Temperature estimation system, hot-dip galvanizing facility, temperature estimation method, and hot-dip galvanizing method

Abstract

To provide a temperature estimation system, hot-dip galvanizing equipment, a temperature estimation method, and a hot-dip galvanizing method capable of acquiring a temperature distribution of the whole hot-dip zinc including a region away from a position of a thermometer installed in a hot-dip galvanizing pot.SOLUTION: The temperature estimation system is a temperature estimation system for estimating a temperature distribution of molten zinc inside a hot-dip galvanizing pot by a machine learning model, wherein the machine learning model is generated using training data in which a plurality of sets of positions where a plurality of thermometers are installed and temperatures of the molten zinc at the positions are input values and the temperature distribution of the molten zinc corresponding to the input values is an output value, the plurality of thermometers are installed in a plurality of regions including at least a first region and a second region along the flow path of the molten zinc generated by the conveyance of the steel sheet, the first region is located on the upstream side of the flow path from the second region, and the second region is included in the bottom surface downstream from the inductor.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to a temperature estimation system, a hot-dip galvanizing facility, a temperature estimation method, and a hot-dip galvanizing method. [Background technology]

[0002] In a hot-dip galvanizing pot for galvanizing using molten zinc, substrate components (e.g., iron) eluted from a substrate (e.g., steel sheet) passing through the molten zinc bath react with the zinc or aluminum in the bath to produce dross, an intermetallic compound. Because the adhesion of dross to the substrate steel sheet causes surface defects, it is important to control the conditions in the molten zinc bath in the plating pot to suppress the generation of dross.

[0003] Dross is said to be more likely to occur when the temperature of the molten zinc bath fluctuates. Therefore, one effective method of suppressing dross generation is to suppress bath temperature fluctuations and maintain it within an appropriate temperature range (for example, Patent Document 1).

[0004] In order to properly control the bath temperature, it is necessary to perform temperature control based on the bath temperature (specifically, heat input from an inductor installed on the side of the pot). Several methods for acquiring the bath temperature have been proposed in the previously cited documents. Patent Document 1 proposes a method in which one temperature sensor is installed above and one below the front surface of the bath, and the respective temperatures are used as control indices. Patent Document 2 proposes a method in which temperature sensors are installed in at least two locations, near the ingot insertion position and near the steel sheet entry position, and the respective temperatures are used as control indices. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2001-107208 [Patent Document 2] Japanese Patent Application Laid-Open No. 2013-234377 Summary of the Invention [Problem to be solved by the invention]

[0006] The temperature of the molten zinc varies over time and space due to the influence of various factors such as heat input from the inductor, heat input from the steel sheet, and the formation of a circulating flow due to the passing of the sheet.

[0007] In Patent Document 1, the temperature sensor position is determined taking into consideration the influence of the strip passing speed, but does not consider the influence of heat input factors such as an inductor (heating means).In addition, in Patent Document 2, the temperature sensor installation position is determined taking into consideration the heat extraction from the ingot and the heat input from the steel strip, but does not consider the influence of flow within the hot-dip galvanizing pot.

[0008] Therefore, conventional techniques are limited to obtaining the local temperature where the thermometer is located, and it is difficult to obtain the temperature distribution of the entire molten zinc, including fluctuations or temperature drops that can lead to the generation of dross.

[0009] In view of the above circumstances, an object of the present disclosure is to provide a temperature estimation system, hot-dip galvanizing equipment, a temperature estimation method, and a hot-dip galvanizing method that make it possible to acquire the temperature distribution of the entire molten zinc, including areas distant from the position of a thermometer installed in a hot-dip galvanizing pot. [Means for solving the problem]

