of the continuous casting billet are selected as target variables for training to verify the effectiveness of the model.

[0169] Optionally, S5 inputs the preprocessed continuous casting production data to be predicted into the trained multi-task learning model based on TabNet to obtain the prediction result of the internal quality of the continuous casting billet, including:

[0170] S51, inputting the preprocessed continuous casting production data to be predicted into the trained multi-task learning model based on TabNet, calculating the probability of internal defect level in the data set, and obtaining the probability value of the defect level;

[0171] S52. Select the defect level with the highest probability value as the final ingot quality prediction result.

[0172] In a feasible implementation method, the real label of the test set is obtained, compared with the prediction result of the ingot quality, and the accuracy of the multi-task learning model based on TabNet is calculated; the incorrectly predicted samples are analyzed to identify the reasons for the poor performance of the multi-task learning model based on TabNet, including: data quality, feature selection, and model complexity. Through the analysis results, the misclassified samples are counted, and the common points of the misclassified samples are found through k-means clustering. The shortcomings of the model are identified, and the feature selection and model complexity are adjusted according to the shortcomings to further improve the prediction ability and accuracy of the model.

[0173] Among them, the present invention adopts the commonalities of different internal quality defect problems, can share feature representation, reduce redundant modeling, and can effectively reduce training costs and inference time while maintaining prediction accuracy, providing an efficient and feasible technical path for solving the internal quality defect problems of continuous casting billets.

[0174] Among them, building a multi-task learning model based on TabNet has significant advantages, especially in processing tabular data; TabNet is a deep learning model based on the attention mechanism, which can achieve efficient representation learning in the process of feature selection and modeling, and is suitable for multi-task learning scenarios; by applying TabNet to the multi-task learning model, the prediction accuracy and reliability of different internal quality defects can be significantly improved. Building a multi-task learning model based on TabNet can not only provide efficient real-time monitoring, but also provide data support for subsequent quality improvement and control measures, promoting the dual improvement of production efficiency and product quality.

[0175] The embodiment of the present invention first obtains a continuous casting production sample data set; preprocesses the continuous casting production sample data set to obtain a preprocessed sample data set; secondly, constructs an initial TabNet-based multi-task learning model; trains the initial TabNet-based multi-task learning model according to the preprocessed sample data set to obtain a trained TabNet-based multi-task learning model; obtains the continuous casting production data to be predicted; finally, preprocesses the continuous casting production data to be predicted to obtain the preprocessed continuous casting production data to be predicted; inputs the preprocessed continuous casting production data to be predicted into the trained TabNet-based multi-task learning model to obtain the continuous casting billet internal quality prediction result.

[0176] The embodiment of the present invention adopts a feature selection method based on prior knowledge, designs a feature selection module, combines prior knowledge with the feature contribution vector obtained by the random forest algorithm, guides feature selection, and makes feature selection more interpretable; the multi-task learning architecture based on TabNet can effectively integrate multiple related tasks, and utilize shared features and dynamic feature selection mechanisms, thereby reducing the training time of the model and improving the generalization ability of the model. An adaptive loss module is designed, which can automatically adjust the weights of different tasks in the total loss according to the training process, effectively optimize the training process of multi-task learning, improve the performance of the model on each task, and enhance the overall learning efficiency and robustness; a method for selecting the initial value of multi-task weights is proposed, and the correlation between the target variable and the input feature is used as the basis for allocating weights, so that the initial value of the weight is closer to the optimal solution, the training process is reduced, and the overfitting phenomenon is effectively avoided. The present invention can improve the performance of predicting internal quality defects of continuous casting billets.

[0177] Figure 6 The invention is a block diagram of a multi-task casting quality defect prediction device based on adaptive loss according to an exemplary embodiment, wherein the device is used in a multi-task casting quality defect prediction method based on adaptive loss. Figure 6 The device includes an acquisition unit 610, a preprocessing unit 620, a construction unit 630, a training unit 640 and a prediction unit 650. Among them:

[0178] An acquisition unit 610 is used to acquire a continuous casting production sample data set;

[0179] A preprocessing unit 620, configured to preprocess the continuous casting production sample data set to obtain a preprocessed sample data set;

[0180] A construction unit 630, used to construct an initial TabNet-based multi-task learning model;

[0181] A training unit 640 is used to train the initial TabNet-based multi-task learning model according to the preprocessed sample data set to obtain a trained TabNet-based multi-task learning model;

[0182] The prediction unit 650 is used to obtain the continuous casting production data to be predicted; preprocess the continuous casting production data to be predicted to obtain the preprocessed continuous casting production data to be predicted; input the preprocessed continuous casting production data to be predicted into the trained TabNet-based multi-task learning model to obtain the continuous casting billet internal quality prediction result.

