Analysis of true and false sticker breakouts’ characteristics based on mould thermal image

Abstract

All 198 samples were obtained for past three years, including true and false sticker breakouts. Based on the mould thermal image of temperature rate, temperature, geometric and shape characteristics were calculated, such as maximum temperature rate, height, width, area, perimeter, circularity, rectangularity and aspect ratio. The difference of true and false sticker breakouts was compared and analysed. The results show that the maximum temperature rate of hot region is 7.13 °C/s for true sticker breakout while it is 5.27 °C/s for false sticker breakout. The minimum temperature rate of cold region is −4.31 °C/s for true sticker breakout while it is −3.43 °C/s for false sticker breakout. The maximum and minimum temperature rates of true sticker breakout are usually more than those of false sticker breakout. When the area of hot region exceeds 24,998 pixels, no sticker breakout has occurred, but there were 10 samples of false sticker breakout. The height, width and areas of hot and cold regions are more concentrated for true sticker breakouts, while false sticker breakout are more scattered. The circularity value of hot region is in the range of 0.17–0.52 for true sticker breakout, which is generally lower than the false sticker breakout (0.13–0.87). The shape of cold region is closer to a circular shape for true sticker breakout.

Keywords

Continuous casting sticker breakout geometric characteristics shape characteristics thermal image mould temperate rate hot and cold region‌

Introduction

The casting speed is more than before along with the development of efficient continuous casting technology. The new coordination between each casting parameter has become a key issue for steel plant.^1–3 If the parameters are not proper, such as friction, heat transfer and poor lubrication of liquid slag,^4–7 it will lead to a series of non-equilibrium problems. Meanwhile, some quality problems will appear in mould, even some accidents. Sticker breakout is the most serious one. It not only damages the continuous caster machine, but also brings a long interruption for continuous casting production.^8–10

At present, some logical judgment and artificial intelligence models^11–14 have been used to predict breakout in steel plant, which has effectively reduced the occurrence of breakout accidents.^15–17 However, in recent years, the number of false alarms increases along with the increase of casting speed in actual production. Those frequent false alarms not only lead to a sudden drop of casting speed, but also decrease the lifespan of continuous caster machine.^18–20 The main reason is that sticker breakout has spatial characteristics, while the current prediction method based on the temperature curve is difficult to describe all spatial characteristics of sticker breakout.^21–23 In response to this issue, some researchers have developed the mould thermal image of temperature rate to monitor sticker breakout in mould.^24–27 It can be used to detect the abnormal condition and spatial characteristics directly. Due to the complexity of different continuous casting conditions, the characteristics of sticker breakout are different to some extent.^28–30 Therefore, it is of great significance to analyse the spatial characteristics of true and false sticker breakouts for conventional slab during continuous casting.^31–33

Based on continuous casting data of steel plant, 99 true sticker breakouts and 99 false sticker breakouts were collected. And characteristics of those samples were extracted with the computer image processing and feature extraction method, such as temperature, geometry and shape. From the view of hot and cold regions in mould thermal image, characteristics were statistically analysed, such as maximum temperature rate, height, width, area, perimeter, circularity, rectangularity and aspect ratio in hot and cold regions. The difference of true and false sticker breakouts was compared and analysed in detail.

Experiment

Caster and samples

The machine is an arc continuous caster. The length of mould copper plates is 900 mm. The common slab widths are 1200, 1307, 1407 and 1558 mm, respectively. The main slab thickness is 235 mm. The casting speed is 0.6–1.6 m/min. The main parameters are included in Table 1.

Table 1.

Continuous caster parameters.

Item	Parameter
Caster type	Arc caster
Strand	2
Casting speed, m/min	0.6–1.6
Mould length, mm	900
Slab width, mm	1200, 1307, 1407, 1558
Slab thickness, mm	235
Drive	Hydraulic drive

To investigate the formation of sticker breakout, thermocouples were arranged at the outside radius, inside radius, left narrow face and right narrow face of the mould copper plate. As shown in Figure 1, there are 72 thermocouples in total. The outside and inside radiuses have 11 columns of thermocouples, and the narrow face has one column. There are three rows of thermocouples with distances from the mould top, 210, 325 and 445 mm, respectively. The distance between each thermocouple is 150 mm.

