A method for detecting clogging of a stopper in a continuous casting process

Abstract

The application discloses a method for detecting clogging of a stopper in a continuous casting process, comprising the following steps: S1, in the early stage of pouring, online machine learning modeling is performed on real-time high-frequency data to obtain a decision coefficient R 2 value for evaluating the accuracy of a model; S2, every m window, the performance of the evaluation model is evaluated by using the high-frequency data in the window to obtain the R 2 value of the evaluation model in the m window; S3, the R 2 value of the window is compared with the decision coefficient R 2 value to perform early warning or alarm; and S4, the steps S2 and S3 are repeated until the pouring of the current pouring is finished. The application can perform early warning of clogging of the stopper in the continuous casting process, remind an operator to remove the clogging or replace the stopper, and ensure the sequence in the production process and the quality of the continuous casting slab.

Description

TECHNICAL FIELD

[0001] The present application relates to a continuous casting process in the field of steelmaking technology, and more particularly to a method for detecting clogging of a stopper during continuous casting. BACKGROUND

[0002] In the pouring process, in the tundish with stopper flow control, the larger size of alumina inclusions is easy to adhere to the position of the stopper rod head or the upper nozzle due to the flow of molten steel from the gap between the stopper and the upper nozzle of the tundish into the mold; the smaller size of inclusions gathers and grows at the bottom of the concave nozzle, and is easy to adhere to the bottom of the submerged nozzle or float to the mold. When the adhered inclusions become more and more, they gradually grow, causing clogging of the stopper. It is very important to find the clogging of the stopper in time to improve the subsequent process.

[0003] When the stopper is clogged, the inner diameter of the nozzle is reduced, and the steel flow is reduced. In order to maintain the pulling speed, the stopper rises, and the opening degree becomes larger. Therefore, the stopper opening degree and the liquid level fluctuation are important characteristic parameters in the continuous casting process, which are often checked in real time. However, due to the different steel flow of each furnace, the argon pressure, flow and other parameters are also different, and the absolute value of the stopper opening degree is difficult to feedback the clogging of the stopper. Therefore, it is necessary to put forward a more effective stopper detection method.

[0004] The existing published documents related to stopper clogging are in the following two aspects:

[0005] (1) Reduce the clogging of the stopper by improving the purity of the molten steel;

[0006] (2) Improve the clogging by nozzle design and control the liquid level fluctuation.

[0007] There are few documents on stopper clogging detection, such as patent application No. 20211065639.5, patent name "continuous casting nozzle clogging analysis method". This patent aims at the deficiency of the existing published documents that are mostly using physical simulation or numerical simulation method to construct the model of clogging phenomenon and cannot determine the degree of clogging. The quantitative evaluation of clogging rate and clogging degree is established. This patent simulates the real working condition of molten steel continuous casting equipment through molten steel continuous casting model, and simulates the clogging process and shedding process, and monitors the state change of the liquid surface, so as to construct the relationship curve between the state change of the liquid surface and the clogging behavior data, which is used for the evaluation of clogging in actual mass production. The disadvantage of this method is that it cannot solve the deficiency caused by simulation test and real continuous casting equipment, and still uses the conventional physical simulation method. SUMMARY

[0008] In view of the above-mentioned defects existing in the prior art, the purpose of the present application is to provide an on-line stopper clogging detection method in a continuous casting pouring process, to carry out stopper clogging early warning in continuous casting pouring, to remind the operator to remove the clogging or replace the stopper, and to ensure the sequence in the production process and the quality of the continuous casting slab.

[0009] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0010] An on-line stopper clogging detection method in a continuous casting pouring process, comprising the following steps:

[0011] S1, at the early stage of pouring start, carrying out on-line machine learning modeling on real-time high-frequency data, to obtain a decision coefficient R 2 value for evaluating the accuracy of the model;

[0012] S2, every time interval m window, applying the high-frequency data in the window to evaluate the performance of the evaluation model, to obtain the R 2 value of the evaluation model in this time interval m window;

[0013] S3, comparing the R 2 value of the window with the decision coefficient R 2 value to execute early warning or alarm;

[0014] S4, repeating step S2 and step S3 until the pouring of the current pouring is finished.

[0015] Preferably, in the step S1, the high-frequency data includes the mold level, the pulling speed, the argon pressure, the argon flow and the stopper opening degree; and further includes the mold width.

[0016] Preferably, the sampling frequency of the high-frequency data is 1s-10s, and the pre-period data for establishing the evaluation model is not more than 20%-40% of the pouring time length.

[0017] The pre-period data for establishing the evaluation model is the data of 600s-2400s time length after the pouring starts.

[0018] Preferably, in the step S1, the method of on-line machine learning modeling is multiple linear regression.

[0019] Preferably, in the step S2, the time interval m is in the range of 600s-2400s.

[0020] Preferably, in the step S3, the R 2 value of the window is compared with the decision coefficient R 2 value, when the two deviate by 30%, a clogging yellow card early warning is executed, indicating that the clogging condition will affect the subsequent slab production quality; when the two deviate by more than 50%, a clogging red card alarm is executed.