[0183] Optionally, the preprocessing unit 620 is used to:

[0184] Using a data mining algorithm to remove missing values ​​and outliers in the continuous casting production sample data set, and obtaining a continuous casting production sample data set after the removal process;

[0185] The processed continuous casting production sample data set is digitized using ONE-HOT coding, and the digitized data is normalized to obtain the preprocessed sample data set.

[0186] Optionally, the initial TabNet-based multi-task learning model includes:

[0187] Feature selection module, TabNet sharing layer, splitting layer, and adaptive loss module;

[0188] Wherein, the feature selection module is used to perform feature selection on input data;

[0189] Wherein, the TabNet shared layer is used to extract high-level features;

[0190] Wherein, the splitting layer includes two fully connected layers and an MLP classification head;

[0191] The adaptive loss module is used to balance the training progress of different tasks.

[0192] Optionally, the training unit 640 is used to:

[0193] Inputting the preprocessed sample data set into the initial TabNet-based multi-task learning model, performing feature selection through a feature selection module, and obtaining a selected feature set;

[0194] The selected feature set is input into the TabNet shared layer to extract high-level features and obtain the feature representation of high-level features;

[0195] The feature representation of the high-level features is input into the splitting layer, and the calculation is performed through two fully connected layers and the MLP classification head. The calculation results are output, and task matching is performed based on the calculation results to obtain the prediction results of the quality defects of the casting;

[0196] According to the prediction results of ingot quality defects, the initial TabNet-based multi-task learning model is trained through the adaptive loss module to obtain a trained TabNet-based multi-task learning model.

[0197] Optionally, the preprocessed sample data set is input into the initial TabNet-based multi-task learning model, and feature selection is performed through a feature selection module to obtain a selected feature set, including:

[0198] Inputting the preprocessed sample data set into the initial TabNet-based multi-task learning model, using a random forest algorithm for calculation to obtain a feature contribution vector;

[0199] According to the feature contribution vector, the prior knowledge vector and threshold are set through the mechanism model to calculate the feature weight;

[0200] Calculate according to the feature weights and the preprocessed sample data set to obtain

[0201] Choose a good feature set.

[0202] Optionally, the training of the initial TabNet-based multi-task learning model through an adaptive loss module according to the prediction result of the quality defects of the casting includes:

[0203] Initialize weights;

[0204] According to the initialization weights, calculate the total loss in a single round; according to the total loss in a single round, calculate the gradient norm of a single task to the overall task; according to the gradient norm of a single task to the overall task, calculate the relative direction training speed of a single task;

[0205] According to the gradient norm of a single task to the overall task, calculate the average gradient norm of all tasks;

[0206] Calculate the gradient loss in a single round based on the average gradient norm of all tasks in a single round, the gradient norm of a single task to the overall task, and the relative direction training speed of a single task;

[0207] Calculate the weights in a single round according to the gradient loss in a single round, and obtain the weights for the next round; back-propagate the update parameters according to the total loss in a single round, and update the weights by normalizing the weights for the next round;

[0208] Determine whether the current number of training rounds reaches the preset value and the total loss descent gradient is less than the set threshold. If so, stop training and output the trained model; if not, return to step S441 to continue training until the preset stop condition is met.

[0209] Optionally, the prediction unit 650 is configured to:

[0210] Inputting the preprocessed continuous casting production data to be predicted into the trained multi-task learning model based on TabNet, calculating the probability of internal defect levels in the data set, and obtaining the probability value of the defect level;

[0211] The defect level with the highest probability value is selected as the final casting billet quality prediction result.

[0212] The embodiment of the present invention first obtains a continuous casting production sample data set; preprocesses the continuous casting production sample data set to obtain a preprocessed sample data set; secondly, constructs an initial TabNet-based multi-task learning model; trains the initial TabNet-based multi-task learning model according to the preprocessed sample data set to obtain a trained TabNet-based multi-task learning model; obtains the continuous casting production data to be predicted; finally, preprocesses the continuous casting production data to be predicted to obtain the preprocessed continuous casting production data to be predicted; inputs the preprocessed continuous casting production data to be predicted into the trained TabNet-based multi-task learning model to obtain the continuous casting billet internal quality prediction result.

[0213] The embodiment of the present invention adopts a feature selection method based on prior knowledge, designs a feature selection module, combines prior knowledge with the feature contribution vector obtained by the random forest algorithm, guides feature selection, and makes feature selection more interpretable; the multi-task learning architecture based on TabNet can effectively integrate multiple related tasks, and utilize shared features and dynamic feature selection mechanisms, thereby reducing the training time of the model and improving the generalization ability of the model. An adaptive loss module is designed, which can automatically adjust the weights of different tasks in the total loss according to the training process, effectively optimize the training process of multi-task learning, improve the performance of the model on each task, and enhance the overall learning efficiency and robustness; a method for selecting the initial value of multi-task weights is proposed, and the correlation between the target variable and the input feature is used as the basis for allocating weights, so that the initial value of the weight is closer to the optimal solution, the training process is reduced, and the overfitting phenomenon is effectively avoided. The present invention can improve the performance of predicting internal quality defects of continuous casting billets.

