Slag defect prediction method, slag defect reduction method, method for manufacturing hot-dip galvanized steel sheet, method for manufacturing galvannealed steel sheet, method for generating slag defect prediction model, slag defect prediction device, and slag defect prediction terminal system

Abstract

The present application provides a method for more effectively reducing dross defects. The dross defect prediction method includes: a step of receiving an input of an operating condition with respect to a dross defect prediction model, the operating condition including a first operating condition related to a galvanizing bath and a second operating condition related to a furnace snout, the dross defect prediction model being a model of machine learning based on operating data including a first parameter related to the galvanizing bath and a second parameter related to the furnace snout as input variables and dross defect information of a steel strip as an output variable; and a step of calculating a predicted value of the dross defect information of the steel strip based on the input operating condition and through the dross defect prediction model.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to a dross defect prediction method, a dross defect reduction method, a method of manufacturing a hot-dip galvanized steel sheet, a method of manufacturing an alloyed hot-dip galvanized steel sheet, a method of generating a dross defect prediction model, a dross defect prediction device, and a dross defect prediction terminal system. BACKGROUND

[0002] A hot-dip galvanized steel sheet, which is one type of hot-dip metal plated steel sheet, is widely used in the fields of building materials, automobiles, home electric appliances, and the like. Moreover, in these uses, the hot-dip galvanized steel sheet is required to be excellent in appearance. In particular, the appearance after painting is strongly affected by surface defects such as plating thickness unevenness, flaws, and foreign matter attachment, and thus it is important to reduce the surface defects of the hot-dip galvanized steel sheet.

[0003] As the surface defects of the hot-dip galvanized steel sheet, defects (hereinafter, also referred to as dross defects) originating from dross attached to the surface of the steel sheet in the zinc plating bath are recognized as one of the surface defects that should be prevented. The dross attached to the surface of the steel sheet causes a pressure wound at the time of secondary processing such as press forming, for example. In addition, if the dross is attached to the surface of the steel sheet, the dross is pressed into the steel sheet by a rolling roll at the time of temper rolling, and a wrinkle-like pattern is formed on the surface of the steel sheet.

[0004] Existing technologies for reducing the generation of such dross defects are proposed. In Patent Literature 1, a method is disclosed in which Fe dissolved in the plating solution due to a temperature variation of the plating bath is precipitated, the dissolved Fe is combined with Al or Zn to generate dross, and thus the output of an inductor is controlled in order to control the temperature of the plating bath to reduce the temperature difference in the plating solution. In Patent Literature 2, a method is disclosed in which the accumulation height of bottom dross is inferred based on the results obtained by measuring the bath temperature at a plurality of positions different in the depth direction within the plating bath. In Patent Literature 3, a method is disclosed in which the wall surface of a snout device connected by being immersed in the plating bath on the feeding side of the zinc plating tank is heated and kept warm.

[0005] Patent Literature 1: Japanese Patent Application Laid-Open No. 2001-107208

[0006] Patent Literature 2: Japanese Patent No. 6137211

[0007] Patent Literature 3: Japanese Patent Application Laid-Open No. 2002-275606

[0008] Patent Literature 1 is based on the finding that the temperature difference in the plating bath is correlated with the accumulation amount of bottom dross. According to the technology described in Patent Literature 1, it is possible to reduce dross defects. However, the accumulation amount of bottom dross is also affected by parameters other than the temperature difference in the plating bath, and thus there is room for improvement in the reduction of dross defects.

[0009] In Patent Literature 2, the amount of accumulation of the bottom dross is predicted only from the temperature information in the plating bath. According to the technology described in Patent Literature 2, it is possible to reduce the dross defects. However, the amount of accumulation of the bottom dross is also affected by parameters other than the temperature information in the plating bath, and thus there is room for improvement in the reduction of the dross defects.