All 99 true sticker breakouts and 99 false sticker breakouts were obtained based on the on-line mould monitoring system during the conventional slab continuous casting. Because the continuous casting was a complex process, false sticker breakouts were picked up from five different conditions, such as casting start-up, normal casting condition, temperature with large fluctuation, temperature with small fluctuation and changed casting status. Those five process conditions encompassed the majority of situations during continuous casting.

Mould thermal image of temperature rate

Figure 2 shows the temperature variation trend during sticker breakout. In Figure 2(a), the temperature variation is small when there is no sticker breakout. In Figure 2(b), a sticker breakout appears, the thermocouple's temperature of the first row will rise obviously, which also has a large temperature rate. In Figure 2(c), along with the movement of continuous casting slab and mould vibration, the sticker point moves down and arrives at the second row location. The temperature of the second row will rise, which is usually called ‘temperature lag’. The temperature of the first row will fall due to the increase of slab thickness, which is usually lower than that of the second row. It is usually called ‘temperature inversion’. In Figure 2(d), as the sticker point leaves the first and second rows and moves towards to the bottom of the mould, the temperature in both rows decreases.

To capture the abnormal temperature change in the mould, a mould thermal image of temperature rate was built up with computer graphics technology. As shown in Figure 3, when there is a sticker breakout in the mould, the overflowed molten steel will connect with mould copper plate directly. It causes a large region with high temperature rate as shown in region A, which is usually called the hot region. When this sticker passed, a cold region with minus temperature rate appears as shown in region B. The hot region is a region with temperature rate higher than 0.3 °C/s, cold region is a region with temperature rate lower than −0.3 °C/s. According to the threshold segmentation algorithm,³⁴ when the temperature rate is 0 or changes little, there is a clear boundary between the cold and hot regions.

Characteristics of extraction of hot and cold regions

To capture the characteristics of hot and cold regions for sticker breakouts, the maximum temperature rate, geometric and shape characteristics were extracted for 99 true sticker breakouts and 99 false sticker breakouts.

Temperature rate

The temperature rates were stored in a vector as shown in formula (1).

\vec{T} = [{\dot{T}}_{1}, {\dot{T}}_{2}, {\dot{T}}_{3} \dots \dots {\dot{T}}_{n}] n = 30

(1)

where,

\vec{T}

is the temperature rate vector, °C/s;

{\dot{T}}_{i}

is the temperature rate at the i second, °C/s; n is the time range for temperature rate, s.

Maximum(minimum) temperature rate: the maximum(minimum) temperature rate is the maximum(minimum) value of temperature rate in the hot (cold) region, represented by formula (2) and (3).

G_{\max} = {\dot{T}}_{i} (if ({\dot{T}}_{i} \geq G_{\max}))

(2)

G_{\min} = {\dot{T}}_{i} (if ({\dot{T}}_{i} \leq G_{\min}))

(3)

where,

G_{\max}

is the maximum value of temperature rate in the hot region, °C/s; if

{\dot{T}}_{i}

is more than or equal to

G_{\max}

{\dot{T}}_{i}

is assigned to

G_{\max}

;

G_{\min}

is the minimum value of temperature rate in the cold region, °C/s; if

{\dot{T}}_{i}

is less than or equal to

G_{\min}

{\dot{T}}_{i}

is assigned to

G_{\min}

Geometric characteristics

Geometric characteristics describe the height, width, area and perimeters of hot and cold regions for true and false sticker breakouts in mould thermal image. It provides an intuitive representation of abnormal regions in two-dimensional space. Height, width and area can be calculated by the following formulas (4), (5) and (6). According to the threshold segmentation algorithm, there is an obvious boundary between the hot region and cold region. If there is a cold region doped in the hot region geometric parameters can be also calculated correctly. When a breakout alarm occurs, both the hot region and the cold region have similar morphologies as shown in Figure 4.

