Regular Article
Molten Steel Level Measurement Based on Temperature Attenuation Characteristic
2016 Volume 56 Issue 12 Pages 2208-2213
Details
2016 Volume 56 Issue 12 Pages 2208-2213
This paper proposes a molten steel level measuring method in a tundish. This method is based on the temperature attenuation characteristic of the temperature tube. Our previous work focused on the temperature distribution of the temperature tube which can be used for molten steel level measurement. In some continuous casters, sticky slag adheres to the tube, thus, blocking the temperature information on the tube. To avoid this problem of lacking true temperature, we analyzed the principle of adhesion and conclude that the adhesive slag thickness of the tube is influenced by whether that part of the tube contacts the slag or the molten steel prior to being lifted. Consequently, for detecting the adhesive slag thickness online with available pyrometer, a heat transfer model is established. Through analysis of the model, it was revealed that the thickness of the adhesive slag remarkably influences the measurable maximum curvature time of the temperature attenuation. In addition, this maximum curvature time is insensitive to other irrelevant factors such as initial temperature and ambient temperature. Therefore, the maximum curvature time of the temperature attenuation of the adhesive slag indirectly indicates the molten steel level. Finally, the experiments prove the feasibility and the sound accuracy of this method with the maximal error being 3 mm. This method is not only suited for the molten steel level measurement, but also applicable to other non-destructive measurement of coating layer thickness.
In the continuous casting of steel, the fluctuation of molten steel level in a tundish directly results in the instability of some key casting conditions including the flow field in the tundish and the crystallizer pressure. These unstable casting conditions will eventually cause the quality of the final product to deteriorate.1,2,3) Fearing quality deterioration, many steel makers wish to precisely control the molten steel level in the tundish. However, as the thickness of the impenetrable slag cannot easily be measured, most steel makers fail to acquire this level of precision. Currently, the majority of steel makers estimate this level with the tundish weight. However, this indirect and rough estimation seriously suffers from the influence of inconsistent weight of slag and irregular impact force given by the molten steel from the ladle. Additionally, the sophisticated inner structure of the tundish further complicates this estimation. Verified with dangerous but relatively accurate manual measurements, the accuracy of the weighting method can be quite frustrating, with the error reaching up to ±30 mm.
To settle this problem, many specialists have attempted various techniques. These previously attempted techniques include some techniques resorting to radioactive, eddy current, response frequency, and microwaves. Radioactive measurement of molten steel level is quite effective. However, though much effort has been done for lowering the radioactivity levels,4,5) the operator’s exposure to radioactive substances still cannot be dismissed. In POSCO, a safer technique, based on the eddy current, has been employed to obtain the mold level in crystallizer, in which its circumstance is similar to the tundish. This method generates the level in high accuracy, however, the measurement range reaches merely 0–300 mm,6,7,8,9) which is too short for the molten steel level in tundish. Through a similar principle, the AGELLIS Company proposes a longer range measurement system for the molten steel level in a tundish. However, as the service life is only about 2 years and the total cost is in excess of ＄100000, this technique has not been widely accepted by the steel maker. In contrast, the technique based on the vibration frequency response of the device10,11) is extremely affordable. However, the resolution of this technique cannot meet the accuracy requirement of the tundish. Microwave radar has been imported for this problem as well.12,13,14) Through the use of sweeping mirrors, this technique simultaneously satisfies the requirement of accuracy and long range for the slag level measurement. However, currently, this technique encounters the difficulty to detect the steel level since the echo which has passed through the non-uniform slag is quite noisy.
