A template matching recognition method based on continuous casting breakout propagation behavior

Abstract

The present invention discloses a template matching identification method based on the propagation behavior of continuous casting steel leakage, which belongs to the field of iron and steel metallurgy, and is characterized by including: establishing a continuous casting steel leakage sample database, obtaining a temperature rate thermal image of continuous casting steel leakage, extracting regional characteristics of continuous casting steel leakage, determining the propagation behavior of continuous casting steel leakage, obtaining continuous casting steel leakage templates with different propagation behaviors, and continuous casting steel leakage identification based on propagation behavior. The present invention combines the steel leakage propagation behavior in the steel continuous casting process with the template matching algorithm, and has the advantages of scientific rationality, strong applicability, good effect, etc., realizes the rapid identification and distinction of true and false continuous casting steel leakage, improves the recognition rate of continuous casting steel leakage, and provides a reliable method for online detection of continuous casting steel leakage.

Description

Technical Field

[0001] The invention belongs to the technical field of continuous casting in iron and steel metallurgy, and is a template matching and recognition method based on continuous casting leak propagation behavior. Background Art

[0002] Steel leakage is a major accident in the continuous casting process, which not only causes huge economic losses, but also damages the continuous casting equipment and interrupts normal production. In order to avoid the occurrence of steel leakage in continuous casting, the steel leakage prediction method has received extensive attention from metallurgical workers.

[0003] It is generally believed that continuous casting breakout occurs near the meniscus, and the tearing of the billet shell caused by poor mold lubrication and excessive friction is the direct cause of breakout. Limited by the special working conditions and invisible characteristics of the continuous casting mold, the mold thermocouple temperature signal is usually used to capture the typical temperature pattern of continuous casting breakout and establish a breakout prediction model. However, this breakout identification and prediction method based on one-dimensional temperature signals has gradually exposed some shortcomings in production practice, mainly manifested in problems such as missed reports and frequent false alarms. The root cause is that the formation and development process of continuous casting breakout has spatial and temporal propagation characteristics, and it is difficult for one-dimensional temperature signals to fully capture the spatial and temporal propagation characteristics of continuous casting breakout.

[0004] Patent document CN111570748 discloses a method for predicting steel leakage in a crystallizer based on image processing. The patent constructs the first three-dimensional spatial coordinates and the second three-dimensional spatial coordinates of the position of the temperature measuring point, the temperature of the temperature measuring point and the temperature rise rate of the temperature measuring point on the basis of arranging temperature measuring points on the side wall of the crystallizer to be measured, and slices the first three-dimensional spatial coordinates and the second three-dimensional spatial coordinates with the preset standard temperature and standard temperature rise rate, respectively, to obtain the temperature section and the temperature rise rate section, and then according to the appearance of the temperature section and the temperature rise rate section and the change of the section parameters, according to the preset warning rules, an early warning prompt is given, which overcomes the contradiction between the missed alarm rate and the false alarm rate in the existing steel leakage algorithm, and reduces the missed alarm rate without increasing the false alarm rate. However, in the actual production process, the temperature of the measuring point in the crystallizer is affected by factors such as the performance of the protective slag and the flow of molten steel, and may fluctuate continuously. The corresponding spatial image slice may have some abnormal interference areas, which leads to a decrease in the prediction accuracy of the method.

[0005] Patent document CN113935416 discloses a method for predicting steel leakage during continuous casting of a multi-state crystallizer. This patent combines the characteristics of thermocouple temperature time series data and uses a time convolutional network to perform feature recognition on the changes in the temperature of each thermocouple inside the crystallizer in the time dimension. This method can better model the time dependence of temperature. Unlike the traditional method of processing time series using recurrent neural networks, it combines with a time convolutional network through gating to effectively alleviate the gradient vanishing and explosion phenomena that occur during training. At the same time, the external feature of the slab pulling speed is also taken into account, and the bonding characteristics are judged more accurately, which improves the accuracy of steel leakage prediction. However, this method has many uncertainties and is not adaptable to changes in processes and steel grades, which can easily lead to false alarms. Summary of the invention

[0006] The purpose of the present invention is to overcome the shortcomings of the prior art and propose a scientific, reasonable, highly applicable and effective template matching recognition method based on the continuous casting leakout propagation behavior. The method combines the continuous casting leakout propagation behavior with template matching to accurately and efficiently identify the continuous casting leakout, providing the necessary conditions for online detection of continuous casting leakout.

