Method and system for predicting mechanical properties of hot-rolled strip steel

Abstract

The invention belongs to the technical field of material performance prediction, and particularly relates to a hot-rolled strip steel mechanical performance prediction method and system based on multi-dimensional feature similarity and integrated instant learning, and the method comprises the steps: firstly obtaining a feature weight through employing a multi-feature quantification method, and introducing a time weight on the basis of the feature weight; a proper modeling sample is selected by comprehensively considering similarity measurement of input characteristics and time factors, and the limitation of single similarity measurement modeling is solved; secondly, a cumulative similarity factor is introduced, and it is ensured that an optimal modeling data sample set is screened out for local modeling; and finally, the output result of each local model is integrated through an integrated learning framework, so that the dependency of the prediction result on a single model is reduced, the model performance is further remarkably enhanced, and the prediction precision and generalization of the mechanical property of the hot-rolled strip steel under the complex production working condition are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of material property prediction, specifically to a method and system for predicting the mechanical properties of hot-rolled strip steel, and more specifically to a method and system for predicting the mechanical properties of hot-rolled strip steel based on multi-dimensional feature similarity and integrated instant learning. Background Art

[0002] Steel plays a cornerstone role in many industrial fields due to its excellent mechanical properties and advantages in economic and environmental sustainability. As a key indicator for evaluating the quality of hot-rolled strip steel, accurate prediction of mechanical properties is of great significance for reducing the frequency of sample testing, ensuring product quality stability, optimizing steel composition, and promoting the development of new products.

[0003] At present, mechanical property prediction methods are mainly divided into two categories: based on traditional physical metallurgy models and based on artificial intelligence models. The construction of traditional physical metallurgy models relies on laboratory test data and is limited by professional knowledge, cost, time and applicability of specific steel grades. Although the artificial intelligence-based models benefit from the development of data collection and management technology, they are still insufficient in prediction accuracy and adaptability in the face of the complex strong coupling and nonlinear relationship between the chemical composition, crystal phase structure and mechanical properties of hot-rolled products.

[0004] By reviewing relevant literature, it can be found that data-driven prediction models for the mechanical properties of hot-rolled strip are divided into two categories: one is global modeling; the other is just-in-time learning (JITL), also known as local modeling. Global modeling relies on a model trained offline from a historical data set once and expects it to provide services for a long time. Although this method has advantages in operating efficiency, it has limitations in dealing with the time-varying nature of industrial data, making it difficult for the model to continuously and stably adapt to the dynamic changes of the industrial process. In addition, global modeling is insufficient in capturing process variable information, which directly affects the prediction ability of the model. In contrast, JITL has attracted widespread attention in the industry because of its ability to deal with specific problems. The JITL model has three core features: first, it starts to build a model when it receives a query request; second, it evaluates the similarity between the sample to be tested and the historical sample based on similarity, and selects relevant samples based on similarity to build a local model; finally, after the prediction is completed, the model will be discarded to improve adaptability and accuracy.

[0005] However, the prediction accuracy of the JITL model is highly dependent on the modeling samples selected by the similarity measurement method. In sample analysis, multiple similarity dimensions are common, and these dimensions may have different effects on the similarity relationship of different samples. Generally, a single similarity measurement method may not be sufficient to fully capture the true similarity structure between samples. Especially in the prediction of mechanical properties of hot-rolled strip, the multidimensional similarity relationship between input and output features makes it difficult for existing technologies to accurately select relevant samples, which is a prominent technical problem in this field. Summary of the invention

[0006] In order to solve the problems existing in the prior art, the main purpose of the present invention is to propose a method and system for predicting the mechanical properties of hot-rolled strip based on multi-dimensional feature similarity and integrated real-time learning.