[0021] Preferably, the determination coefficient R 2 value is represented as R 2 0, the R 2 value of the window is represented as R 2 m;

[0022] The deviation between R 2 m and R 2 0 is the nodule index, and the calculation method is as follows:

[0023] Nodule index = | (R 2 m-R 2 0) / R 2 0 |.

[0024] Preferably, the nodule index is transmitted to the L2 level process monitoring or quality judgment system for subsequent slab quality analysis.

[0025] Preferably, when the stopper opening degree is equal to 80% of the historical maximum stopper opening degree, the nodule red card alarm is also executed.

[0026] Preferably, the step S4 further includes repeating steps S1 to S4 when a tundish replacement is detected.

[0027] The online stopper nodule detection method provided by the application is based on the real-time model established in the early stage of the pouring process, and monitors the subsequent pouring process, and proposes a nodule index. When the nodule index reaches 30% or more, a nodule yellow and red card alarm is performed, and the nodule index is transmitted to the L2 system for quality prediction model. The method does not depend on the structure of the crystallizer and the state of the equipment, and can be transplanted to other continuous casting equipment without obstacles, and has the following beneficial effects:

[0028] (1) The method is based on the real-time nodule model established in the early stage of the pouring process, and the real-time nodule model is applied to monitor the subsequent pouring process, which can monitor the influence of nodule on production sequence and quality;

[0029] (2) The method applies the data of the current pouring, and does not depend on the structure of the crystallizer and the state of the equipment, and can be transplanted without obstacles. BRIEF DESCRIPTION OF DRAWINGS

[0030] Figure 1 is a flowchart of the online stopper nodule detection method in the continuous casting pouring process of the application. DETAILED DESCRIPTION

[0031] In order to better understand the above technical solutions of the application, the technical solutions of the application will be further described below in combination with the drawings and examples.

[0032] Combining Figure 1 The application provides a method for detecting clogging of a stopper in a continuous casting process, which comprises the following steps:

[0033] S1, at the early stage of the start of casting, online machine learning modeling is performed on real-time high-frequency data to obtain a decision coefficient R 2 value of an evaluation model, which is represented as R 2 0; the decision coefficient R 2 value is used to evaluate the accuracy of the model;

[0034] The high-frequency data includes the crystallizer liquid level, the pulling speed, the argon pressure, the argon flow and the stopper opening degree;

[0035] The crystallizer width is also included to cover the case of online width adjustment of the crystallizer;

[0036] The frequency of the high-frequency data is 1s-10s, and the data used for the establishment of the evaluation model is not more than 20%-40% of the casting length;

[0037] The data used for the establishment of the evaluation model is the data within a time length of 600s-2400s from the start of casting;

[0038] The method of online machine learning modeling is multiple linear regression;

[0039] S2, every time interval m window, the high-frequency data in the window is applied to evaluate the performance of the evaluation model, and the R 2 value of the evaluation model in the time interval m window is obtained, which is represented as R 2 m;

[0040] The time interval m is in the range of 600s-2400s;

[0041] S3, the R 2 value of the window is compared with the decision coefficient R 2 0, that is, R 2 m and R 2 0 are compared, when the deviation reaches 30%, a clogging yellow card warning is performed, indicating that the clogging will affect the quality of the subsequent slab production; when the deviation reaches more than 50%, a clogging red card alarm is performed;

[0042] The deviation between R 2 m and R 2 0 is the clogging index, which is calculated as follows:

[0043] Clogging index = | (R 2 m-R 2 0) / R 2 0 |;

[0044] The execution of the clogging red card alarm can also add the stopper opening degree value, when the stopper opening degree is equal to 80% of the historical highest stopper opening degree, the clogging red card alarm is also executed;

[0045] In addition to being used for real-time clogging alarm, the clogging index is also transmitted to the process monitoring or quality determination system (referred to as L2 system) of L2 level, for subsequent slab quality analysis;

[0046] Generally, the smelting industry has L1 to L4 level systems, the L1 system is an instrument information system, mainly a high frequency data system, and the L2 system is a model system, which is a system with lower frequency;

[0047] S4, repeat step S2 and step S3 until the pouring of the current pouring is finished, and the repeated step is only to the end of the tundish; when the tundish replacement is detected, steps S1 to S4 need to be repeated in order to cover the influence of the tundish structure on the evaluation model.

[0048] In summary, the online stopper clogging detection method in the continuous casting pouring process is based on the real-time clogging model established in the early stage of the pouring process, and the subsequent pouring process is monitored by using the real-time clogging model, which can reduce the influence of clogging on production sequence and quality. The method detects whether the subsequent pouring is "smooth" by using the "smooth" pouring of the current pouring, does not depend on the mold structure and equipment state, and can be transplanted without obstacles.

[0049] Embodiment

[0050] Reference is made to Figure 1 As shown, the online stopper clogging detection method according to the patent is used for online stopper clogging detection of a No. 3 continuous casting machine of a steelmaking enterprise. The No. 3 continuous casting machine is double-flow (even flow and odd flow) pouring, and the method is applied to each flow number of each pouring for each flow number to detect whether clogging occurs in the stopper of each flow channel of each pouring.