Height (H): the height is used to describe the number of pixel points in the vertical direction of hot or cold region in the mould. The height can be calculated by formula (4). The height of the hot region is denoted as $H_{h}$ , and the cold region is denoted as $H_{c}$ , as shown in Figure 4(a).

H = Y_{top} - Y_{down}

(4)

Width (W): the width is used to describe the number of pixel points in the horizontal direction of hot or cold region in the mould. It can be calculated by formula (5). The width of hot region is denoted as $W_{h}$ , and the cold region is denoted as $W_{c}$ , as shown in Figure 4(b).

W = X_{right} - X_{left}

(5)

Area (S): the area is used to describe the number of pixels occupied by cold or hot region in the mould thermal image, it can be calculated by formula (6). When the pixel $(x, y)$ is located inside the abnormal region, the value of $f (x, y)$ is 1; otherwise it is 0. As shown in Figure 4(c), the area of hot region is denoted as $S_{h}$ and the cold region is denoted as $S_{c}$ .

S = \sum_{y = Y_{top}}^{Y_{down}} \sum_{x = X_{left}}^{X_{right}} f (x, y)

(6)

Perimeter (P): the perimeter is used to describe the outer boundary contour of cold or hot region, calculated by the sum of adjacent boundary pixels. The perimeter of the hot region is denoted as $P_{h}$ , and the perimeter of the cold region is denoted as $P_{c}$ , as shown in Figure 4(d).

Shape characteristics

The hot and cold regions have some distinct differences for true and false sticker breakouts in shape. Therefore, the characteristics of circularity, rectangularity and aspect ratio were calculated.

Circularity: the circularity means whether a shape approaches a circle. It can be calculated by equation (7).

C = \frac{4 π A}{P^{2}}

(7)

where, A is the area of a hot or cold region; P is the perimeter. The circularity ranges from 0 to 1. When the value is close to 1, the shape is close to a circle.

Rectangularity: the rectangularity means that a shape approaches a rectangle. It can be calculated by equation (8).

R = \frac{W \times H}{A}

(8)

where, W is the width; H is the height; A is the area. When the value of rectangularity is close to 1, the shape is close to a square.

Aspect ratio: it describes the elongation or breadth of a region as represented by equation (9). The value is less than 1, a more elongated image. If the value is greater than 1, it is a more expansive image.

A R = \frac{W}{H}

(9)

Analysis of the hot and cold regions’ characteristics

Maximum (minimum) temperature rate

Figure 5 shows the maximum and minimum temperature rates of hot and cold regions for true and false sticker breakouts. As shown in Figure 5(a), the maximum temperature rate of hot region is mainly concentrated in 1.06–7.13 °C/s for true sticker breakout, while the range is 0.28–5.27 °C/s for false sticker breakout. It can be seen that the maximum temperature rate of hot region is often higher for true sticker breakout than false sticker breakout. When the maximum temperature rate is less than 1.06 °C/s, the true sticker breakout did not occur. The minimum temperature rate of the cold region is −4.31 °C/s for true sticker breakout, while the minimum temperature rate of the cold region is −3.43 °C/s for false sticker breakout. As shown in Figure 5(b), the cooling rate of cold region is also lower for true sticker breakout than false sticker breakout. This is mainly due to the repeated tearing and healing of sticker region, which leads to the thick strand shell. At the same time, the sticker point leaves the former position. Therefore, the local temperature will decrease from the highest point rapidly. The combined effect leads to a faster cooling of cold region for sticker breakout.

