A method and system for determining a cross crack based on a tissue prediction

Abstract

The application discloses a kind of based on organization prediction transverse crack determination method and system, comprising the following steps: S1, real-time collection L2 system slab identification information, when finding slab cutting, from L2 system collection the production performance information of this slab;S2, according to production performance information calculation obtains cooling curve, i.e. The temperature-time curve in the process from liquid to solid state;S3, according to production performance information, cooling curve, through slab organization prediction model module carries out the prediction of slab solidification organization information;S4, based on production performance information, slab solidification organization information, through transverse crack determination model module carries out the prediction of slab transverse crack, for the slab predicted as transverse crack carries out offline processing.The application fully considers the organization characteristics and performance in production of transverse crack occurrence, can reduce misjudgment rate, is favorable for improving hot charging hot delivery rate.

Description

Technical Field

[0001] This invention relates to the continuous casting process in the field of steelmaking technology, and more specifically, to a method and system for determining transverse cracks based on microstructure prediction. Background Technology

[0002] Hot charging and hot delivery of continuously cast billets, along with continuous rolling technology, can significantly reduce equipment investment and production costs, and improve product competitiveness. This requires hot charging and hot delivery of defect-free billets. Transverse cracks are mostly subcutaneous cracks, difficult to identify using infrared cameras. Therefore, real-time and effective online prediction of transverse crack defects on the surface of continuously cast billets is a highly important issue in continuous casting billet production.

[0003] In recent years, with the rapid development of information technology, especially the successful application of big data and AI technologies in industries such as the internet and healthcare, significant economic benefits have been brought about, leading to their application in industry. Machine learning / deep learning methods are beginning to be applied to the online detection of longitudinal cracks in cast billets. Publication No. CN111618265A, entitled "An Online Detection Method for Longitudinal Cracks in Continuously Cast Billets Based on K-Nearest Neighbor Classification," Publication No. CN111666710A, entitled "A Method for Predicting Longitudinal Cracks in Continuously Cast Billets Using Logistic Regression Classification," and Publication No. CN111680448A, entitled "A Method for Predicting Longitudinal Cracks in Continuously Cast Billets Based on Support Vector Machine (SVM) Classification," disclose methods for predicting longitudinal cracks in continuously cast billets using different machine learning classification algorithms (K-nearest neighbors, logistic regression, and support vector machine (SVM)). These methods involve splicing the temperature change rate of longitudinal cracks and the temperature of thermocouples in the same row under normal operating conditions to obtain temperature samples and a sample library. The classification algorithm is then used to classify the sample library and the preprocessed results of the online real-time detected thermocouple temperatures in the same row to identify and predict longitudinal cracks in continuously cast billets.

[0004] The publicly disclosed patents that apply machine learning algorithms are limited to the prediction of longitudinal cracks in slabs, and none involve applications specifically for online prediction of transverse cracks in slabs.

[0005] Based on this, the patent applicant filed a patent application with application number 202210015470.X, entitled "An Online Prediction Method for Transverse Cracks in Continuously Cast Slabs." This patent uses anomaly detection and a transverse crack tendency index of the steel grade to comprehensively determine whether transverse cracks will occur. In this patent, the transverse crack tendency index of the steel grade is set to 0 (no occurrence) or 1 (possible occurrence) based on whether transverse cracks have occurred in the past and whether precipitates have formed in the steel grade. If the transverse crack tendency index is 0, further, the anomaly probability P of real-time production data is calculated based on machine learning. When P is between 0.4 and 0.6, transverse cracks occur. The advantage of this method is that it considers the necessary conditions for the occurrence of conventional transverse cracks, namely, the occurrence of nitride precipitates such as AlN, Nb(CN), VN, and BN at grain boundaries. However, the shortcomings of this method are:

[0006] (1) Nitride precipitation at grain boundaries is a necessary condition for the formation of transverse cracks, but it is not a sufficient condition. The influence of grain size, carbon equivalent, etc. must also be considered. For example, when the grain size is small enough, the steel has good toughness, and the formation of precipitates will not lead to the formation of transverse cracks.

[0007] (2) The above patent application 202210015470.X is based on the steel tapping mark, not on the actual composition in the steelmaking process. In the actual composition, a small change in the N content can greatly affect the precipitation structure.