[0010] Patent Literature 3 is a method that focuses on the correlation between the temperature of the wall surface of the furnace nose device and the dross present in the furnace nose, but as described above, the dross includes not only the top dross that floats on the surface of the bath but also the bottom dross that settles at the bottom of the plating solution and accumulates. Therefore, there is a problem that it is not possible to prevent the dross defects by only reducing the dross that floats on the surface of the bath in the furnace nose. SUMMARY

[0011] The present disclosure was completed in view of the above-described situation, and aims to provide a method that more effectively reduces dross defects.

[0012] In order to achieve the above-described object, the present inventors have repeatedly conducted intensive research and found the following matters, and completed the present disclosure, that is, to consider a plurality of operating factors that affect the generation behavior of the dross in the galvanizing bath to accurately predict the generation of the dross defects and to be able to perform appropriate and rapid re-setting of the operating conditions based on the prediction result, thereby being able to more effectively reduce the dross defects.

[0013] The present disclosure was completed based on the above-described findings. That is, the gist structure of the present disclosure is as follows.

[0014] [1] A dross defect prediction method of a steel strip detected on the downstream side of a galvanizing bath in a continuous hot dip galvanizing device including an annealing furnace, the galvanizing bath, and a furnace nose, the galvanizing bath being formed with a galvanizing bath, the furnace nose being provided on the delivery side of the annealing furnace and configured such that the front end thereof is immersed in the galvanizing bath, wherein

[0015] The dross defect prediction method includes:

[0016] receiving an input of an operating condition with respect to a dross defect prediction model, the operating condition including a first operating condition related to the galvanizing bath and a second operating condition related to the furnace nose, the dross defect prediction model being a model of machine learning based on inputting operating data including a first parameter related to the galvanizing bath and a second parameter related to the furnace nose as an input variable and outputting dross defect information of the steel strip as an output variable; and

[0017] calculating a predicted value of the dross defect information of the steel strip based on the input operating condition and by the dross defect prediction model.

[0018] [2] The dross defect prediction method according to the above [1], wherein

[0019] The continuous hot-dip galvanizing apparatus has a gas blowing apparatus on the delivery side of the galvanizing bath,

[0020] The dross defect prediction model is a model of the machine learning performed based on a third parameter related to the gas blowing apparatus as the input variable, and the operating conditions further include a third operating condition related to the gas blowing apparatus.

[0021] [3] The dross defect prediction method for a steel strip according to the above [1] or [2], wherein

[0022] The first parameter includes one or two parameters selected from the temperature of the galvanizing bath and the Al concentration in the galvanizing bath.

[0023] [4] The dross defect prediction method according to any one of the above [1] to [3], wherein

[0024] The second parameter includes one or more parameters selected from the dew point in the furnace nose, the hydrogen concentration in the furnace nose, the oxygen concentration in the furnace nose, the temperature of the steel strip in the furnace nose, and the atmosphere temperature in the furnace nose.

[0025] [5] The dross defect prediction method according to any one of the above [1] to [4], wherein

[0026] The first operating condition includes one or more selected from the bath temperature control output of the galvanizing bath, the strip speed of the steel strip, and the press-in amount of the support roll in the galvanizing bath with respect to the steel strip.

[0027] [6] The dross defect prediction method according to any one of the above [1] to [5], wherein

[0028] The second operating condition includes the flow ratio of the mixed gas supplied to the furnace nose.

[0029] [7] A dross defect reduction method for a steel strip, wherein

[0030] The dross defect reduction method for a steel strip includes:

[0031] calculating a predicted value of the dross defect information of the steel strip on the downstream side of the galvanizing bath using the dross defect prediction method according to any one of the above [1] to [6]; and

[0032] re-setting the operating conditions based on the predicted value of the dross defect information.

[0033] [8] A method for manufacturing a hot-dip galvanized steel sheet, wherein

[0034] The continuous hot-dip galvanizing apparatus whose operating conditions are controlled using the dross defect reduction method described in the above [7] is used to form a zinc plating layer on the surface of the steel strip to form a hot-dip galvanized steel sheet.