Geometric characteristics

Height

Figure 6 shows the height characteristics of hot and cold regions for true and false sticker breakouts. As shown in Figure 6(a), the heights of the hot region are between 80 and 181 pixels for true sticker breakout, located in the lower–middle part in mould, while this value of false sticker breakout is between 28 and 181 pixels. It means that the height cannot be very small for true sticker breakout. As shown in Figure 6(b), the height of the cold region is between 34 and 80 pixels for true sticker breakout, located in the upper-middle part of the mould. The height of the cold region is between 15 and 123 pixels for the false sticker breakout. It means that the height of the cold region is more concentrated for true sticker breakout. Based on the formation mechanism of sticker breakout, it usually occurs near the meniscus. Due to the effect of vibration and continuous casting in the mould, strand shell is torn and healed repeatedly. A cold region appears at the top of the sticker point, which is usually small and more concentrated in the vertical direction.

Width

Figure 7 shows the width characteristics of hot and cold regions for true and false sticker breakouts. As shown in Figure 7(a), the width of the hot region is between 80 and 316 pixels for true sticker breakout, and the actual propagation distance is approximately 400–1580 mm in the mould. While the width of the hot region ranges from 19 to 423 pixels for the false sticker, and the actual propagation distance is approximately 95–2115 mm in mould. The horizontal propagation of the hot region exceeds two rows of thermocouples for sticker breakout, while it is less than one row of thermocouples for false sticker breakout. In Figure 7(b), the width of the cold region ranges from 14 to 182 pixels for the true sticker breakout, and the actual propagation distance is approximately 70–910 mm in the mould. While this value of false sticker breakout is in the range of 18–515 pixels, and the actual propagation distance is approximately 90–2575 mm in the mould. The minimum horizontal propagation distance of cold region of true sticker breakout is similar to that of false sticker breakout, but the maximum propagation distance is significantly less than that of the false sticker breakout, indicating that the cold region of the true sticker breakout is more concentrated.

Area

Figure 8 shows the area characteristics of hot and cold regions for true and false sticker breakouts. As shown in Figure 8(a), the area of hot region is between 1559 and 24 998 pixels for the true sticker breakout, while the false sticker breakout is between 699 and 39 664 pixels. The area of the hot region is more concentrated for the true sticker breakout than that of the false sticker breakout. When the area of the hot region exceeds 24 998 pixels, no sticker breakout has occurred. But there are 10 samples of false sticker breakout, which were mainly caused by process operations, such as startup of continuous casting and quick change of submerged nozzle. In Figure 8(b), the area of the cold region is between 519 and 12 743 pixels for the true sticker breakout, and the false sticker breakout is between 249 and 25 342 pixels. The area of the cold region is smaller and more concentrated for the true sticker breakout. When the area exceeds 12 743 pixels, no sticker breakout has occurred, but there are four samples of false sticker breakout.

Perimeter

Figure 9 shows the perimeter characteristics of hot and cold regions for true and false sticker breakouts. As shown in Figure 9(a), in the mould thermal image, the perimeter of the hot region is between 326 and 1114 pixels for the true sticker breakout, and the actual propagation distance is approximately 163–5570 mm in the mould. The perimeter of the hot region is 120–1614 pixels for the false sticker breakout, and the actual propagation distance is approximately 600–8070 mm in the mould. In Figure 9(b), the perimeter of the cold region is 73–474 pixels for the true sticker breakout, and the actual propagation distance is approximately 365–2370 mm in the mould. The perimeter of the cold region is within 60–1004 pixels for the false sticker breakout, and the actual propagation distance is approximately 300–5020 mm in the mould. It should be noted that the perimeter of the true sticker breakout is more concentrated, while the perimeter of the false sticker breakout is more dispersed.