[0007] To achieve the above object, the technical solution adopted by the present invention is: a template matching recognition method based on continuous casting breakout propagation behavior, which mainly includes the following steps:

[0008] 1) Establish a continuous casting leak sample database

[0009] ① Use the crystallizer online monitoring system to obtain real-time data of the continuous casting process;

[0010] ②According to the continuous casting on-site records, obtain steel leakage examples, extract the pouring data at the time of steel leakage, and establish a continuous casting steel leakage sample database;

[0011] 2) Obtain the temperature rate thermal image of continuous casting breakout

[0012] Based on the measured casting data, the corresponding relationship between temperature and color space was established by using interpolation algorithm and computer graphics technology, and the temperature of each copper plate was reflected on a plane to achieve thermal imaging of the temperature change rate of the copper plate in the crystallizer;

[0013] 3) Extract regional features of continuous casting leaks

[0014] ① Obtain the minimum value X of the horizontal coordinate of the continuous casting leakage hot area hot-min and the maximum value X hot-max ;

[0015] ② Obtain the minimum value X of the horizontal coordinate of the continuous casting leakage cold area cold-min and the maximum value X cold-max ;

[0016] ③ If the minimum horizontal coordinate value X of the hot area hot-min Less than X cold-min , then X hot-min Assign to X bo-min Otherwise, X cold-min Assign to X bo-min ;

[0017] ④ If the maximum value of the horizontal coordinate of the hot area is X hot-max Greater than X cold-max , then X hot-max Assign to X bo-max Otherwise, X cold-max Assign to X bo-max ;

[0018] ⑤ Calculate the width W of continuous casting leaky steel using formula (1);

[0019] W=X bo-max -X bo-min (1)

[0020] 4) Determine the propagation behavior of continuous casting breakout

[0021] ① If the width W of the leaking area is less than W lim , it is judged as a typical propagation leak;

[0022] ② If the width W of the leaking area is greater than W lim , it is judged as large-span leakage;

[0023] ③ If the leakage area only propagates to the right side, it is determined as a one-way leakage on the right side;

[0024] ④ If the leakage area only propagates to the left side, it is determined as left-side unidirectional leakage;

[0025] 5) Obtain continuous casting leaky steel templates with different propagation behaviors

[0026] According to the typical propagation, large span propagation, unidirectional left propagation and unidirectional right propagation behavior characteristics of continuous casting breakout, the continuous casting breakout sample set is divided into four subsets. The K-means clustering algorithm is used to calculate the minimum Euclidean distance between each sample and the cluster center using formula (2), and the cluster center with the minimum Euclidean distance is used as the template.

[0027]

[0028] Where, Q is the continuous casting leak sample set; Q i is a subset of steel leakage samples of four propagation behaviors; q is a continuous casting steel leakage sample; k is the number of cluster centers of the sample; p is the randomly selected initial cluster center;

[0029] 6) Continuous casting leak identification based on propagation behavior

[0030] ①Use upsampling or downsampling methods to pre-process the obtained hot and cold areas of steel leakage so that their width is consistent with the template;

[0031] ② Use continuous casting steel leakage templates with different propagation behaviors for matching. If the maximum matching degree is greater than the threshold T, it is judged as continuous casting steel leakage; if the maximum matching degree is less than the threshold T, it is judged as pseudo continuous casting steel leakage.

[0032] Among them, the width threshold of the steel leakage area is W lim The number of cluster centers k of the sample is set to 4; the template matching threshold T is set to 0.6.

[0033] This method is suitable for identifying steel leaks in continuous casting of slabs and round billets.

[0034] The beneficial effect of the template matching recognition method based on the continuous casting steel leakage propagation behavior of the present invention is: combining the steel leakage propagation behavior in the steel continuous casting process with the template matching algorithm, selecting continuous casting steel leakage templates with different propagation behaviors for matching, improving the recognition rate of continuous casting steel leakage, reducing the false alarm rate of pseudo steel leakage, and providing an effective method for online detection of continuous casting steel leakage. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a flow chart of continuous casting leaky steel template matching and identification;

[0036] Figure 2 It is the crystallizer temperature rate thermogram;

[0037] Figure 3 It is a typical propagation template for continuous casting breakout;

[0038] Figure 4 It is a large span propagation template for continuous casting leaky steel;

[0039] Figure 5 It is a one-way left-side propagation template for continuous casting leakage;

[0040] Figure 6 It is a one-way right-side propagation template for continuous casting leakage. DETAILED DESCRIPTION

[0041] The present invention is further described below with reference to the embodiments, but the present invention is not limited thereto.