[0010] (1) A temperature estimation system according to an embodiment of the present disclosure includes: A temperature estimation system for estimating the temperature distribution of molten zinc inside a hot-dip galvanizing pot, a plurality of thermometers installed on a wall surface of the hot-dip galvanizing pot to measure the temperature of the molten zinc at the installed positions; An estimation unit that acquires temperatures measured from the plurality of thermometers and positions of the thermometers, and estimates the temperature distribution of the molten zinc based on a machine learning model previously generated by machine learning and the acquired temperatures and positions; an output unit that outputs the estimated temperature distribution of the molten zinc, The machine learning model is generated using teacher data in which a plurality of sets of positions where the plurality of thermometers are installed and the temperatures of the molten zinc at the positions are input values, and the temperature distribution of the molten zinc corresponding to the input values ​​is output values, the hot-dip galvanizing pot includes therein a sink roll and a support roll for transporting the steel sheet, and an inductor for heating the molten zinc located on the bottom side of the sink roll, The plurality of thermometers are installed in a plurality of regions including at least a first region and a second region along a flow path of the molten zinc generated by the conveyance of the steel plate, the first region is located upstream of the second region in the flow path, The second region is included in the bottom surface downstream from the inductor.

[0011] (2) As one embodiment of the present disclosure, in (1), With respect to the wall surface connected to the bottom surface, the wall surface that collides with the flow path passing between the sink roll and the inductor is defined as the front surface, and the wall surfaces located on both sides of the flow path passing between the sink roll and the inductor are defined as side surfaces, and the first region is included in the bottom surface upstream of the front surface or the inductor.

[0012] (3) As one embodiment of the present disclosure, in (1), With respect to the wall surface connected to the bottom surface, the wall surface that collides with the flow path passing between the sink roll and the inductor is defined as the front surface, and the wall surfaces located on both sides of the flow path passing between the sink roll and the inductor are defined as side surfaces, and the first region is included in the side surfaces.

[0013] (4) The hot-dip galvanizing equipment according to an embodiment of the present disclosure includes: The hot-dip galvanizing pot includes a temperature estimation system according to any one of (1) to (3), in which the temperature distribution of the molten zinc is estimated.

[0014] (5) A temperature estimation method according to an embodiment of the present disclosure includes: A temperature estimation method executed by a temperature estimation system that estimates a temperature distribution of molten zinc inside a hot-dip galvanizing pot, comprising: measuring the temperature of the molten zinc at positions where the thermometers are installed on the wall surface of the hot-dip galvanizing pot; A step of acquiring temperatures measured from the plurality of thermometers and positions of the thermometers, and estimating a temperature distribution of the molten zinc based on a machine learning model previously generated by machine learning and the acquired temperatures and positions; and outputting the estimated temperature distribution of the molten zinc, The machine learning model is generated using teacher data in which a plurality of sets of positions where the plurality of thermometers are installed and the temperatures of the molten zinc at the positions are input values, and a temperature distribution of the molten zinc corresponding to the input values ​​is output values, the hot-dip galvanizing pot includes therein a sink roll and a support roll for transporting the steel sheet, and an inductor for heating the molten zinc located on the bottom side of the sink roll; The plurality of thermometers are installed in a plurality of regions including at least a first region and a second region along a flow path of the molten zinc generated by the conveyance of the steel plate, the first region is located upstream of the second region in the flow path, The second region is included in the bottom surface downstream from the inductor.

[0015] (6) A hot-dip galvanizing method according to an embodiment of the present disclosure includes: The surface of the steel sheet is plated using the hot-dip galvanizing pot in which the temperature distribution of the molten zinc is estimated by the temperature estimation method (5). [Effects of the Invention]

[0016] According to the present disclosure, it is possible to provide a temperature estimation system, hot-dip galvanizing equipment, temperature estimation method, and hot-dip galvanizing method that are capable of acquiring the temperature distribution of the entire molten zinc, including areas away from the position of a thermometer installed in a hot-dip galvanizing pot. [Brief explanation of the drawings]