Shape characteristics

Circularity

In terms of shape, there are some significant differences between true and false sticker breakouts too. Therefore, the circularity of hot and cold regions were calculated. To show circularity changes, they are lined up from the minimum value to maximum value as shown in Figure 10. In Figure 10(a), the circularity value of the hot region is 0.17–0.52 for the true sticker breakout, which is significantly lower than the false sticker breakout at 0.13–0.87, indicating that the shape of the false sticker breakout is closer to circular in the mould thermal image. The circular of the hot region is relatively low for true sticker breakout, because true sticker breakouts have diversity. In Figure 10(b), the circularity range of the cold region is 0.41–0.89 for the true sticker breakout, which is generally higher than the false sticker breakout (0.2–0.88). The shape of the cold region is closer to a circular shape for true sticker breakout. All statistical results show that the circularity of the hot region for the true sticker breakout is lower, while this value of the cold region is higher. This is one of the main differences between true and false sticker breakouts.

Figure 11 shows the perimeter change along with circularity for true and false sticker breakouts. In Figure 11(a), the circularity of the hot region is below 0.6 for all 99 true sticker breakouts, and the perimeter decreases as the circularity increases. Among them, there are 85 samples of true sticker breakouts with circularity value below 0.4, and the perimeter is in the range of 500–1114 pixels. In Figure 11(b), the perimeter of hot and cold regions varies irregularly with circularity for false sticker breakouts. When the circularity is below 0.4, the perimeter of the hot region is in the range of 227–1614 pixels. In Figure 11(c), the circularity of the cold region is all above 0.4, but the perimeter is all below 500 pixels for true sticker breakout. These data show that the shape of the cold region is generally close to circular for true sticker breakout. Under the action of continuous casting, the region of the sticker point always moves downward to the mould outlet, and the thickness of the strand shell continuously increases without large-scale expansion. In Figure 11(d), there are 15 samples of the cold region whose circularities are below 0.4 for the false sticker breakout, which has distinct differences from the true sticker breakout that is generally above 0.4.

To further understand the circularity characteristics of true and false sticker breakouts, circularity value 0.5 was taken as the division point. The statistical analysis of circularity of hot and cold regions was carried out for true and false sticker breakouts. As shown in Figure 12, when the circularity of true sticker breakout is less than or equal to 0.5, there are 97 samples in the hot region and eight samples in the cold region. When the circularity of true sticker breakout is more than 0.5, there are two samples in the hot region and 91 samples in the cold region. The circularity value of the hot region is significantly lower than that of the cold region for true sticker breakout, further indicating that the shape of the cold region is closer to circular for the true sticker breakout. There is a distinct difference in the shapes of the hot and cold regions for true sticker breakouts. When the circularity of the false sticker breakout is less than or equal to 0.5, there are 49 samples in the hot region and 26 samples in the cold region. When the circularity of false sticker breakout is more than 0.5, there are 50 samples in the hot region and 73 samples in the cold region. The median roundness distribution of the hot region is almost equal on both sides, while the circularity of the cold region is slightly more than 0.5 for the false sticker breakout. It can be seen that circularity can be an important criterion for distinguishing between true and false sticker breakouts.

Rectangularity

To further describe the shape characteristics, the rectangularity values of hot and cold regions were calculated for true and false sticker breakouts. As shown in Figure 13(a), the rectangularity of the hot region is in the range of 0.3–0.8 for the true sticker breakout, which is generally lower than false sticker breakout (0.26–0.94), indicating that the shape of the hot region is closer to a rectangular for the false sticker breakout. In Figure 13(b), the rectangularity of the cold region ranges from 0.5 to 0.96 for the true sticker breakout, which is generally higher than the false sticker breakout (0.27–0.95), indicating that the shape of the cold region is closer to rectangular for the true sticker breakout.