[0007] According to one aspect of the present invention, the present invention provides the following technical solution: A method for predicting mechanical properties of hot-rolled strip steel based on multi-dimensional feature similarity and integrated real-time learning comprises the following steps: S1. Collect and pre-process the hot-rolled strip production data set to obtain the hot-rolled production process parameters X[X 1 , X 2 , …,X n ] as the independent variable, and the hot-rolled strip mechanical properties Y[Y 1 , Y 2 , …, Y n ] is the dependent variable, and the production data of each roll of hot-rolled strip steel is taken as one data group to construct the historical data set of the prediction model; S2. For the historical data set, the feature quantification method A is used to calculate the hot rolling production process parameters X[X 1 , X 2 , …,X n ]Effects of hot rolled strip mechanical properties Y[Y 1 , Y 2 , …, Y n ]; at the same time, each data group is sorted according to the corresponding hot-rolled strip production time and assigned a time weight; S3. Calculate the feature-time multi-dimensional similarity measurement value between the sample to be tested and the samples in the historical data set according to the feature weight and the time weight, select multiple samples with high similarity to the sample to be tested from the historical data set by accumulating similarity factors, and use them as training data sets to build a local model based on real-time learning, and output the prediction results of the mechanical properties of hot-rolled strip based on feature quantification method A; S4, using at least one other feature quantization method (feature quantization method B, C, D...) different from feature quantization method A, and repeating steps S2-S3; S5. An integrated learning strategy is adopted to perform weighted integration on the prediction results of multiple local models obtained under different feature quantification methods, and the prediction results of the mechanical properties of hot-rolled strip based on multi-dimensional feature weighted similarity and integrated real-time learning are output as the final output results of the model.

[0008] As a preferred solution of the hot-rolled strip mechanical property prediction method based on multi-dimensional feature similarity and integrated real-time learning described in the present invention, wherein: in the step S1, data that meets one of the following three conditions in the data set preprocessing process is regarded as abnormal data: (1) The measured value is outside the predetermined range; (2) Data is missing; (3) The 3σ principle is not met.

[0009] As a preferred solution of the hot-rolled strip mechanical property prediction method based on multi-dimensional feature similarity and integrated real-time learning described in the present invention, wherein: in the step S1, the hot rolling production process parameter X[X 1 , X 2 , …, X n ] including but not limited to chemical element content, furnace temperature, rough rolling start temperature, rough rolling start thickness, rough rolling final thickness, finishing rolling start temperature, finishing rolling final temperature, finishing rolling final thickness, coiling temperature, etc.; hot rolled strip mechanical properties Y[Y 1 , Y 2 , …, Y n ]Including but not limited to tensile strength, yield strength, elongation, etc.

[0010] As a preferred solution of the hot-rolled strip mechanical properties prediction method based on multi-dimensional feature similarity and integrated instant learning described in the present invention, wherein: in the step S2, the feature quantification method includes but is not limited to Pearson correlation analysis (PPC), Spearman correlation analysis (SCC), mutual information correlation analysis (MICC), mean decreasing accuracy (MDA), etc.

[0011] As a preferred solution of the hot-rolled strip mechanical property prediction method based on multi-dimensional feature similarity and integrated real-time learning described in the present invention, wherein: in the step S3, the feature-time multi-dimensional similarity metric value calculation formula of the sample to be tested and the sample in the historical data set is: in: Represents the feature-time multidimensional similarity measure between the sample to be tested and the historical samples, For the j A historical sample, is the sample to be tested, is the feature weight similarity measure between the sample to be tested and the historical sample, is the time weight of historical samples, It is an adjustable parameter.

[0012] As a preferred solution of the hot-rolled strip mechanical property prediction method based on multi-dimensional feature similarity and integrated real-time learning described in the present invention, wherein: in the step S3, the feature weight similarity metric value is introduced by taking Pearson correlation analysis (PPC) as an example Calculation method: The formula for calculating feature weights using Pearson correlation analysis (PPC) is: in: n represents the number of samples, and Respectively represent the sample dimensions d The input and output observations of and Represent the sample input mean and output mean respectively; Combining the feature weights calculated by Pearson correlation analysis (PCC) with the traditional Euclidean similarity, the formula for calculating the feature weight similarity measure is: in: is the jth historical sample; For test samples; d is the dimension of the input data; represents the input feature weight; is the similarity between the test sample and the historical sample.

[0013] As a preferred solution of the hot-rolled strip mechanical property prediction method based on multi-dimensional feature similarity and integrated real-time learning described in the present invention, wherein: in step S3, the cumulative similarity factor L The calculation formula is: Among them: the numerator represents the sum of the similarity contributions of the first n groups of samples with higher similarity, and the denominator represents the sum of the similarities between the test sample and all historical samples.