[0008] The above two points will improve the accuracy of transverse crack prediction, but will misjudge normal slabs as defective slabs, leading to an increased misjudgment rate, which is not conducive to improving the hot charging and hot delivery rate of slabs. Summary of the Invention

[0009] To address the aforementioned deficiencies in existing technologies, the present invention aims to provide a transverse crack determination method and system based on microstructure prediction. This method fully considers the microstructure characteristics of transverse crack occurrence and actual production performance, thereby reducing the misjudgment rate and improving the hot charging and hot delivery rate.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] On the one hand, a transverse crack determination method based on microstructure prediction includes the following steps:

[0012] The S1 and L2 systems (L2 level systems for production model and control) collect slab identification information in real time. When a slab is detected to be cut, the L2 system collects the production performance information of that slab.

[0013] S2. Calculate and obtain the cooling curve based on production performance information, that is, the temperature-time curve of the slab from liquid to solid state;

[0014] S3. Based on production performance information and cooling curves, predict the solidification structure of the slab using the slab structure prediction model module.

[0015] S4. Based on production performance information and slab solidification structure information, the transverse crack judgment model module is used to predict transverse cracks in the slab. Slabs predicted to have transverse cracks are removed from the production line.

[0016] Preferably, in step S1, the production performance information includes the intermediate package detection components, the temperatures of each thermocouple on the crystallizer, and the cooling modes of the primary and secondary cooling systems of the crystallizer; and / or

[0017] The production performance information also includes production stability information; and / or

[0018] The production stability information includes the slab position and slab width.

[0019] Preferably, in step S2, the cooling curve is the conventional cooling curve of the continuous casting machine.

[0020] Preferably, in step S3, the slab microstructure prediction model module includes a composition-cooling curve-microstructure database and a microstructure prediction calculation model;

[0021] The composition-cooling curve-microstructure database includes the microstructure of continuous casting machine tap marks according to nominal composition and common cooling modes;

[0022] The organization prediction calculation model is modeled using the similarity principle.

[0023] Preferably, the modeling of the slab microstructure prediction calculation model further includes:

[0024] First, based on the cooling mode in the production performance information obtained in step S2, data with the same cooling mode are found in the composition-cooling curve-microstructure database. Second, among the data with the same cooling mode, the top 3 to 20 cases with the most similar composition to the slab steel grade are found based on the composition similarity calculation. The microstructure of the target slab is the set of microstructures of the most similar cases, and the proportion of each microstructure is the average of the top 3 to 20 cases.

[0025] Preferably, the similarity calculation of the steel composition of the slab is based on the carbon equivalent principle, that is, the most similar case is the one with the smallest carbon equivalent value distance.

[0026] The closest is the one whose absolute value is the distance between C equivalent (Case 1) and C equivalent (Case 2).

[0027] Preferably, the similarity calculation of the slab steel composition is based on the principle of carbon equivalent and nitrogen content. The most similar case is found among the top 3 to 10 cases with similar carbon equivalent, and the case with the most similar nitrogen content (smallest nitrogen content distance) is selected.

[0028] Preferably, in step S3, the solidification structure information of the slab includes grain size, precipitates, and the number of precipitates.

[0029] Preferably, in step S4, the transverse crack determination model module obtains the model based on machine learning of historical data.

[0030] When performing machine learning, the variables in the data are carbon equivalent, grain size, precipitates, amount of precipitates, slab position, and slab width adjustment indicator, while the target of the data is the occurrence of transverse cracks.

[0031] On the other hand, a transverse crack detection system based on microstructure prediction includes:

[0032] The data acquisition module is used to collect slab production performance information and production stability information in the L2 system;

[0033] The cooling curve calculation module calculates and obtains the cooling curve and the temperature-time curve of the slab from liquid to solid state based on production performance information.

[0034] The slab microstructure prediction model module is used to predict the solidification microstructure of slabs based on production performance information and cooling curves.

[0035] The transverse crack detection model module, based on production performance information and slab solidification structure information, is used to predict transverse cracks in slabs.

[0036] The transverse crack determination system based on microstructure prediction is used to implement the transverse crack determination method based on microstructure prediction.