In our previous work,15,16,17) we proposed a molten steel level measurement technique based on the vertical temperature distribution of the tundish. The principle of this technique is illustrated in Fig. 1. This system inserts a temperature tube into the molten steel and lifts it up every 10 minutes. During the lifting process, a CCD pyrometer is employed to retrieve the temperature information of the tube. Combining the temperature information and the heat transfer model, the system acquires the vertical temperature distribution in the tundish. By analyzing the curvature of the vertical temperature distribution, this system precisely locates the molten steel level. Based on the above system, monocular vision18) and laser device19) plugins are introduced to acquire the slag level in real-time. Assuming the thickness of the slag is constant for 10 minutes, the system obtains the molten steel level in real time by the last molten steel level and the fluctuation of the slag level during the period.
Molten steel level measurement based on thermal image analysis.
However, during the product popularization, this system encounters a new problem. This problem is due to the fact that various kinds of slag in certain steel continuous casters is extremely sticky. Consequently, during the lifting process, there is a certain amount of slag adhered to the surface of the temperature tube. Due to the enhanced temperature retention of the slag, the surface temperature acquired by the CCD pyrometer is much different from the true temperature which is required for the analysis. Therefore, the measuring principle must be reestablished to fit such a new circumstance.
This paper proposes a novel molten steel level measurement method based on the time characteristic of the temperature attenuation. First, the paper analyzes the principle of the slag adhesion, pointing out that the thickness of adhesion depends on the surface micro-structure of the tube. In addition, the surface micro-structure of the tube depends on whether it comes into contact with the molten steel or the liquid slag prior to being lifted. Second, according to the analysis of the heat transfer model, the maximum curvature time of the temperature attenuation is a sound indicator for the thickness of adhesive slag, since the maximum curvature time is sensitive to thickness of adhesive slag and insensitive to the irrelevant parameters such as the ambient temperature and the initial temperature. Third, with the temperature data obtained in the tundish, this paper proves the validity of this method with the experimental results. In conclusion, a series of verifications with manual measurements is provided. These measurements demonstrate the repetitiveness and the accuracy of this method, with the maximal error below 3 mm.
The thickness distribution of adhesive slag in a used temperature tube can be analyzed to determine the relative position of the tube and the molten steel level before the tube is lifted. The Schmitt’s model offers a theoretical explanation for this phenomenon.20) The degree to which the slag sticks to the temperature tube is mainly dependent on slag and tube material composition, as well as the micro-structure on the surface of the temperature tube. Temperature tubes are made of about 75% Al2O3, 20% graphite, and 5% carbonaceous binder; slag is mainly comprised of CaO, MgO, and SiO2, but at individual proportions that differ considerably across different steel-making plants. The materials of slag are mutually reactive - they ultimately take the form of silicates which adhere to Al2O3.
During casting, the Al2O3 of the tube slightly solutes in the slag at the border of the tube and the slag above the molten steel level, leaving non-stick, graphite-rich structures on the surface of the tube. Conversely, below the molten steel level, Al2O3 is insoluble to the molten steel to which the graphite is soluble - this physical feature leads to the formation of 70-um, Al2O3-rich micro-structures with relatively little graphite at the border of the molten steel and the tube.21) In effect, during the lifting process, there is relatively little slag adhered at the position of the tube above the molten steel level and relatively thick slag at the position originally under the steel level. Accordingly, the molten steel level can be accurately located by measuring the adhesive thickness distribution.
We established a heat transfer model to investigate the relationship between surface temperature characteristics and adhesive slag thickness, then utilized the model to detect the thickness of the adhesive slag with a commercially available CCD pyrometer. There are two stages to the heat transfer model: “pre-lifting”, and “lifting”.
In the first stage, the temperature tube was placed at the location shown in the left side of Fig. 1 until reaching thermal balance. The boundary conditions included vertical slag thickness of 50 mm and outer and inner tube radius of 40 mm and 15 mm, respectively; the model was asymmetrical about the y axis, ambient temperature was 600°C, molten steel temperature was 1530°C, heat convection between the air and the temperature tube was 8 W/m2, heat convection between the slag and the tube was 3000 W/m2, and heat convection between the molten steel and the tube was 3000 W/m2. Temperature tube and slag emissivity were both 0.8. Detailed thermal physical parameters of the materials are listed in Table 1.