[0014] As a preferred solution of the method for predicting the mechanical properties of hot-rolled strip based on multidimensional feature similarity and integrated instant learning described in the present invention, wherein: in the step S3, the local model construction method based on instant learning includes but is not limited to multivariate linear regression (MLR), partial least squares regression (PLS), support vector machine (SVM), etc.

[0015] As a preferred solution of the hot-rolled strip mechanical property prediction method based on multi-dimensional feature similarity and integrated real-time learning described in the present invention, wherein: in the step S5, the weighted integration of the prediction results of multiple local models depends on the performance of each local model on the verification data set, r The mean absolute error of the models ( MAE ) is calculated as: in: and Respectively i The observed and predicted values ​​of the test samples, m is the number of test set samples.

[0016] As a preferred solution of the hot-rolled strip mechanical property prediction method based on multi-dimensional feature similarity and integrated real-time learning described in the present invention, wherein: in step S5, each model MAE Definition r The weight of the sub-model output result is: Among them: The numerator represents the r The sub-model e is the base number, MAE is the function value of the exponential, and the denominator represents all sub-models with e is the base number, MAE is the sum of the function values ​​of the exponentials.

[0017] According to one aspect of the present invention, the present invention provides the following technical solution: A hot-rolled strip mechanical properties prediction system based on multi-dimensional feature similarity and integrated real-time learning, comprising: Production data set collection and preprocessing module: collect hot-rolled strip production data sets and preprocess them based on hot-rolled production process parameters X[X 1 , X 2 , …, X n ] as the independent variable, and the hot-rolled strip mechanical properties Y[Y 1 , Y 2 , …, Y n ] is the dependent variable, and the production data of each coil of hot-rolled strip steel is taken as one data group to construct the historical data set of the prediction model; Weight determination module: For the historical data set, the feature quantification method A is used to calculate the hot rolling production process parameters X[X 1 , X 2 , …, X n ]Effects of hot rolled strip mechanical properties Y[Y 1 , Y 2 , …, Yn ]; at the same time, each data group is sorted according to the corresponding hot-rolled strip production time and assigned a time weight; Multi-dimensional feature similarity calculation and local model construction module: According to the feature weight and time weight, the feature-time multi-dimensional similarity measurement value of the sample to be tested and the samples in the historical data set is calculated, and multiple samples with high similarity to the sample to be tested are selected from the historical data set by accumulating similarity factors. They are used as training data sets to build a local model based on real-time learning, and the prediction results of the mechanical properties of hot-rolled strip based on feature quantification method A are output; at least one other feature quantification method (feature quantification method B, C, D...) different from feature quantification method A is used to output the prediction results of the mechanical properties of hot-rolled strip based on it; Ensemble learning final result output module: adopts ensemble learning strategy to weightedly integrate the prediction results of multiple local models obtained under different feature quantification methods, and outputs the prediction results of hot-rolled strip mechanical properties based on multi-dimensional feature weighted similarity and ensemble instant learning as the final output result of the model.

[0018] The beneficial effects of the present invention are as follows: The present invention proposes a method and system for predicting the mechanical properties of hot-rolled strip based on multi-dimensional feature similarity and integrated instant learning. Firstly, a plurality of feature quantification methods are used to obtain feature weights, and time weights are introduced on the basis of feature weights. A similarity metric that comprehensively considers input features and time factors is constructed to select suitable modeling samples, thereby solving the limitations of modeling with a single similarity metric. Secondly, a cumulative similarity factor is introduced to ensure that the optimal modeling data sample set is screened out for local modeling. Finally, an integrated learning framework is used to integrate the output results of each local model, thereby reducing the dependence of the prediction results on a single model, thereby significantly enhancing the model performance and improving the prediction accuracy and generalization of the mechanical properties of hot-rolled strip under complex production conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 the structures shown in these drawings without paying creative work.

[0020] Figure 1 It is a schematic diagram of a framework for predicting the mechanical properties of hot-rolled strip based on multi-dimensional feature weighting and integrated instant learning provided by an embodiment of the present invention.