[0214] Figure 7 is a schematic diagram of the structure of a multi-task casting quality defect prediction device based on adaptive loss provided by an embodiment of the present invention, such as Figure 7 As shown, the multi-task casting quality defect prediction device based on adaptive loss may include the above Figure 6 The multi-task casting quality defect prediction device based on adaptive loss is shown. Optionally, the multi-task casting quality defect prediction device based on adaptive loss 710 may include a first processor 2001.

[0215] Optionally, the adaptive loss-based multi-task casting quality defect prediction device 710 may further include a memory 2002 and a transceiver 2003 .

[0216] The first processor 2001, the memory 2002 and the transceiver 2003 may be connected via a communication bus.

[0217] Combine the following Figure 7The components of the adaptive loss-based multi-task casting quality defect prediction device 710 are specifically introduced as follows:

[0218] The first processor 2001 is the control center of the multi-task casting quality defect prediction device 710 based on adaptive loss, which can be a processor or a general term for multiple processing elements. For example, the first processor 2001 is one or more central processing units (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiment of the present invention, such as one or more microprocessors (digital signal processor, DSP), or one or more field programmable gate arrays (field programmable gate array, FPGA).

[0219] Optionally, the first processor 2001 can perform various functions of the adaptive loss-based multi-task casting quality defect prediction device 710 by running or executing a software program stored in the memory 2002 and calling data stored in the memory 2002.

[0220] In a specific implementation, as an embodiment, the first processor 2001 may include one or more CPUs, such as Figure 7 CPU0 and CPU1 are shown in FIG.

[0221] In a specific implementation, as an embodiment, the multi-task casting quality defect prediction device 710 based on adaptive loss may also include multiple processors, such as Figure 7 The first processor 2001 and the second processor 2004 are shown in FIG. Each of these processors may be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0222] The memory 2002 is used to store the software program for executing the solution of the present invention, and is controlled to be executed by the first processor 2001. The specific implementation method can refer to the above method embodiment, which will not be repeated here.

[0223] Optionally, the memory 2002 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2002 may be integrated with the first processor 2001, or may exist independently, and may be accessed through the interface circuit ( Figure 7 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0224] The transceiver 2003 is used to communicate with a network device or a terminal device.

[0225] Optionally, the transceiver 2003 may include a receiver and a transmitter ( Figure 7 The receiver is used to implement a receiving function, and the transmitter is used to implement a sending function.

[0226] Optionally, the transceiver 2003 may be integrated with the first processor 2001, or may exist independently, and communicate with the first processor 2001 through the interface circuit ( Figure 7 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0227] It should be noted that Figure 7 The structure of the adaptive loss-based multi-task ingot quality defect prediction device 710 does not constitute a limitation on the router, and the actual knowledge structure recognition device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0228] In addition, the technical effects of the multi-task ingot quality defect prediction device 710 based on adaptive loss can refer to the technical effects of the multi-task ingot quality defect prediction method based on adaptive loss described in the above method embodiment, which will not be repeated here.

[0229] It should be understood that the first processor 2001 in the embodiment of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0230] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0231] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0232] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0233] In the present invention, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" 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, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0234] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0235] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0236] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0237] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, 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 through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0238] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0239] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0240] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0241] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A multi-task casting quality defect prediction method based on adaptive loss, characterized in that: The method comprises: S1. Obtain a continuous casting production sample data set; S2, preprocessing the continuous casting production sample data set to obtain a preprocessed sample data set; S3. Build an initial TabNet-based multi-task learning model; S4. Training the initial TabNet-based multi-task learning model according to the preprocessed sample data set to obtain a trained TabNet-based multi-task learning model; S5. Acquire the continuous casting production data to be predicted; preprocess the continuous casting production data to be predicted to obtain the preprocessed continuous casting production data to be predicted; input the preprocessed continuous casting production data to be predicted into the trained TabNet-based multi-task learning model to obtain the continuous casting billet internal quality prediction result.

2. The method for predicting quality defects of casting billets based on multi-task adaptive loss according to claim 1, characterized in that: The step S2 of preprocessing the continuous casting production sample data set to obtain a preprocessed sample data set includes: S21, using a data mining algorithm to remove missing values ​​and outliers in the continuous casting production sample data set to obtain a continuous casting production sample data set after the removal process; S22. Use ONE-HOT coding to digitize the processed continuous casting production sample data set, normalize the digitized data, and obtain a preprocessed sample data set.