[0017] [Figure 1] FIG. 1 is a diagram showing an example of the configuration of a hot-dip galvanizing pot in which the temperature distribution of molten zinc is estimated by a temperature estimation system according to an embodiment of the present disclosure. [Figure 2] FIG. 2 is a diagram illustrating the flow path of molten zinc generated inside the hot-dip galvanizing pot. [Figure 3] FIG. 3 is a diagram illustrating an example configuration of a temperature estimation system according to an embodiment of the present disclosure. [Figure 4] FIG. 4 is a diagram for explaining the first region and the second region. [Figure 5] FIG. 5 is a diagram showing the change in temperature estimation error depending on the number of sensors. [Figure 6] FIG. 6 is a diagram illustrating appropriate positions according to the number of sensors. [Figure 7] FIG. 7 is a flowchart of a sensor position setting method. [Figure 8] FIG. 8 is a flowchart of a method for creating training data. [Figure 9] FIG. 9 is a diagram illustrating the temperature distribution of molten zinc obtained by fluid simulation. [Figure 10] FIG. 10 is a diagram for explaining the machine learning model. [Figure 11] FIG. 11 is a flowchart of a method for generating a machine learning model. [Figure 12] FIG. 12 is a diagram illustrating the results of estimating the temperature distribution of molten zinc using the technique of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0018] Hereinafter, a temperature estimation system, a hot-dip galvanizing equipment, a temperature estimation method, and a hot-dip galvanizing method according to an embodiment of the present disclosure will be described with reference to the drawings.

[0019] (Configuration overview) FIG. 1 is a diagram showing an example of the configuration of a hot-dip galvanizing pot in which the temperature estimation system according to this embodiment estimates the temperature distribution of molten zinc. The hot-dip galvanizing pot includes a sink roll and support rolls for transporting the steel sheet, and an inductor for heating the molten zinc located below the sink roll on the bottom side. The hot-dip galvanizing pot is filled with molten zinc, through which the steel sheet continuously passes. In the example shown in FIG. 1, the steel sheet enters the bath from the top (s1), then turns around the sink roll, changes direction, and is pulled up to the upper part of the galvanizing bath (s2) via the support roll. In addition, in the example shown in FIG. 1, the inductor is installed near the bottom of the side of the bath, from which heated molten zinc is supplied.

[0020] In this embodiment, the hot-dip galvanizing facility includes a hot-dip galvanizing pot. That is, the hot-dip galvanizing pot constitutes a part of the hot-dip galvanizing facility. The hot-dip galvanizing facility performs a hot-dip galvanizing method for applying a coating treatment to the surface of a steel sheet using the hot-dip galvanizing pot.

[0021] Figure 2 is a diagram illustrating the flow path of molten zinc generated by the transportation of a steel sheet inside the hot-dip galvanizing pot of Figure 1. As the steel sheet passes through the hot-dip galvanizing pot, the molten zinc around the steel sheet is pulled by the movement of the steel sheet. As a result, a flow path (circulation flow) of molten zinc circulating inside the hot-dip galvanizing pot is formed.

[0022] From the steel sheet entry point to just before the sink roll, the circulating flow follows the steel sheet (f1). At the sink roll, the circulating flow leaves the steel sheet and flows toward the front and side of the hot-dip galvanizing pot (f2). The circulating flow then flows near the front of the hot-dip galvanizing pot (f3), near the bottom (f4), near the rear (f5), and near the top (f6) before returning to the steel sheet entry point. The walls of the hot-dip galvanizing pot are sometimes referred to as the bottom, front, side, or rear, and are shown in Figures 1 and 2. The wall surface opposite the top surface where the steel sheet enters is the bottom surface. The wall surface connected to the bottom surface, where the flow path between the sink roll and the inductor collides, is the front surface. The walls on both sides of the flow path between the sink roll and the inductor are the side surfaces. The wall surface opposite the front surface is the rear surface.

[0023] Generally, in a hot-dip galvanizing pot, thermometers (or temperature sensors, hereinafter sometimes simply referred to as "sensors") are installed discretely on the wall surface. However, it is required to estimate the temperature distribution of the entire molten zinc, including areas away from the wall surface (for example, near the center inside the hot-dip galvanizing pot). If a thermometer is installed in an area away from the wall surface, the thermometer and the members supporting the thermometer will disturb the internal temperature and flow of the molten zinc. Furthermore, from the viewpoint of facility layout and cost, it is difficult to install a large number of thermometers, and the number of thermometers installed is usually limited to a few to several tens. Therefore, it is required to install as few thermometers as possible on the wall surface.