[0021] Figure 2 It is a schematic diagram of data set division in a simulated hot rolling process provided by an embodiment of the present invention.

[0022] Figure 3 It is a schematic diagram of the MDA principle provided by an embodiment of the present invention.

[0023] Figure 4 It is a feature importance diagram under different feature analysis methods provided by an embodiment of the present invention.

[0024] Figure 5 It is a prediction error graph of each local model under different cumulative similarity factors provided by an embodiment of the present invention.

[0025] Figure 6 It is a comparison chart of prediction results of different models provided by an embodiment of the present invention.

[0026] Figure 7 It is a scatter plot comparison of prediction results of different models provided by an embodiment of the present invention.

[0027] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0028] The following will be described clearly and completely in conjunction with the technical solutions in the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] The present invention proposes a method and system for predicting the mechanical properties of hot-rolled strip based on multi-dimensional feature similarity and integrated real-time learning. The technical solution of the present invention is further explained below in conjunction with specific embodiments.

[0030] Example Figure 1 : is a schematic diagram of a framework for hot-rolled strip mechanical property prediction based on multi-dimensional feature weighting and integrated real-time learning in this embodiment, comprising the following steps: S1. Collect and pre-process the hot-rolled strip production data set to obtain the hot-rolled production process parameters X[X 1 , X 2 , …,X n ] as the independent variable, and the hot-rolled strip mechanical properties Y[Y 1 , Y 2 , …, Y n ] is the dependent variable, and the production data of each roll of hot-rolled strip steel is taken as one data group to construct the historical data set of the prediction model; Hot rolling production process parameters X[X 1 , X 2 , …, Xn ]Including chemical element (C, Si, Mn, P, S, Cu, Ni, Cr, V, Al, N, Nb, Ti) content, furnace temperature (TT), rough rolling start temperature (RRST), rough rolling start thickness (TRR), rough rolling final thickness (FTRR), finishing rolling start temperature (FRST), finishing rolling final temperature (FDT), finishing rolling final thickness (THK), coiling temperature (CT), hot rolled strip mechanical properties Y[Y 1 , Y 2 , …, Y n ] including tensile strength (TS), yield strength (YS) and elongation (EL); the data pre-processing method considers that the measurement data that meets one of the following three conditions are regarded as abnormal data: (1) the measurement value exceeds the predetermined range; (2) the data is missing; (3) the 3σ principle is not met; in order to more accurately simulate and reflect the actual production process, a data set division method is adopted, such as Figure 2 As shown in the figure, the original data is divided into three parts according to the production time series: (1) historical data set: composed of the earliest production data; (2) intermediate data set: lagged samples caused by objective factors in the actual production process; (3) the remaining part is the test sample set.

[0031] After data preprocessing, 10,000 samples were retained. The statistical distribution of each parameter in the data set is shown in Table 1: Table 1 Statistical distribution of parameters in the sample The samples are sorted according to the production time sequence and divided into historical data sets, intermediate data sets and test sets. The three data sets contain 5970, 30 and 4000 samples respectively. In order to eliminate the influence of different dimensional differences, the Min-Max normalization method is used to normalize the original data: in: and Respectively represent the maximum and minimum values ​​of the d-th feature, is the original value of the dth feature of the ith sample, Represents the normalized value of the dth feature of the ith sample.

[0032] S2. For the historical data set, Pearson correlation analysis (PCC) is used to calculate the characteristic weights of the influence of hot rolling production process parameters on the mechanical properties of hot rolled strip; at the same time, each data group is sorted according to the corresponding hot rolled strip production time and assigned a time weight; S3. Calculate the feature-time multi-dimensional similarity metric value between the sample to be tested and the samples in the historical data set according to the feature weight and time weight, and adaptively select multiple samples with high similarity to the sample to be tested from the historical data set as training data sets to build a local model based on real-time learning, and output the prediction results of the mechanical properties of hot-rolled strip steel based on Pearson correlation analysis (PCC); S4, respectively using Spearman correlation analysis (SCC), mutual information correlation analysis (MICC), mean decreasing accuracy (MDA, the principle diagram is as follows Figure 3 Steps S2-S3 are repeated using feature quantization methods such as those shown in FIG.