Aspect ratio

Aspect ratio is used to describe the proportional relationship between the width and length of a region, which is a ratio between horizontal pixel and vertical pixel of a region. Therefore, the aspect ratio of true and false sticker breakouts were calculated and drawn along with rectangularity for true and false sticker breakouts. In Figure 14(a), the linear fitting slope of hot region is K = 2.73 for the true sticker breakout. There are 82 samples whose rectangularity values are between 0.3 and 0.61, and aspect ratio values are between 0.44 and 1.89. There are some samples with large rectangularity and aspect ratio, because there is a segmentation phenomenon of hot region as shown in Figure 15(a). The rectangularity and an aspect ratio of the hot region are 0.72 and 2.32, which deviates from the concentrated region of the true sticker breakout. More attention should be paid to such phenomena during the image processing. This is different from Figure 15(b) and (c). As shown in Figure 14(b), the linear fitting slope is K = 1 for the cold region of the true sticker breakout. The cold region mainly concentrates between rectangularity 0.78 and 0.93 and the aspect ratio 0.74–2.29. In Figure 15(a), the rectangularity and aspect ratio of the cold region are 0.7 and 2.09, respectively. While in Figure 15(b) and (c), the rectangularity values of the cold region are 0.84 and 0.91, respectively, and aspect ratios are 1.11 and 1.94, respectively. Those have rectangular characteristics in the mould thermal image and less shape differences. Figure 14(c) and (d) shows the aspect ratio change along with rectangularity for the false sticker breakout. The rectangularity ranges from 0.26 to 0.94 and the aspect ratio ranges from 0.37 to 4.12 in the hot region, while the rectangularity ranges from 0.27 to 0.95 and the aspect ratio ranges from 0.59 to 4.77 in the cold region. The fitted slope values are −0.41 and −0.44, respectively. When the slope is negative, the aspect ratio of the hot and cold regions decreases with an increase of rectangularity for the false sticker breakout. Because there is no horizontal propagation phenomenon for false sticker breakout, the width of the false sticker breakout remains small even with a larger rectangularity. Therefore, the relationship between rectangularity and the aspect ratio can be used as an important basis for judging true and false sticker breakouts.

Identification of true and false sticker breakouts

Based on the statistical results in Table 2, the characteristic values of the hot and cold regions are listed and summarised for the true and false sticker breakouts:

Temperature rate characteristics: the heating and cooling rates of the true sticker breakout are greater than those of the false sticker breakout.

Geometric characteristics: except the width of the cold region, the height, width, area and perimeters of the hot and cold regions are more concentrated for the true sticker breakouts. When these geometric parameters exceed the range of the characteristic value, there is no true sticker breakout, which can be used as an important basis to recognise the false sticker breakout.

Shape characteristics: the circularity, rectangularity and aspect ratio of the hot region are more concentrated for true sticker breakout, whose characteristic values are in the range of false sticker breakout. However, circularity, rectangularity and aspect ratio of the cold region for the true sticker breakout have a cross with those of the false sticker breakout.

Table 2.

Statistical analysis of the temperature rate, geometry and shape characteristics for 198 true and false sticker breakouts.

Item		True sticker breakout		False sticker breakout
Item		Hot region	Cold region	Hot region	Cold region
Temperature rate	$G_{m}$ (°C/s)	1.06–7.13	−0.35 to −4.31	0.28–5.27	−0.28 to −3.43
Geometric characteristics	H (pixels)	80–181	34–80	28–181	15–123
	W (pixels)	80–316	14–182	19–423	18–515
	S (pixels)	1559–24 998	519–12 743	699–39 664	249–25 342
	P (pixels)	326–1114	73–474	120–1614	60–1004
Shape characteristics	C	0.17–0.52	0.41–0.89	0.13–0.87	0.2–0.88
	R	0.3–0.8	0.5–0.96	0.26–0.94	0.27–0.95
	AR	0.44–2.69	0.4–3.14	0.37–4.12	0.59–4.77

Conclusion

Temperature rate characteristics: the maximum temperature rate of the hot region is between 1.06 and 7.13 °C/s for the true sticker breakout, while the false sticker breakout is between 0.28 and 5.27 °C/s. The minimum temperature rate of the cold region is −4.31 °C/s for the true sticker breakout, lower than the false sticker breakout −3.43 °C/s. Compared to the false sticker breakout, the true sticker breakout has a greater temperature increase of the hot region while a larger temperature decrease of the cold region.