[0037] The present invention provides a transverse crack determination method and system based on microstructure prediction. It primarily analyzes the necessary conditions for transverse crack formation from the perspective of the solidification mechanism. When predicting and identifying transverse cracks, it considers the grain size and precipitates after solidification, reducing the probability of misclassifying normal slabs as transverse cracks. Simultaneously, it combines historical performance data feature values ​​for transverse crack identification, further reducing the false positive rate and improving the accuracy of transverse crack prediction. Attached Figure Description

[0038] Figure 1 This is a flowchart illustrating the transverse crack determination method based on tissue prediction of the present invention.

[0039] Figure 2 This is a schematic diagram of the framework of the transverse crack determination system based on tissue prediction of the present invention. Detailed Implementation

[0040] To better understand the above-mentioned technical solutions of the present invention, the technical solutions of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0041] Combination Figure 1 As shown, the present invention provides a transverse crack determination method based on microstructure prediction, comprising the following steps:

[0042] S1. Collect slab identification information from the L2 system (L2 level system for production model and control) in real time. When a slab is cut, collect the production performance information of the slab from the L2 system.

[0043] Production performance information includes the composition of the intermediate package, the temperature of each thermocouple on the crystallizer, and the cooling modes of the first and second cooling modes of the crystallizer; production performance information also includes production stability information, including slab position and slab width, etc.

[0044] S2. Calculate and obtain the cooling curve based on production performance information, that is, the temperature-time curve of the slab from liquid to solid state;

[0045] The cooling profile is the conventional cooling profile of the continuous casting machine. Typically, the cooling mode of a continuous casting machine is a defined conventional mode class, and there are a certain number of cooling profiles based on these cooling modes. Step S2 involves calculating the cooling profile mode of the slab at that time;

[0046] S3. Based on production performance information and cooling curves, the solidification structure information of the slab is predicted through the slab structure prediction model module. The solidification structure information of the slab includes grain size, precipitates, and the number of precipitates.

[0047] The slab microstructure prediction model module includes a composition-cooling curve-microstructure database and a microstructure prediction calculation model;

[0048] The composition-cooling profile-microstructure database includes the microstructure of continuous casting machine tap marks according to nominal composition and common cooling modes;

[0049] The organization's predictive computation model is modeled using the similarity principle.

[0050] The modeling of the organization's predictive computational model further includes:

[0051] First, based on the cooling mode in the production performance information obtained in step S2, find data with the same cooling mode in the composition-cooling curve-microstructure database; second, among the data with the same cooling mode, find the case with the most similar composition to the slab steel grade according to the similarity of the slab steel grade composition, and the microstructure of the target slab is the microstructure of the most similar case.

[0052] The similarity calculation of slab steel composition follows the carbon equivalent principle, that is, the most similar case is the one with the closest carbon equivalent value.

[0053] The similarity calculation of slab steel composition follows the principle of carbon equivalent and nitrogen content. The most similar case is found among the top 3 to 20 cases with similar carbon equivalent, and the case with the smallest difference in nitrogen content is selected.

[0054] S4. Based on production performance information and slab solidification structure information, the transverse crack judgment model module is used to predict transverse cracks in the slab. Slabs predicted to have transverse cracks are removed from the production line.

[0055] The transverse crack detection model module is obtained through machine learning based on historical data;

[0056] When performing machine learning, the variables in the data are carbon equivalent, grain size, precipitates, amount of precipitates, slab position, and slab width adjustment indicator, while the target of the data is the occurrence of transverse cracks.

[0057] Combination Figure 2 As shown, the present invention also provides a transverse crack determination system based on microstructure prediction, comprising:

[0058] The data acquisition module is used to collect slab production performance information and production stability information in the L2 system;

[0059] The cooling curve calculation module calculates and obtains the cooling curve and the temperature-time curve of the slab from liquid to solid state based on production performance information.

[0060] The slab microstructure prediction model module is used to predict the solidification microstructure of slabs based on production performance information and cooling curves.

[0061] The transverse crack detection model module, based on production performance information and slab solidification structure information, is used to predict transverse cracks in slabs.

[0062] The transverse crack determination method based on microstructure prediction of the present invention is realized by using the microstructure prediction transverse crack determination system based on the present invention.

[0063] As can be seen from the above, the transverse crack determination method and system based on microstructure prediction provided by this invention mainly analyzes the necessary conditions for transverse crack occurrence from the perspective of the solidification mechanism of transverse crack generation. When predicting and identifying transverse cracks, it considers the grain size and precipitates after solidification, reducing the probability of misclassifying normal slabs as transverse cracks. Simultaneously, it combines historical performance data feature values ​​for transverse crack identification, further reducing the false positive rate and improving the accuracy of transverse crack prediction. Finally, this method does not require additional detection instruments and, based on the microstructure of the steel after solidification, is easily adaptable to various production lines.