3. The method for predicting quality defects of casting billets based on multi-task adaptive loss according to claim 1, characterized in that: The initial TabNet-based multi-task learning model includes: Feature selection module, TabNet sharing layer, splitting layer, and adaptive loss module; Wherein, the feature selection module is used to perform feature selection on input data; Wherein, the TabNet shared layer is used to extract high-level features; Wherein, the splitting layer includes two fully connected layers and an MLP classification head; The adaptive loss module is used to balance the training progress of different tasks.

4. The method for predicting quality defects of casting billets based on multi-task adaptive loss according to claim 3 is characterized in that: The step S4 of training the initial TabNet-based multi-task learning model according to the preprocessed sample data set to obtain a trained TabNet-based multi-task learning model includes: S41, inputting the preprocessed sample data set into the initial TabNet-based multi-task learning model, performing feature selection through a feature selection module, and obtaining a selected feature set; S42, inputting the selected feature set into the TabNet shared layer to extract high-level features and obtain feature representation of the high-level features; S43, inputting the feature representation of the high-level feature into the splitting layer, performing calculations through two fully connected layers and the MLP classification head, outputting the calculation results, performing task matching according to the calculation results, and obtaining the prediction results of the quality defects of the ingot; S44. According to the prediction results of the ingot quality defects, the initial TabNet-based multi-task learning model is trained through an adaptive loss module to obtain a trained TabNet-based multi-task learning model.

5. The method for predicting quality defects of casting billets based on multi-task adaptive loss according to claim 4, characterized in that: The step S41 of inputting the preprocessed sample data set into the initial TabNet-based multi-task learning model, performing feature selection through a feature selection module, and obtaining a selected feature set includes: S411, inputting the preprocessed sample data set into the initial TabNet-based multi-task learning model, using a random forest algorithm for calculation, and obtaining a feature contribution vector; S412, according to the feature contribution vector, setting the prior knowledge vector and the threshold through the mechanism model, and calculating the feature weight; S413: Calculate based on feature weights and preprocessed sample data sets to obtain Choose a good feature set.

6. The method for predicting quality defects of casting billets based on multi-task adaptive loss according to claim 4, characterized in that: The S44 trains the initial TabNet-based multi-task learning model through an adaptive loss module according to the prediction results of the quality defects of the casting, including: S441, initializing weights; S442, calculating the total loss in a single round according to the initialization weight; calculating the gradient norm of a single task to the overall task according to the total loss in a single round; calculating the relative direction training speed of a single task according to the gradient norm of the single task to the overall task; S443, calculating the average gradient norm of all tasks according to the gradient norm of a single task to the overall task; S444, calculating the gradient loss in a single round according to the average gradient norm of all tasks in a single round, the gradient norm of a single task to the overall task, and the relative direction training speed of a single task; S445, calculating the weights in a single round according to the gradient loss in a single round, obtaining the weights of the next round; back-propagating the update parameters according to the total loss in the single round, and updating the weights by normalizing the weights of the next round; S446: Determine whether the current number of training rounds has reached the preset value and the total loss descent gradient is less than the set threshold. If so, stop training and output the trained model; if not, return to step S441 to continue training until the preset stop condition is met.

7. The multi-task casting quality defect prediction method based on adaptive loss according to claim 1 is characterized in that: The step S5 inputs the pre-processed continuous casting production data to be predicted into the trained multi-task learning model based on TabNet to obtain the prediction result of the internal quality of the continuous casting billet, including: S51, inputting the preprocessed continuous casting production data to be predicted into the trained multi-task learning model based on TabNet, calculating the probability of internal defect level in the data set, and obtaining the probability value of the defect level; S52. Select the defect level with the highest probability value as the final ingot quality prediction result.

8. A multi-task casting quality defect prediction device based on adaptive loss, the multi-task casting quality defect prediction device based on adaptive loss is used to implement the multi-task casting quality defect prediction method based on adaptive loss as claimed in any one of claims 1 to 7, characterized in that: The device comprises: An acquisition unit, used for acquiring a continuous casting production sample data set; A preprocessing unit, used for preprocessing the continuous casting production sample data set to obtain a preprocessed sample data set; Construction unit, used to build the initial TabNet-based multi-task learning model; A training unit, used to train the initial TabNet-based multi-task learning model according to the preprocessed sample data set to obtain a trained TabNet-based multi-task learning model; The prediction unit is used to obtain the continuous casting production data to be predicted; preprocess the continuous casting production data to be predicted to obtain the preprocessed continuous casting production data to be predicted; input the preprocessed continuous casting production data to be predicted into the trained TabNet-based multi-task learning model to obtain the continuous casting billet internal quality prediction result.

9. A multi-task casting quality defect prediction device based on adaptive loss, characterized in that: The multi-task casting quality defect prediction device based on adaptive loss includes: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program codes, which can be called by a processor to execute the method according to any one of claims 1 to 7.

Images