[0035] [9] A method for manufacturing an alloyed hot-dip galvanized steel sheet, wherein the continuous hot-dip galvanizing apparatus whose operating conditions are controlled using the dross defect reduction method described in the above [7] is used,

[0036] The continuous hot-dip galvanizing apparatus further includes a reheating apparatus at a position downstream of the zinc plating bath and upstream of the defect detection device,

[0037] In the zinc plating bath, a zinc plating layer is formed on the surface of the steel strip to form a hot-dip galvanized steel sheet,

[0038] The hot-dip galvanized steel sheet is further subjected to an alloying treatment by the reheating apparatus to form an alloyed hot-dip galvanized steel sheet.

[0039]

[10] A method for generating a dross defect prediction model for a steel strip detected downstream of a zinc plating bath in a continuous hot-dip galvanizing apparatus including an annealing furnace, the zinc plating bath, and a furnace nose, the annealing furnace annealing a steel strip, the zinc plating bath having a zinc plating bath, and the furnace nose being provided at a delivery side of the annealing furnace and being configured to have an end immersed in the zinc plating bath, wherein

[0040] The method for generating the dross defect prediction model includes:

[0041] a step of acquiring operating data including a first parameter related to the zinc plating bath and a second parameter related to the furnace nose; and

[0042] a step of performing machine learning using the acquired operating data as input variables and using dross defect information of a steel strip detected downstream of the zinc plating bath as output variables.

[0043]

[11] The method for generating a dross defect prediction model according to the above

[10] , wherein

[0044] In the step of performing machine learning, one or more machine learning algorithms selected from a neural network, decision tree learning, random forest, support vector regression, Gaussian process, and k-nearest neighbor method are used.

[0045]

[12] A dross defect prediction device that predicts a dross defect of a steel strip detected on a downstream side of a galvanizing bath of a continuous hot dip galvanizing apparatus including an annealing furnace, the galvanizing bath in which a galvanizing bath is formed, and a furnace snout provided on a delivery side of the annealing furnace and configured such that a front end portion is immersed in the galvanizing bath, wherein

[0046] The dross defect prediction device includes:

[0047] an acquisition unit that acquires an operation condition including a first operation condition related to the galvanizing bath and a second operation condition related to the furnace snout; and

[0048] a control unit that inputs the operation condition to a dross defect prediction model to calculate a predicted value of dross defect information of the steel strip, the dross defect prediction model being a model of machine learning performed based on operation data including a first parameter related to the galvanizing bath and a second parameter related to the furnace snout as input variables and the dross defect information of the steel strip as an output variable.

[0049]

[13] A dross defect prediction terminal system, wherein:

[0050] the dross defect prediction device according to

[12] ; and

[0051] a terminal device that receives a user input related to a change in the operation condition and transmits user input information based on the user input to the dross defect prediction device,

[0052] the dross defect prediction device includes a control unit that changes at least a part of the operation condition based on the user input information as a changed operation condition, and calculates a predicted value of the dross defect information of the steel strip based on the changed operation condition and through the dross defect prediction model.

[0053] According to the present disclosure, it is possible to more effectively reduce dross defects. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 is a detailed view of a hot dip galvanizing bath peripheral apparatus of a continuous hot dip metal plating apparatus.

[0055] Figure 2 (a) of FIG. 1 is a schematic view of a dross defect, Figure 2 (b) of FIG. 1 is an optical microscope-based observation image of the dross defect, Figure 2 (c) of FIG. 1 is an image after image processing of the dross defect, Figure 2 (d) of FIG. 1 is an optical microscope-based observation image of the dross defect X in the A-A' cross section.

[0056] Figure 3 Fig. 1 is a diagram showing an outline of a continuous hot dip metal plating apparatus.

[0057] Figure 4 Fig. 2 is a diagram for explaining a gas supply method to a soaking zone.

[0058] Figure 5 Fig. 3 is a diagram showing a gas supply method to a furnace nose.

[0059] Figure 6 Fig. 4 (a) is a plan view showing an outline of a galvanizing bath, Figure 6 Fig. 4 (b) is a front view showing an outline of the galvanizing bath.