[0064] This invention relates to a method and system for determining transverse cracks based on microstructure prediction. This method reduces the false positive rate of transverse crack detection, which is beneficial for improving the hot charging and hot delivery rate. Furthermore, the method is simple, widely applicable, and requires no additional detection instruments, making it suitable for various production lines. Timely prediction of defective billets, allowing for corrective measures or sorting, is of significant importance for ensuring production continuity, improving product quality, and reducing production costs.

[0065] Those skilled in the art should recognize that the above embodiments are merely illustrative of the present invention and are not intended to limit the present invention. Any variations or modifications to the above embodiments that are within the spirit and essence of the present invention will fall within the scope of the claims of the present invention.

Claims

1. A method for determining transverse cracks based on microstructure prediction, characterized in that, Includes the following steps: S1. Collect slab identification information in real time from the L2 system of production model and control. When a slab is found to be cut, collect the production performance information of the slab from the L2 system of production model and control. S2. Calculate and obtain the cooling curve based on production performance information, that is, the temperature-time curve of the slab from liquid to solid state; S3. Based on production performance information and cooling curves, predict the solidification structure of the slab using the slab microstructure prediction model module. The slab microstructure prediction model module includes a composition-cooling curve-microstructure database and a microstructure prediction calculation model; The composition-cooling curve-microstructure database includes the microstructure of continuous casting machine tap marks according to nominal composition and common cooling modes; The organization prediction calculation model is modeled using the similarity principle. The modeling of the organization prediction calculation model further includes: First, based on the cooling mode in the production performance information obtained in step S2, data with the same cooling mode are found in the composition-cooling curve-microstructure database. Second, among the data with the same cooling mode, the top 3 to 20 cases most similar to the slab steel grade are found based on composition similarity calculation. The microstructure of the target slab is the set of microstructures of the most similar cases, and the proportion of each microstructure is the average of the top 3 to 20 cases. The similarity calculation of the steel composition of the slab is based on the carbon equivalent principle, that is, the most similar case is the one with the smallest carbon equivalent value distance; S4. Based on production performance information and slab solidification structure information, the transverse crack detection model module is used to predict transverse cracks in the slab. Slabs predicted to have transverse cracks are removed from the production line. The transverse crack determination model module is obtained through machine learning based on historical data; When performing machine learning, the variables in the data are carbon equivalent, grain size, precipitates, amount of precipitates, slab position, and slab width adjustment indicator, while the target of the data is the occurrence of transverse cracks.

2. The transverse crack determination method based on microstructure prediction according to claim 1, characterized in that: In step S1, the production performance information includes the components detected in the intermediate package, the temperatures of each thermocouple on the crystallizer, the cooling modes of the primary and secondary cooling systems of the crystallizer, and production stability information. The production stability information includes the slab position and slab width.

3. The transverse crack determination method based on microstructure prediction according to claim 1, characterized in that: In step S2, the cooling curve is the conventional cooling curve of the continuous casting machine.

4. The transverse crack determination method based on microstructure prediction according to claim 1, characterized in that: The similarity calculation of the slab steel composition is based on the principle of carbon equivalent and nitrogen content. The most similar case is found among the top 3 to 10 cases with similar carbon equivalent, and the case with the similarest nitrogen content is selected.

5. The transverse crack determination method based on microstructure prediction according to claim 1, characterized in that: In step S3, the solidification structure information of the slab includes grain size, precipitates, and the number of precipitates.

6. A transverse crack detection system based on microstructure prediction, characterized in that, include: The data acquisition module is used to collect slab production performance information from the L2 system of production model and control. The cooling curve calculation module calculates and obtains the cooling curve and the temperature-time curve of the slab from liquid to solid state based on production performance information. The slab microstructure prediction model module is used to predict the solidification microstructure of slabs based on production performance information and cooling curves. The transverse crack detection model module, based on production performance information and slab solidification structure information, is used to predict transverse cracks in slabs. The transverse crack determination system based on microstructure prediction is used to implement the transverse crack determination method based on microstructure prediction as described in any one of claims 1-5.

Images