[0024] To address these issues, the temperature estimation method implemented by the temperature estimation system according to the present embodiment uses a machine learning model (trained model) to estimate the temperature distribution, including regions away from the wall surface. As will be described in detail later, pairs of thermometer positions and temperatures at those positions are used as input variables (also referred to as input values ​​or explanatory variables) for the machine learning model. There are multiple pairs of input variables. Furthermore, the temperature distribution of the hot-dip galvanized coating is used as the output variable (also referred to as output value or target variable) for the machine learning model. The temperature estimation system according to the present embodiment enables estimation of the temperature distribution of the entire molten zinc coating, including regions away from the wall surface, by expressing the relationship between the temperatures measured by the thermometers and the temperature distribution in the target region using a machine learning model. Furthermore, to improve the accuracy of estimation using such a machine learning model, it is important to install as few thermometers as possible in appropriate locations. Therefore, the generated machine learning model is evaluated using an index (estimated contribution) that indicates the contribution of each candidate thermometer to the temperature distribution estimation. By preferentially selecting thermometer positions with higher estimated contributions, it is possible to determine the installation locations of sensors suitable for temperature distribution estimation.

[0025] (Temperature estimation system) FIG. 3 is a diagram showing an example of the configuration of a temperature estimation system according to this embodiment. The temperature estimation system includes multiple thermometers, an estimation unit, and an output unit. The multiple thermometers are installed on the wall surface of the hot-dip galvanizing pot and measure the temperature of the molten zinc at the installed positions. The thermometers may be installed so that the temperature-sensing portions are on the inner wall surface. The estimation unit acquires the temperatures measured by the multiple thermometers and their positions, and estimates the temperature distribution of the molten zinc based on a machine learning model previously generated by machine learning and the acquired temperatures and positions. As described above, the machine learning model uses multiple sets of the installed positions of the multiple thermometers and the temperatures of the molten zinc at those positions as input variables, and the temperature distribution of the molten zinc as an output variable. The output unit outputs the temperature distribution of the molten zinc estimated by the estimation unit.

[0026] Here, the estimation unit and the output unit may be configured by a device such as a computer. For example, a temperature estimation system may be configured by multiple thermometers and a computer. In this case, the device configuring the estimation unit and the output unit may not be a single device (computer), but may be configured by multiple devices located in multiple locations and capable of sending and receiving data to each other via a network. The computer may include, for example, a memory, a hard disk drive (storage device), and a CPU (processing device). The program can be stored in the hard disk drive, and when executed by the CPU, it is read from the hard disk drive to the memory. The estimation unit and the output unit may be realized by, for example, a CPU that reads and executes the program. Furthermore, the machine learning model may be stored in the storage device and read by the CPU functioning as the estimation unit.

[0027] FIG. 4 is a diagram illustrating the first and second regions. The first and second regions in FIG. 4 also correspond to those in FIG. 2. A plurality of thermometers are installed in a plurality of regions, including at least the first and second regions, along the flow path of the molten zinc. Here, the second region is included on the bottom surface downstream of the inductor. The first region is located upstream of the flow path from the second region. In terms of the flow path, the second region is defined as the region close to f4 in FIG. 2. The first region is defined as the region close to f1 to f3 in FIG. 2. The first region may be included on the front surface or the bottom surface upstream of the inductor, as shown in FIG. 4, for example. As another example, the first region may be included on a side surface. The locations of the thermometers were determined using a sensor number reduction method based on estimated contributions, which will be described later, and were installed in at least the first and second regions.