Geometric characteristics: the heights of the hot and cold regions for true sticker breakouts are between 80–181 and 34–80 pixels, respectively, while the heights of hot and cold regions for false sticker breakouts ranges from 28 to 181 and 15 to 123 pixels, respectively. The width of the cold region for true sticker breakout is between 14 and 182 pixels, and the width of the cold region for false sticker breakout is between 18 and 515 pixels. In addition to the width of the cold region, the height, width, area and perimeter are more concentrated in both the hot and cold regions for true sticker breakouts, which are all in the range of characteristic values of the false sticker breakout. When these geometric characteristic values exceed, there is no sticker breakout, which can be used as an important basis for recognising false sticker breakout.

Shape characteristic: the circularity of the hot region does not exceed 0.5 for 97 samples true sticker breakouts, which is lower than the false sticker breakout. The circularity of the cold region is more than 0.4, and there are 91 samples whose values are more than 0.5 for the true sticker breakouts. The propagation of the hot region shows an irregular shape, while the shape of the cold region is close to circular for the true sticker breakout. The linear regression slope values of the rectangularity and the aspect ratio for the true sticker breakout are K = 2.73 of the hot region and K = 1 for the cold region. The true sticker breakout has an expansion in the horizontal direction, while the false sticker breakout does not have horizontal expansion, whose linear regression slope is negative.

Footnotes

Acknowledgements

This research was funded by the National Natural Science Foundation of China (51704073/51974056) and Science and Technology Research Project of Jilin Provincial Education Department (JJKH20230111KJ).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship and/or publication of this article: This work was supported by the Science and Technology Research Project of Jilin Provincial Education Department, National Natural Science Foundation of China (grant numbers: JJKH20230111KJ, 51704073/51974056).

ORCID iDs

Yu Liu

Yali Gao

Data availability

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

References

Qian

Xie

, et al. Recovery mechanism of the sticking-type breakout during continuous casting of steel. Ironmak Steelmak 2019; 46: 259–268.

Wang

Zheng

Zhu

, et al. An integrated production batch planning approach for steelmaking-continuous casting with cast batching plan as the core. Comput Ind Eng 2022; 173: 108636.

. Formation mechanism and influence factors of the sticker between solidified shell and mold in continuous casting of steel. High Temp Mat Pr-isr 2019; 38: 192–198.

Qin

Zhu

Zheng

, et al. Molten steel breakout prediction based on thermal friction measurement. J Iron Steel Res Int 2011; 18: 24–35.

Zhang

Han

Liu

, et al. Effect of mineralogical structure of flux film on slab quality for medium carbon steel. Trans Indian Inst Met 2018; 71: 1803–1807.

Fang

Ding

, et al. Analysis of mold friction in a continuous casting using wavelet transform. Metall Mater Trans B 2018; 49: 558–568.

Paswan

Ansu

Adeleke

, et al. Understanding casting behaviour of low carbon high manganese steel through detailed characterization of mould powder and mould top slag. Ironmak Steelmak 2023; 50: 286–294.

Wang

Zhou

Zhang

. Review of artificial intelligence algorithms-based wind turbine condition monitoring and fault diagnosis techniques. J Northeast Electr Power Univ 2024; 44: 42–51.

Zhang

, et al. Investigation on non-uniform friction behaviors of slab during continuous casting based on an inverse algorithm. J Mater Process Technol 2021; 288: 116871.

10.

Ray

Basak

. Local heat flux profiles and interfacial thermal resistance in steel continuous casting. J Mater Process Tech 2018; 255: 605–610.

11.

Zhang

Sheng

, et al. Application of GA-ACO algorithm in thin slab continuous casting breakout prediction. Trans Indian Inst Metals 2023; 76: 145–155.

12.

Liu

Wang

, et al. Recognition of sticker breakout based on different propagation behaviors and template matching. Ironmak Steelmak 2023; 50: 94–100.