[0042] like Figure 1 As shown, a template matching recognition method based on continuous casting breakout propagation behavior comprises the following steps:

[0043] Step 1: Establish a continuous casting leak sample database

[0044] ① Use the crystallizer online monitoring system to obtain real-time data of the continuous casting process;

[0045] ②According to the continuous casting on-site records, obtain steel leakage examples, extract the pouring data at the time of steel leakage, and establish a continuous casting steel leakage sample database.

[0046] Step 2: Obtain the temperature rate thermal image of continuous casting leakage

[0047] Based on the measured casting data, the corresponding relationship between temperature and color space was established by using interpolation algorithm and computer graphics technology, and the temperature of each copper plate was reflected on a plane to achieve thermal imaging of the temperature change rate of the copper plate in the crystallizer, such as Figure 2 shown.

[0048] Step 3: Extract regional features of continuous casting leakage

[0049] ① Obtain the minimum value X of the horizontal coordinate of the continuous casting leakage hot area hot-min and the maximum value X hot-max ;

[0050] ② Obtain the minimum value X of the horizontal coordinate of the continuous casting leakage cold area cold-min and the maximum value X cold-max ;

[0051] ③ If the minimum horizontal coordinate value X of the hot area hot-min Less than X cold-min , then X hot-min Assign to X bo-min Otherwise, X cold-min Assign to X bo-min ;

[0052] ④ If the maximum value of the horizontal coordinate of the hot area is X hot-max Greater than X cold-max , then X hot-max Assign to X bo-max Otherwise, X cold-max Assign to X bo-max ;

[0053] ⑤ Use formula (1) to calculate the width W of continuous casting leak.

[0054] W=X bo-max -X bo-min (1)

[0055] Step 4: Determine the propagation behavior of continuous casting breakout

[0056] ① If the width W of the leaking area is less than 230 pixels, it is judged as a typical propagation leaking;

[0057] ② If the width W of the leaking area is greater than 230 pixels, it is determined to be a large-span leaking area;

[0058] ③ If the leakage area only propagates to the right side, it is determined as a one-way leakage on the right side;

[0059] ④ If the steel leakage area only propagates to the left side, it is judged as left-side unidirectional steel leakage.

[0060] Step 5: Obtain continuous casting leaky steel templates with different propagation behaviors

[0061] According to the typical propagation, large span propagation, unidirectional left propagation and unidirectional right propagation behavior characteristics of continuous casting breakout, the continuous casting breakout sample set is divided into four subsets. The K-means clustering algorithm is used to calculate the minimum Euclidean distance between each sample and the cluster center using formula (2), and the cluster center with the minimum Euclidean distance is used as the template.

[0062]

[0063] Where, Q is the continuous casting leak sample set; Q i is a subset of steel leakage samples with four propagation behaviors; q is a continuous casting steel leakage sample; k is the number of cluster centers of the sample, which is set to 4; p is the randomly selected initial cluster center.

[0064] The continuous casting leaky steel templates with different propagation behaviors are obtained through clustering algorithm. Figure 3 It is a typical propagation template for continuous casting breakout. Figure 4 It is a large-span propagation template for continuous casting leakage. Figure 5 It is a one-way left-side propagation template for continuous casting leakage. Figure 6 It is a one-way right-side propagation template for continuous casting leakage.

[0065] Step 6: Continuous casting leak identification based on propagation behavior

[0066] ①Use upsampling or downsampling methods to pre-process the obtained hot and cold areas of steel leakage so that their width is consistent with the template;

[0067] ② Use continuous casting steel leakage templates with different propagation behaviors for matching. If the maximum matching degree is greater than the threshold value of 0.6, it is judged as continuous casting steel leakage; if the maximum matching degree is less than the threshold value of 0.6, it is judged as pseudo continuous casting steel leakage.

[0068] This method is suitable for identifying steel leaks in continuous casting of slabs and round billets.