[0028] Figure 5 shows the change in temperature estimation error depending on the number of sensors. The temperature estimation error is the difference between the estimated temperature value and the actual temperature value, and is calculated using the mean square error in the example of Figure 5. As shown in Figure 5, the temperature estimation error increases when the number of sensors is two or less or 256 or more. In contrast, when the number of sensors is between four and 128, the temperature estimation error remains small and stable. The temperature estimation error reaches a minimum value when the number of sensors is 32, but when the number of sensors is between four and 128, the error increase rate compared to the minimum value is within 5%, indicating that the estimation accuracy does not decrease significantly. Therefore, the appropriate range for the number of sensors is between four and 128, and sufficient estimation accuracy can be maintained even if the number of sensors is reduced within this range. In Figure 5, the 5% increase rate of error compared to the minimum value is indicated by the horizontal dashed line. Also, in Figure 5, the dashed curve shows the general trend of the change in temperature estimation error based on the calculated values ​​(plotted points).

[0029] The reason why the mean squared error increases when the number of sensors is 256 or more is thought to be because an increase in the number of input variables to the machine learning model causes overfitting, resulting in a decrease in prediction accuracy for unknown data.In addition, the reason why the mean squared error increases when the number of sensors is 2 or less is thought to be because the information obtained from the sensors cannot adequately represent the internal state.

[0030] Figure 6 is a diagram illustrating appropriate sensor installation positions depending on the number of sensors. As the number of sensors decreases, appropriate sensor installation positions tend to be concentrated on the front and bottom surfaces of the hot-dip galvanizing pot wall. When the number of sensors is reduced to four, the only appropriate sensor installation positions are the first and second areas shown in Figure 4.

[0031] The reason why the first and second regions have such a large influence as temperature measurement locations is presumed to be as follows: The first region corresponds to the region through which the flow passes immediately after detaching from the steel plate. Because the fluid whose temperature has changed due to heat input or output from the steel plate passes near the temperature sensor installed in the first region, it is believed that temperature fluctuations caused by the steel plate temperature can be accurately captured. The second region corresponds to the region where the circulating flow and the fluid flowing in from the inductor merge. Because the fluid whose temperature has risen due to heat input from the inductor passes near the temperature sensor installed in the second region, it is believed that temperature fluctuations caused by the inductor can be accurately captured. Below, examples of the sensor position setting method, training data creation method, and machine learning model generation method used in this embodiment are described. In this embodiment, each method is described as being executed by the temperature estimation system, but this is merely an example. An external computer (not shown in FIG. 3) may execute each method in advance before the temperature is measured by the temperature estimation system.

[0032] (Sensor position setting method) FIG. 7 is a flowchart showing an example of a process of a sensor position setting method, which is a method for appropriately setting a sensor position.

[0033] The temperature estimation system acquires a machine learning model (step S1). The machine learning model is generated or updated in advance using a machine learning model generation method described below. The temperature estimation system verifies the estimation accuracy of the machine learning model (step S2). If the required estimation accuracy is met (Yes in step S3), the temperature estimation system evaluates the estimation contribution of each input (i.e., data for each thermometer position) (step S4). The estimation contribution may be evaluated using a known method for measuring the usefulness of input variables in a machine learning model, such as permutation importance. In this embodiment, the estimation contribution of each input is calculated as a scalar value, and a larger value indicates a greater contribution to the internal state estimation at the target thermometer position. The temperature estimation system sorts the inputs in order of their estimation contribution, excludes the bottom half (i.e., the bottom 50% of the total) (step S5), and updates the model by learning using the top half (i.e., the top 50% of the total) of inputs (step S6). The model is updated using the same method as the machine learning model generation method described below. After updating, the process returns to step S1. If the temperature estimation system does not satisfy the required estimation accuracy (No in step S3), it acquires the machine learning model before the update (step S7) and ends the process. By repeatedly performing the above process as long as the machine learning model satisfies the required accuracy, the temperature estimation system can finally obtain a machine learning model that satisfies the required accuracy and has reduced thermometer positions. In the example of FIG. 7, the number of thermometer positions is halved (steps S5 and S6), but the number to be excluded (the number used in updating the model) is not limited to 50% of the total and can be adjusted.