| Density (kg/m3) | Specific heat (J/kg·K) | Thermal conductivity (W/(m·K)) | |
|---|---|---|---|
| Tube | 2900 | 1000 | 8.7 |
| Slag | 2500 | 1170 | 1.0 |
In the second stage, the temperature tube was lifted up 100 mm step-wise at a speed of 25 mm/s over the course of 4 s. In each step, the initial temperature distribution of the step is loaded from the end of the previous step and the contact boundary is recalculated. During our experiment, to simulate adhesion, 1 mm adhesive slag was set to rise with the tube on the original interface between the adhering slag and the tube, and 2 mm adhesive slag stuck to the part of the tube immersed in the liquid steel prior to lifting.
The final state of the temperature distribution is illustrated in Fig. 2. After retrieving the temperature data from typical locations A, B, C, D, E and F after they were exposed to air, the data was used to draw the temperature attenuation curves shown in Fig. 3. The curves of locations A, B and C, which had identical adhesive thickness, share a very similar pattern but are remarkably different from the curves of locations D, E and F, which had different adhesive thicknesses. All of the curves did share an important common characteristic: That each temperature attenuation curve is comprised of two parts, 1) where the surface temperature drops dramatically due to the space radiation, observing from the CCD pyrometer, seemingly only the adhesive slag (which has low thermal conductivity) takes part in the procedure, and 2) where the surface temperature drops slowly and steadily as a result of thermal conduction between temperature tube and slag, as temperature tube (which has a high specific heat) dominates the procedure. There is a clear maximum curvature point between the two parts.
Temperature distribution at the end of the lifting process.(Online version in color.)
Temperature attenuation curves of typical locations. (Online version in color.)
As illustrating in Fig. 4, we calculated the curvature of the curves in Fig. 3 after 6 order polynomial fitting and their corresponding maximum to locate the precise position of the maximum curvature point. The maximum curvature point of 2-mm adhesive slag was apparently later than that of 1-mm adhesive slag. We alternated different adhesive thicknesses and picked the points with same initial temperature after being exposed to air, and then were able to observe the relation between the adhesive thickness and the location of the maximum curvature point more intuitively as shown in Fig. 5. Similarly, we calculated the curvatures of these curves and found their maximum curvature points as shown in Fig. 6. Obviously, the maximum curvature time increases with the adhesive thickness. Although the maximum curvature point of 4-mm adhesion is not clear, in most cases, the adhesive thickness falls into the range of 0.5 mm to 2.0 mm. Accordingly, maximum curvature time of temperature attenuation serves as an indicator of adhesion thickness.
Temperature attenuation curvature of typical locations. (Online version in color.)
Temperature attenuation curves of varying adhesive thickness. (Online version in color.)
Temperature attenuation curvatures of varying adhesive thickness. (Online version in color.)
For maximum curvature time to indeed be applied to detect adhesion thickness, it must be insensitive to irrelevant factors (e.g., initial temperature and ambient temperature). To ensure the necessary insensitivity, we investigated the correlation between maximum curvature time and adhesion thickness, initial temperature, and ambient temperature: First, we conducted five simulations at ambient temperature of 600°C and adhesion thickness of 0 mm, 0.5 mm, 1 mm, 1.5 mm, and 2 mm, respectively, then searched for the position at which initial temperature was 1300°C and calculated the maximum curvature time in each of the five temperature attenuation curves (Column 1, Table 2). Next, we found the position at which initial temperature was 1200°C among the five simulations and calculated the maximum curvature time in each of the five temperature attenuation curves (Column 2, Table 2). Finally, we conducted another five simulations at ambient temperature of 400°C, and similar to the first step, searched for the position at which initial temperature was 1300°C and calculated the maximum curvature time in each of the five temperature attenuation curves (Column 3, Table 2).