13.

Zhang

Liu

Shen

, et al. A review of ensemble learning algorithms used in remote sensing applications. Applied Sciences 2022; 12: 8654.

14.

Wang

Yao

. Integrated model of ACWGAN-GP and computer vision for breakout prediction in continuous casting. Metall Mater Trans B 2022; 53: 2873–2883.

15.

Duan

Wang

Bai

, et al. Modeling of breakout prediction approach integrating feature dimension reduction with K-means clustering for slab continuous casting. Int J Adv Manuf Technol 2020; 109: 2707–2718.

16.

Zhou

Deng

. Novel mold breakout prediction and control technology in slab continuous casting. J Process Control 2015; 29: 1–10.

17.

Duan

Wang

Bai

, et al. Integrated approach to density-based spatial clustering of applications with noise and dynamic time warping for breakout prediction in slab continuous casting. Metall Mater Trans B 2019; 50: 43–53.

18.

Bartocha

Wróbel

Szajnar

. The influence of casting velocity on structure of Al continuous ingots. Arch Metall Mater 2017; 62: 1609–1613.

19.

Singh

Das

, et al. Mathematical model and plant investigation to characterize effect of casting speed on thermal and solidification behavior of an industrial slab caster. ISIJ Int 2018; 58: 2308–2317.

20.

Wang

Liu

, et al. Influence of mould curvature on the metallurgical performances of beam blank continuous casting. Ironmak Steelmak 2020; 47: 929–941.

21.

Liu

Tian

Wang

, et al. Influence of processing parameters on slab stickers during continuous casting. High Temp Mat Pr-isr 2020; 39: 228–235.

22.

Ito

Nabeshima

Miki

, et al. Evaluation of solidified shell thickness by thermocouple in mold. ISIJ Int 2019; 52: 2239–2246.

23.

Liu

Wang

Sun

, et al. Research on a new detection method of slab surface crack in mould during continuous casting. Metall Res Technol 2018; 115: 08.

24.

Liu

Wang

, et al. Space-time characteristics of true and false sticker breakouts in mould during continuous casting. Ironmak Steelmak 2021; 48: 901–908.

25.

Deng

, et al. Development and application of mould breakout prediction system with online thermal map for steel continuous casting. Ironmak Steelmak 2015; 42: 194–208.

26.

Liu

Wang

Yao

, et al. Visualization and recognition method of sticker breakout of mold during continuous casting process. J Iron Steel Res Int 2015; 27: 37–41.

27.

Liu

Wang

, et al. Computer vision detection of mold breakout in slab continuous casting using an optimized neural network. Int J Adv Manuf Technol 2017; 88: 557–564.

28.

Xin

. An improved FLC frequency modulation control strategy with adaptive ability. J Northeast Electr Power Univ 2023; 43: 55–62.

29.

Zhang

Chen

. Vienna rectifier fault diagnosis based on CEEMD feature extraction and optimized RF classification. J Northeast Electr Power Univ 2023; 43: 23–31.

30.

Duan

Wei

, et al. Longitudinal crack detection approach based on principal component analysis and support vector machine for slab continuous casting. Steel Res Int 2021; 92: 2100168.

31.

Duan

Wang

Bai

, et al. Application of k-means clustering for temperature timing characteristics in breakout prediction during continuous casting. Int J Adv Manuf Tech 2020; 106: 4777–4787.

32.

Liu

Wang

Yao

, et al. Effect of casting parameters on sticker breakout and its propagation behaviour during slab continuous casting. Ironmak Steelmak 2014; 41: 748–755.

33.

Cui

Wang

Yang

, et al. Phase equilibria of fluoride-free boracic mould flux for steel continuous casting. Steel Res Int 2015; 86: 662–669.

34.

Liu

Wang

, et al. Visual detection based on computer vision for sticker breakout in slab continuous casting. Ironmak Steelmak 2015; 42: 417–423.