[0051] The online stopper clogging detection method in the continuous casting pouring process of the embodiment specifically includes the following steps:

[0052] Note: The model is used to detect the pouring of a continuous casting machine, and the pouring is checked for each pouring. In the same pouring process, the tundish replacement may be encountered, which means that the stopper is replaced and the calculation is restarted (still in the same pouring). The embodiment is implemented on a continuous casting machine, and the results of different pourings (pouring number identification) are different. Therefore, I am not very clear about how to do the supplementary embodiment, but the results under different pouring conditions are supplemented.

[0053] S1, when a new casting is identified from the L2 level data source, collect the mold level, casting speed, argon pressure, argon flow rate, and stopper opening degree data from the L1 instrument information every 2s for 600s-1800s from the start of casting;

[0054] S2, using the above data, establish a multiple linear regression model based on the mold level, casting speed, argon pressure, argon flow rate, and stopper opening degree, and evaluate the model using the above data to obtain the R 2 value (denoted as R 2 0);

[0055] S3, starting from 1800s, collect the mold level, casting speed, argon pressure, argon flow rate, and stopper opening degree data for a time length of 900s, and apply these data to evaluate the linear regression model to obtain the R 2 value (denoted as R 2 m) for the time length; calculate the clogging index for the time length;

[0056] S4, when R 2 m deviates from R 2 0 by 30%, issue a yellow card warning; when the deviation reaches 50%, issue a red card warning. Transmit the clogging index to the L2 level process monitoring or quality determination system for storage for analysis by process personnel when slab quality problems occur.

[0057] S5: if no tundish replacement is detected (note: if no tundish replacement is detected, it means that one stopper is used for this casting, and recalculation is not needed, otherwise recalculation is needed), repeat steps S2-S4 every 900s;

[0058] S6: if tundish replacement is detected, repeat steps S1-S5;

[0059] S7: repeat steps S1-S6 until the casting of this casting is completed.

[0060] In No. 3 continuous casting machine, the number of continuous casting is not many, when the stopper alarm occurs, it is often in the middle and later stage of a casting, the yellow card alarm has less effect on the casting progress, and more affects the slab quality, while the red card alarm affects the casting progress, if it is not the end of casting (tail blank), measures need to be taken. According to this feature, when the operator hears the yellow card alarm, he checks the argon pressure value during casting, if the argon pressure value is normal (<50MPa), the casting continues normally, otherwise the stopper is replaced. If a red card alarm occurs, if the casting slab is not in the tail at this time, measures need to be taken to replace the stopper to prevent large liquid level fluctuations after clogging. The following table is a typical case and measures after the enterprise implements this method.

[0061]

[0062]

[0063] Those skilled in the art will recognize changes, modifications, and variations of the embodiments described herein can be made without departing from the spirit of the present application and that the scope of the application should be judged by the breadth of the claims and can be supported by any of the examples disclosed herein.

Claims

1. A method for on-line clogging detection of a stopper in a continuous casting process, characterized by, The method comprises the following steps: S1, in the early stage of pouring starting, online machine learning modeling is performed on real-time high-frequency data to obtain a decision coefficient R for evaluating model accuracy 2 value; The high-frequency data comprises crystallizer liquid level, casting speed, argon pressure, argon flow and stopper opening degree; and further comprises crystallizer width; The sampling frequency of the high-frequency data is 1s-10s, and the early-stage data for establishing the evaluation model is not more than 20%-40% of the length of the pouring time; The early-stage data for establishing the evaluation model is the data of the length of 600s-2400s since the start of pouring; The method of online machine learning modeling is multiple linear regression; S2, every time interval m window, the high-frequency data in the window is applied to evaluate the performance of the evaluation model, and the R value of the evaluation model in the time interval m window is obtained 2 value; S3, compare the R of the window. 2 Value and the determination coefficient R 2 The values ​​are used to issue early warnings or alarms. When the deviation between the two values ​​reaches 30%, a yellow warning for nodule formation is issued, indicating that nodule formation will affect the subsequent slab production quality. When the deviation exceeds 50%, a red alarm for nodule formation is issued. The decision coefficient R 2 The value is denoted as R 2 0, the R 2 The value is denoted as R 2 m; Comparative R 2 m and R 2 The deviation from 0 between m and R is the nodule index, which is calculated as follows: ； S4, repeating step S2 and step S3 until the end of the current pouring.

2. The method for on-line clogging detection of a stopper in a continuous casting process according to claim 1, characterized in that: In the step S2, the time length m is in the range of 600s-2400s.

3. The method of on-line clogging detection of a stopper in a continuous casting process according to claim 1, characterized in that: The nodulation index is transmitted to the process monitoring or quality determination system of the L2 level, which is used for subsequent slab quality analysis.

4. The method of on-line clogging detection of a stopper in a continuous casting process according to claim 1, characterized in that: When the stopper opening degree is equal to 80% of the historical highest stopper opening degree, a nodulation red card alarm is also executed.

5. The method of on-line clogging detection of a stopper in a continuous casting process according to claim 1, characterized in that: In the step S4, when the tundish replacement is detected, steps S1-S4 need to be repeated.

Images