[0033] In view of the fact that the selection of fixed data length may be limited by the adaptability of the model to different industrial environments, the present invention adopts an adaptive method based on cumulative similarity factor to dynamically determine the number of similar samples to make the selected samples more reasonable; the calculation formula of cumulative similarity factor L is: Among them: the denominator represents the sum of similarities between the test sample and all historical samples, and the numerator represents the previous sample with the greatest similarity. n The sum of similarity contributions of group samples.

[0034] To evaluate the model performance, the root mean square error ( RMSE ), mean absolute error ( MAE ) and mean absolute percentage error ( MAPE ), which is expressed as follows: in: m is the number of test samples, For the i The true value of the samples, for i The predicted value of a sample.

[0035] Figure 4 Taking tensile strength (TS) as an output feature, the absolute values ​​of feature weights assigned by feature quantification methods such as Pearson correlation analysis (PPC), Spearman correlation analysis (SCC), mutual information correlation analysis (MICC), and mean decreasing accuracy (MDA) are ranked, and only the values ​​of the first six features are displayed; Figure 4 It can be seen that different feature quantization methods assign different weight values ​​to the same feature.

[0036] Figure 5 Shows the difference LThe prediction performance of each local model for different mechanical properties under different values ​​( Figure 5 (a) represents tensile strength, Figure 5 (b) represents the yield strength, Figure 5 Where (c) represents elongation); from Figure 5 It can be seen that MAE along with L The change of shows a trend of decreasing first and then increasing; specifically, when L When the value is small, due to the lack of training samples, the model is prone to overfitting, resulting in large prediction errors; L As the number of training samples increases, the model can use more training samples, thereby improving the accuracy of prediction; however, when L As the value continues to increase, more irrelevant samples are introduced, making the nonlinear relationship in the data set more complicated, which in turn affects the predictive performance of the linear model.

[0037] S5. An integrated learning strategy is adopted to perform weighted integration on the prediction results of multiple local models, and the prediction results of the mechanical properties of hot-rolled strip based on multi-dimensional feature weighted similarity and integrated real-time learning are output as the final output results of the model.

[0038] To verify the superiority of the multi-dimensional feature weighted and integrated just-in-time learning model (MDFWS-IJITL), MDFWS-IJITL was compared with the JITL models (MDA-JITL, MICC-JITL, PCC-JITL and SCC-JITL) and the global linear regression model (G-LR) with different feature weighted similarity metrics. The prediction results of each model for different mechanical properties are shown in Figure 2. Figure 6 As shown ( Figure 6 (a) represents the root mean square error, Figure 6 (b) represents the mean absolute error, Figure 6 (c) DB mean absolute percentage error). Figure 6 It can be seen that G-LR has the highest prediction error for the three mechanical properties. This is because G-LR only performs modeling once and uses the established model for a long time. This modeling method ignores the time-varying characteristics of the hot-rolled strip production process and cannot better reflect the dynamic changes of industrial data, which leads to a large prediction error. Compared with G-LR, the error of the JITL model that introduces feature weights and time weights is reduced. The reason is that JITL selects relevant samples for re-modeling for each predicted sample, and the constructed model is more targeted. In addition, the constructed JITL model takes into account the feature weights and time weights, which further improves the correlation between the selected modeling samples and the samples to be tested.

[0039] For the prediction of different mechanical properties, each local model showed different prediction effects. For example, compared with other JITL models, MDA-JITL has the lowest MAE and RMSE in tensile strength prediction and the best prediction effect, while PCC-JITL has the best prediction performance in elongation compared with other JITL models. This is because different JITL models assign different weights to different features when selecting relevant modeling samples. The size of the feature weight reflects the correlation of the input feature to the output feature. If the assigned feature weight is closer to the true correlation between the input and output, the selection of relevant samples during the model construction process will be more accurate, thereby ensuring the effectiveness of the constructed model. Although the introduction of feature weights can improve the accuracy of relevant samples, a single JITL model can only take into account one-sided correlation, and there are often multiple similarity dimensions between the input and output features of hot-rolled strip samples. Figure 6 It can be seen that the MDFWS-IJITL model outperforms all JITL models in predicting different mechanical properties. This is as mentioned earlier, because each JITL model only analyzes the correlation between input and output data from one side, it is difficult to consider all the information of the original data in modeling. MDFWS-IJITL can analyze the correlation between input and output data in a variety of ways and integrate them into a framework. This sample selection method based on multi-dimensional feature weighted similarity reflects the more realistic correlation between samples as much as possible and improves the accuracy of relevant sample selection. In addition, the integration strategy can automatically assign appropriate weights to each JITL model, improving the robustness of the regression model. These characteristics make the MDFWS-IJITL method superior to other methods.