[0060] Figure 7 Fig. 5 is a diagram showing an outline of a gas purging apparatus.

[0061] Figure 8 Fig. 6 is a diagram showing an outline of a defect detection device.

[0062] Figure 9 Fig. 7 is a coordinate graph showing a relationship between a dew point and a number of drosses per unit length in a furnace nose.

[0063] Figure 10 Fig. 8 is a functional block diagram showing a structure of a dross defect prediction model generation device.

[0064] Figure 11 Fig. 9 is a diagram showing an outline of a dross defect reduction method.

[0065] Figure 12 Fig. 10 is a functional block diagram showing a structure of a dross defect prediction device.

[0066] Figure 13 Fig. 11 is a flowchart showing one example of an operation of a dross defect prediction model generation device.

[0067] Figure 14 Fig. 12 is a schematic diagram showing a structure of a dross defect prediction model.

[0068] Figure 15 Fig. 13 is a flowchart showing one example of an operation of a dross defect prediction device.

[0069] Figure 16 Fig. 14 is a functional block diagram showing a structure of a dross defect prediction system. DETAILED DESCRIPTION

[0070] In one example, a hot-dip galvanized steel sheet is manufactured using a continuous hot-dip galvanizing apparatus configured to continuously perform a series of processes including heating, cooling, galvannealing, and alloying treatment of hot-dip galvanizing. A steel sheet supplied from a supply side of the continuous hot-dip galvanizing apparatus is annealed while passing through a continuous annealing furnace. A steel strip guided from a discharge side of the continuous annealing furnace to a galvanizing bath through a furnace nose of the continuous annealing furnace is introduced into a molten zinc bath formed in the galvanizing bath, thereby forming a hot-dip galvanized layer on the surface. Further, the steel strip on which the hot-dip galvanized layer is formed is pulled up from the molten zinc bath, and a purge gas is blown from gas blowing nozzles arranged on both sides of the steel strip, thereby scraping off excess molten zinc adhering to the surface of the steel strip, and a hot-dip galvanized steel sheet having an adhering amount (hereinafter also referred to as a unit area weight) of the hot-dip galvanized layer adjusted is obtained.

[0071] In the production of such a hot-dip galvanized steel sheet, dross generated in the galvanizing bath adheres to the surface of the steel strip, thereby generating a dross defect. In addition, in the present specification, "dross" refers to an intermetallic compound generated by a reaction of Fe dissolved from the steel strip with bath components (Al, Zn) in the galvanizing bath. The dross is roughly classified into Fe-Al dross and Fe-Zn dross depending on the kind of the bath component that reacts with Fe. The Fe-Al dross has a smaller density than the density of molten zinc, and is called top dross. As shown in FIG. 1A, the top dross 5 floats on the surface of the plating solution of the galvanizing bath. In contrast, the Fe-Zn dross has a larger density than the density of molten zinc, and is called bottom dross. As shown in FIG. 1B, the bottom dross 6 settles at the bottom of the plating solution of the galvanizing bath. Figure 1 Figure 1 As shown in FIG. 1B, the bottom dross 6 settles at the bottom of the plating solution of the galvanizing bath. The bottom dross settled at the bottom of the plating solution is floated by flow in the plating solution, and thus, when the steel strip is pulled up from the surface of the bath, any dross adheres to the surface of the steel strip in the bath or on the surface of the bath. At this time, a dross defect is generated from dross of a size of a certain amount or more among the dross adhering to the surface of the steel strip. Thus, in the present embodiment, dross of a size of a certain amount or more adhering to the surface of the steel sheet is used as an index when detecting a dross defect.