[0069] Finally, based on the casting data of a domestic steel plant, 120 cases of true and false continuous casting were matched, and the results are shown in Table 1. When the four templates of typical propagation, large span propagation, left unidirectional propagation and right unidirectional propagation were selected, the reporting rate of continuous casting leakage was 100%, and there was no wrong identification, which shows that the four templates can identify continuous casting leakage more accurately and efficiently, providing a reliable method for continuous casting leakage prediction.

[0070] Table 1 Template matching recognition results

[0071]

[0072] The above-described embodiments merely express the implementation methods of the present invention, but they cannot be understood as a prediction of the scope of the patent of the present invention. It should be pointed out that for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A template matching recognition method based on continuous casting breakout propagation behavior, characterized in that: The steel leak propagation behavior in the steel continuous casting process is combined with the template matching algorithm, and the steel leaks with different propagation behaviors are used as templates to achieve rapid identification of true and false continuous casting steel leaks. The specific identification steps are as follows: 1) Establish a continuous casting leak sample database ① Use the crystallizer online monitoring system to obtain real-time data of the continuous casting process; ②According to the continuous casting on-site records, obtain steel leakage examples, extract the pouring data at the time of steel leakage, and establish a continuous casting steel leakage sample database; 2) Obtain the temperature rate thermal image of continuous casting breakout Based on the measured casting data, the corresponding relationship between temperature and color space was established by using interpolation algorithm and computer graphics technology, and the temperature of each copper plate was reflected on a plane to achieve thermal imaging of the temperature change rate of the copper plate in the crystallizer; 3) Extract regional features of continuous casting leaks ① Obtain the minimum value X of the horizontal coordinate of the continuous casting leakage hot area hot-min and the maximum value X hot-max ; ② Obtain the minimum value X of the horizontal coordinate of the continuous casting leakage cold area cold-min and the maximum value X cold-max ; ③ If the minimum horizontal coordinate value X of the hot area hot-min Less than X cold-min , then X hot-min Assign to X bo-min Otherwise, X cold-min Assign to X bo-min ; ④ If the maximum value of the horizontal coordinate of the hot area is X hot-max Greater than X cold-max , then X hot-max Assign to X bo-max Otherwise, X cold-max Assign to X bo-max ; ⑤ Calculate the width W of continuous casting leaky steel using formula (1); W = X bo-max -X bo-min (1) 4) Determine the propagation behavior of continuous casting breakout ① If the width W of the leaking area is less than W lim , it is judged as a typical propagation leak; ② If the width W of the leaking area is greater than W lim , it is determined to be a large-span steel leakage; ③ If the leakage area only propagates to the right side, it is determined as a one-way leakage on the right side; ④ If the leakage area only propagates to the left side, it is determined as left-side unidirectional leakage; 5) Obtain continuous casting leaky steel templates with different propagation behaviors According to the typical propagation, large span propagation, unidirectional left propagation and unidirectional right propagation behavior characteristics of continuous casting breakout, the continuous casting breakout sample set is divided into four subsets. The K-means clustering algorithm is used to calculate the minimum Euclidean distance between each sample and the cluster center using formula (2), and the cluster center with the minimum Euclidean distance is used as the template. Where, Q is the continuous casting leak sample set; Q i is a subset of steel leakage samples of four propagation behaviors; q is a continuous casting steel leakage sample; k is the number of cluster centers of the sample; p is the randomly selected initial cluster center; 6) Continuous casting leak identification based on propagation behavior ①Use upsampling or downsampling methods to pre-process the obtained hot and cold areas of steel leakage so that their width is consistent with the template; ② Use continuous casting leaky steel templates with different propagation behaviors for matching. If the maximum matching degree is greater than the threshold T, it is determined to be continuous casting leaky steel; If the maximum matching degree is less than the threshold value T, it is judged as pseudo continuous casting leakage.

2. The template matching recognition method based on continuous casting breakout propagation behavior according to claim 1 is characterized in that: The width threshold of the leaking area W lim The number of cluster centers k of the sample is set to 4; the template matching threshold T is set to 0.

6.

3. The template matching recognition method based on continuous casting breakout propagation behavior according to claim 1 is characterized in that: This method is suitable for identifying steel leaks in continuous casting of slabs and round billets.

Images