[0034] (How to create training data) In training a machine learning model, input data includes multiple pairs of thermometer positions and the temperatures at those positions, and output data includes the temperature distribution of the molten zinc. However, since it is difficult to observe the temperature distribution of the molten zinc in an actual machine, data on the wall temperature and the temperature distribution of the molten zinc must be obtained by alternative means such as model experiments. In this embodiment, a method is adopted in which the input data and output data are calculated using fluid analysis.

[0035] FIG. 8 is a flowchart of a method for creating training data. First, the internal structure of the hot-dip galvanizing pot is modeled on a computer using a known method. Then, the temperature estimation system provides the steel sheet speed, steel sheet temperature, and inductor heat input as conditions for fluid analysis (step S101). Then, a thermal fluid analysis of the molten zinc bath is performed by fluid simulation based on the input conditions (step S102), and the temperature distribution on the wall surface and inside of the hot-dip galvanizing pot according to the operation is calculated (step S103). FIG. 9 is a diagram illustrating the temperature distribution of molten zinc obtained by fluid simulation.

[0036] The temperature estimation system then sets sampling points at equal intervals on the wall surface and the estimation area (or estimation surface) (step S104) and acquires the temperature distribution at the multiple sampling points (point cloud) (step S105). The temperature estimation system then calculates the vectors of the temperature distribution on the wall surface and the estimation area (step S106). In other words, a vector representation of the temperature distribution of the molten zinc relative to the wall surface temperature is obtained. The temperature estimation system can create a large amount of training data by setting multiple values ​​for the fluid analysis conditions of steel plate speed, steel plate temperature, and inductor heat input and executing a series of processes.

[0037] (How to generate a machine learning model) As shown in Figure 10, the machine learning model has the function of inputting temperatures measured by multiple sensors installed on the exterior wall in vector format and outputting the temperature distribution of the molten zinc to be estimated in vector format. The internal structure of the machine learning model is not particularly limited, but since the correlation between the input and output is nonlinear, a nonlinear activation function is inserted between layers. In this embodiment, a machine learning model consisting only of a fully connected layer is constructed, and a ReLU function, which is a nonlinear function, is used as the activation function. The machine learning model is generated by machine learning using training data created by the above-mentioned fluid analysis.

[0038] 11 is a flowchart of a method for generating a machine learning model. The temperature estimation system acquires temperature distributions (vectors) on the wall surface and in the estimation area (step S201). The temperature estimation system learns a machine learning model that estimates an estimation surface temperature vector (output) from the wall surface temperature vector (input) (step S202), and generates the machine learning model (step S203).

[0039] The effects of the present disclosure will be specifically described below based on examples, but the present disclosure is not limited to these examples.

[0040] As an example of the present disclosure, an example of the results of estimating the internal state of a hot-dip galvanizing apparatus will be described. In this example, the estimation target was a central cross section of a hot-dip galvanizing pot, and the temperature distribution on this plane was estimated using the method of the above embodiment. The number of temperature sensors used for the estimation was four, which is the lower limit of the appropriate range of the number of sensors mentioned above.

[0041] FIG. 12 shows an example of the temperature distribution in the central cross section of the hot-dip galvanizing pot in this example. Here, the true values ​​(numerical analysis results) of the temperature distribution under the same operating conditions are shown in FIG. 9. Comparing FIG. 9 and FIG. 12, there is a consistent tendency for the fluid temperature to be high near the steel plate (f1 and f2 in FIG. 2) and low near the bottom (f4 in FIG. 2). It was confirmed that the method disclosed herein can estimate the internal state of the hot-dip galvanizing pot with high accuracy. In FIG. 9 and FIG. 12, isotherms are used to indicate how much higher a temperature is from a predetermined reference temperature (unit: K). The numbers in FIG. 9 and FIG. 12 indicate the temperature of each isotherm in K.

[0042] As described above, in the temperature estimation system, hot-dip galvanizing equipment, temperature estimation method, and hot-dip galvanizing method according to the present embodiment, the above configuration makes it possible to acquire the temperature distribution of the entire molten zinc, including areas distant from the position of the thermometer installed in the hot-dip galvanizing pot.