[0040] Figure 7 The figure shows the comparison of the prediction accuracy of different mechanical properties by different prediction models, where (a-1)~(a-6), (b-1)~(b-6) and (c-1)~(c-6) are the details of the predicted values ​​and true values ​​of tensile strength (TS), yield strength (YS) and elongation (EL) under different models, respectively. The black lines in (a-1)~(a-6) represent the relative error boundary of 5%, and the black lines in (b-1)~(b-6) and (c-1)~(c-6) represent the relative error boundary of 10%. Figure 7It can be seen that G-LR cannot achieve high accuracy due to one-time modeling. In contrast, the accuracy of each JITL model is improved compared with global modeling due to the introduction of dual weights of features and time. Compared with each JITL model, the MDFWS-IJITL model introduces multi-dimensional feature weights and time weights to improve the accuracy of relevant sample selection, so that the MDFWS-IJITL model shows better fitting effects in different mechanical property predictions. The distribution of prediction results on the diagonal has stronger aggregation and is basically within the set error area. The comparison of the scatter plots of the prediction results of each model further demonstrates the superiority of MDFWS-IJITL in the prediction of mechanical properties of tropical steel.

[0041] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A method for predicting mechanical properties of hot-rolled strip steel based on multi-dimensional feature similarity and integrated real-time learning, characterized in that: The steps include: S1, collect hot-rolled strip production data set and pre-process it, using hot-rolled production process parameters X[X1, X2, …, X n ] as the independent variable, and the hot-rolled strip mechanical properties Y[Y1, Y2, …, Y n ] is the dependent variable, and the production data of each coil of hot-rolled strip steel is taken as one data group to construct the historical data set of the prediction model; S2. Based on the historical data set, the feature quantification method A is used to calculate the hot rolling production process parameters X[X1, X2, …, X n ] on the mechanical properties of hot rolled strip Y[Y1, Y2, …, Y n ]; at the same time, each data group is sorted according to the corresponding hot-rolled strip production time and assigned a time weight; S3. Calculate the feature-time multi-dimensional similarity measurement value between the sample to be tested and the samples in the historical data set according to the feature weight and the time weight, select multiple samples with high similarity to the sample to be tested from the historical data set by accumulating similarity factors, and use them as training data sets to build a local model based on real-time learning, and output the prediction results of the mechanical properties of hot-rolled strip based on feature quantification method A; S4, using at least one other feature quantization method different from feature quantization method A, and repeating steps S2-S3; S5. An integrated learning strategy is adopted to perform weighted integration on the prediction results of multiple local models obtained under different feature quantification methods, and the prediction results of the mechanical properties of hot-rolled strip based on multi-dimensional feature weighted similarity and integrated real-time learning are output as the final output results of the model.

2. The method for predicting mechanical properties of hot-rolled strip based on multi-dimensional feature similarity and integrated real-time learning according to claim 1 is characterized in that: In step S1, data that meets one of the following three conditions during the data set preprocessing process is considered as abnormal data: (1) The measured value is outside the predetermined range; (2) Data is missing; (3) The 3σ principle is not met.

3. The method for predicting mechanical properties of hot-rolled strip based on multi-dimensional feature similarity and integrated real-time learning according to claim 1 is characterized in that: In step S1, the hot rolling production process parameters X[X1, X2, …, X n ] including chemical element content, furnace temperature, rough rolling start temperature, rough rolling start thickness, rough rolling final thickness, finishing rolling start temperature, finishing rolling final temperature, finishing rolling final thickness, coiling temperature; hot rolled strip mechanical properties Y[Y1, Y2, …, Y n ]Including tensile strength, yield strength and elongation.