| Adhesion Thickness (mm) | Initial temp 1300°C | Initial temp 1200°C | Initial temp 1300°C |
|---|---|---|---|
| Ambient temp 600°C | Ambient temp 600°C | Ambient temp 400°C | |
| 0 | 0.056 s | 0.055 s | 0.057 s |
| 0.5 | 0.56 s | 0.55 s | 0.57 s |
| 1 | 0.80 s | 0.77 s | 0.83 s |
| 1.5 | 1.14 s | 1.10 s | 1.18 s |
| 2 | 1.42 s | 1.34 s | 1.49 s |
Figure 7 illustrates the relation between initial temperature, ambient temperature, maximum curvature time, and adhesive thickness. There was sound consistency between adhesion thickness and maximum curvature time. At initial temperature of 1300°C and ambient temperature of 600°C, while the true initial temperature had an offset of 100°C or the actual ambient temperature deviated by 200°C, the error of the detected thickness did not exceed 6%. This accuracy is sufficient for molten steel level measurement because the variation in adhesion thickness between the slag layer and the molten steel layer usually reaches the scale of 200% (or higher,) therefore, the maximum curvature time of temperature attenuation fully satisfies the necessary requirements of using the adhesion thickness indicator for molten steel level measurement.
Maximum curvature time increases with adhesion thickness, initial temperature, ambient temperature. (Online version in color.)
The data measured according to the tundish was in accordance with the data acquired through simulations. During the lifting procedure, we used a CCD thermal imager to obtain a series of thermal images of the lifting temperature tube; Fig. 8 shows the final image taken, and Fig. 9 shows the temperature attenuation data gathered for two typical points (A and B in Fig. 8) and Fig. 10 shows the curvatures and maximum curvature time of them. After filtering the data by 6-order polynomial fitting, the patterns of the temperature attenuation curves appeared extremely similar to those of the simulation, again with a clear maximum curvature point between the two parts of each temperature attenuation curve. According to Fig. 7, the adhesive thicknesses on points A and B are 1.0 mm and 2.0 mm, both of which coincide with our measurement of the temperature tube after it was cooled down and destructed.
Thermal image of lifted temperature tube.
Temperature attenuation curve of 2 typical locations. (Online version in color.)
Temperature attenuation curvatures of 2 Typical locations. (Online version in color.)
By converting all the temperature attenuation data to adhesive thickness data via the method described above, we obtained the adhesive thickness distribution shown in Fig. 11; a column of vertical adhesive thickness distribution after geometric correction is also given in the right part of the figure. By comparing the manually gathered molten steel level and the level acquired by above method, we confirmed that the step of adhesive thickness appearing in the distribution precisely indicated the molten steel level (at which over 75% of the pixels have a step).
Adhesive thickness distribution of temperature tube.
A lot of replications of these experiments showed that the phenomenon of the thickness step consistently occurred on the temperature tube. Further, the thickness step matched the molten steel level precisely and constantly as long as the temperature tube was inserted into the molten steel for more than five minutes. Table 3 shows some of the typical verification data we gathered from the experiments, where the maximal error of the proposed molten steel level measurement method was 3 mm.
| System measurement | Manual measurement |
|---|---|
| 768 mm | 766 mm |
| 802 mm | 804 mm |
| 780 mm | 779 mm |
| 844 mm | 841 mm |
This paper describes a novel method of measuring molten steel level based on the temperature attenuation characteristics of adhesive slag. Based on the readily observable and repeatable phenomenon of the adhesion thickness step, this method acquires the molten steel level precisely and reliably; according to the results of our experiments, the error of this method is less than 3 mm. The proposed method solves the problem of adherent slag interference to the molten steel level measurement system, and is applicable not only to molten steel level measurement, but also to other non-destructive measurement of paint-coat thickness.
This study was supported by the State Key Laboratory of Synthetical Automation for Process Industries (2013ZCX06), Fundamental Research Funds for the Central Universities (N140404022) and Natural Science Foundation of Liaoning Province (2015020085).