[0043] Although the embodiments of the present disclosure have been described based on the drawings and examples, it should be noted that those skilled in the art would easily be able to make various modifications or alterations based on the present disclosure. Therefore, it should be noted that these modifications and alterations are included within the scope of the present disclosure. For example, the functions included in each component or step can be rearranged so as not to cause logical inconsistencies, and multiple components or steps can be combined or divided into one. The embodiments of the present disclosure can also be realized as a program executed by a processor included in an apparatus or a storage medium on which a program is recorded. It should be understood that these are also included within the scope of the present disclosure.

Claims

1. A temperature estimation system for estimating the temperature distribution of molten zinc inside a hot-dip galvanizing pot, a plurality of thermometers installed on a wall surface of the hot-dip galvanizing pot to measure the temperature of the molten zinc at the installed positions; An estimation unit that acquires temperatures measured from the plurality of thermometers and positions of the thermometers, and estimates the temperature distribution of the molten zinc based on a machine learning model previously generated by machine learning and the acquired temperatures and positions; an output unit that outputs the estimated temperature distribution of the molten zinc, The machine learning model is generated using teacher data in which a plurality of sets of positions where the plurality of thermometers are installed and the temperatures of the molten zinc at the positions are input values, and a temperature distribution of the molten zinc corresponding to the input values ​​is output values, the hot-dip galvanizing pot includes therein a sink roll and a support roll for transporting the steel sheet, and an inductor for heating the molten zinc located on the bottom side of the sink roll; The plurality of thermometers are installed in a plurality of regions including at least a first region and a second region along a flow path of the molten zinc generated by the conveyance of the steel plate, the first region is located upstream of the second region in the flow path, The second region is included in the bottom surface downstream from the inductor.

2. The temperature estimation system of claim 1, wherein the wall surface connected to the bottom surface is defined as the front surface, where the flow path passing between the sink roll and the inductor collides, and the wall surfaces located on both sides of the flow path passing between the sink roll and the inductor are defined as side surfaces, and the first region is included in the front surface or the bottom surface upstream of the inductor.

3. The temperature estimation system of claim 1, wherein the wall surface connected to the bottom surface is defined as the front surface, where the flow path passing between the sink roll and the inductor collides, and the wall surfaces located on both sides of the flow path passing between the sink roll and the inductor are defined as sides, and the first region is included in the sides.

4. A hot-dip galvanizing facility comprising the hot-dip galvanizing pot, the temperature distribution of the molten zinc of which is estimated by the temperature estimation system according to any one of claims 1 to 3.

5. A temperature estimation method executed by a temperature estimation system that estimates a temperature distribution of molten zinc inside a hot-dip galvanizing pot, comprising: measuring the temperature of the molten zinc at positions where the thermometers are installed on the wall surface of the hot-dip galvanizing pot; A step of acquiring temperatures measured from the plurality of thermometers and positions of the thermometers, and estimating a temperature distribution of the molten zinc based on a machine learning model previously generated by machine learning and the acquired temperatures and positions; and outputting the estimated temperature distribution of the molten zinc, The machine learning model is generated using teacher data in which a plurality of sets of positions where the plurality of thermometers are installed and the temperatures of the molten zinc at the positions are input values, and a temperature distribution of the molten zinc corresponding to the input values ​​is output values, the hot-dip galvanizing pot includes therein a sink roll and a support roll for transporting the steel sheet, and an inductor for heating the molten zinc located on the bottom side of the sink roll; The plurality of thermometers are installed in a plurality of regions including at least a first region and a second region along a flow path of the molten zinc generated by the conveyance of the steel plate, the first region is located upstream of the second region in the flow path, The temperature estimation method, wherein the second region is included in the bottom surface downstream from the inductor.

6. A hot-dip galvanizing method, comprising: plating a surface of the steel sheet using the hot-dip galvanizing pot in which a temperature distribution of the molten zinc is estimated by the temperature estimation method according to claim 5.

Images