4. The method for predicting mechanical properties of hot-rolled strip steel based on multi-dimensional feature similarity and integrated real-time learning according to claim 1, characterized in that: In step S2, the feature quantification method includes Pearson correlation analysis, Spearman correlation analysis, mutual information correlation analysis, and mean decreasing precision.

5. The method for predicting mechanical properties of hot-rolled strip steel based on multi-dimensional feature similarity and integrated real-time learning according to claim 1, characterized in that: In step S3, the feature-time multi-dimensional similarity measurement value calculation formula between the sample to be tested and the samples in the historical data set is: in: Represents the feature-time multidimensional similarity measure between the sample to be tested and the historical samples, For the j A historical sample, is the sample to be tested, is the feature weight similarity measure between the sample to be tested and the historical sample, is the time weight of historical samples, It is an adjustable parameter.

6. The method for predicting mechanical properties of hot-rolled strip steel based on multi-dimensional feature similarity and integrated real-time learning according to claim 5, characterized in that: In step S3, the cumulative similarity factor L The calculation formula is: Among them: the numerator represents the sum of the similarity contributions of the first n groups of samples with higher similarity, and the denominator represents the sum of the similarities between the test sample and all historical samples.

7. The method for predicting mechanical properties of hot-rolled strip steel based on multi-dimensional feature similarity and integrated real-time learning according to claim 1, characterized in that: In step S3, the local model construction method based on instant learning includes multiple linear regression, partial least squares regression, and support vector machine.

8. The method for predicting mechanical properties of hot-rolled strip steel based on multi-dimensional feature similarity and integrated real-time learning according to claim 1, characterized in that: In step S5, the weighted integration of the prediction results of multiple local models depends on the performance of each local model on the validation data set. r The mean absolute error of the models ( MAE ) is calculated as: in: and Respectively i The observed and predicted values ​​of the test samples, m is the number of test set samples.

9. The method for predicting mechanical properties of hot-rolled steel strip based on multi-dimensional feature similarity and integrated real-time learning according to claim 8, characterized in that: In step S5, each model MAE Definition r The weight of the sub-model output result is: Among them: The numerator represents the r The sub-model e is the base number, MAE is the function value of the exponential, and the denominator represents all sub-models with e is the base number, MAE is the sum of the function values ​​of the exponentials.

10. A hot-rolled strip mechanical properties prediction system based on multi-dimensional feature similarity and integrated real-time learning, characterized in that: include: Production data set collection and preprocessing module: collects hot-rolled strip production data sets and preprocesses them based on the hot-rolling production process parameters X[X1, X2, …, X n ] as the independent variable, and the hot-rolled strip mechanical properties Y[Y1, Y2, …, Y n ] is the dependent variable, and the production data of each coil of hot-rolled strip steel is taken as one data group to construct the historical data set of the prediction model; Multi-dimensional feature similarity calculation and local model construction module: For historical data sets, feature quantification method A is used to calculate hot rolling production process parameters X[X1, X2, …, X n ] on the mechanical properties of hot rolled strip Y[Y1, Y2, …, Y n ]; at the same time, each data group is sorted according to the corresponding hot-rolled strip production time and assigned a time weight; According to the feature weight and time weight, the feature-time multi-dimensional similarity measurement value between the sample to be tested and the samples in the historical data set is calculated, and multiple samples with high similarity to the sample to be tested are selected from the historical data set by accumulating similarity factors, and used as training data sets to build a local model based on real-time learning, and output the prediction results of the mechanical properties of the hot-rolled strip based on the feature quantization method A; at least one other feature quantization method different from the feature quantization method A is used, and the prediction results of the mechanical properties of the hot-rolled strip based on it are output; Ensemble learning final result output module: adopts ensemble learning strategy to weightedly integrate the prediction results of multiple local models obtained under different feature quantification methods, and outputs the prediction results of hot-rolled strip mechanical properties based on multi-dimensional feature weighted similarity and ensemble instant learning as the final output result of the model.

Images