[0072] Figure 2 FIG. 1 is a diagram showing a schematic view and an observed image of a dross defect. Figure 2 (a) of FIG. 1 is a schematic view of a dross defect X on the surface of a hot-dip galvanized steel sheet G. Figure 2 (b) of FIG. 1 is an observed image of the dross defect X based on an optical microscope. Figure 2 (c) of FIG. 1 is an image obtained by performing image processing on the observed image of (b). Figure 2 (d) of FIG. 1 is an image obtained by performing image processing on the observed image of (c). Figure 2 (e) of FIG. 1 is an image obtained by performing image processing on the observed image of (d). Figure 2 The observed image of the dross defect X in the A-A' cross section of (a) is based on an optical microscope. As shown in (a), the dross defect X is observed from the surface of the hot-dip galvanized steel sheet G. Figure 2 ​The dross defect X generated on the surface of the hot-dip galvanized steel sheet G is recognized as a dot-shaped defect of about 100 μm in size as shown in (a). Figure 2 An optical microscope image of the portion enclosed by a broken line in (a) is as shown in Figure 2 Figure 2 An image in which the bright and dark portions are emphasized by image processing of (b) is Figure 2 (c) of FIG. 10. By performing image processing on an image taken from the surface of such a hot-dip galvanized steel sheet, it is possible to recognize the dross defect X as a dot-shaped defect. In the present embodiment, a dross defect detection device based on such a principle is used to perform detection of the dross defect X. In addition, Figure 2 (d) of FIG. 10 is an observation image obtained by cutting the hot-dip galvanized steel sheet G in the A-A' cross section and observing the cross section with an optical microscope. It is understood that the dross adhering to the surface of the steel sheet in the galvanizing bath is pressed into the steel sheet by the extrusion of the rolling rolls during the flattening rolling, thereby generating the dross defect X.

[0073] In the present disclosure, in order to manufacture a hot-dip galvanized steel sheet having reduced dross defects as described above, a dross defect prediction method is provided. Hereinafter, one embodiment of the present disclosure will be mainly described with reference to the drawings. The embodiment shown below exemplifies a device and a method for embodying the technical idea of the present disclosure, but the present disclosure is not limited to the embodiment below. In addition, in the present specification, a numerical range indicated using "~" means a range including the numerical values recited before and after "~" as lower limit values and upper limit values.

[0074] <<Dross Defect Prediction System>>

[0075] In Figure 16 , the structure of the dross defect prediction system according to the present embodiment is shown. The dross defect prediction system shown in Figure 12 performs the dross defect prediction method according to the present embodiment. As shown in Figure 16 , the dross defect prediction system includes a dross defect prediction device 84, a dross defect prediction model generation device 83, and a running performance database (DB) 85. The dross defect prediction system predicts the dross defect of a steel strip detected on the downstream side of the galvanizing bath in a continuous hot-dip galvanizing apparatus 100. The prediction of the dross defect can be a prediction of both the front and back surfaces of the steel strip, or a prediction of either the front surface or the back surface.

[0076] <<Continuous Hot-Dip Galvanizing Apparatus>>

[0077] First, with reference to Figure 3 ​The structure of the continuous hot dip galvanizing apparatus 100 according to one embodiment of the present disclosure will be described. In one example, the continuous hot dip galvanizing apparatus 100 has a longitudinal continuous annealing furnace 30 in which the heating zone 20, the soaking zone 21, and the cooling zones 22, 23 are arranged in this order side by side, and a zinc plating bath 1 as a hot dip galvanizing apparatus located downstream of the steel sheet passing direction of the cooling zone 23. In the present embodiment, the cooling zones 22, 23 include a first cooling zone 22 (fast cooling zone) and a second cooling zone 23 (slow cooling zone). The front end of the furnace nose 2 connected to the delivery side of the second cooling zone 23 is immersed in the zinc plating bath 1, and the continuous annealing furnace 30 and the zinc plating bath 1 are connected via the furnace nose 2. The gas in the continuous annealing furnace 30 flows from the downstream to the upstream of the furnace, and is discharged from the steel sheet guide inlet at the lower portion of the heating zone 20. The steel strip S supplied from the entry side of the continuous hot dip galvanizing apparatus is annealed in the continuous annealing furnace 30 in the order of the heating zone 20, the soaking zone 21, the cooling zones 22, 23, and the furnace nose 2. Then, the steel strip S guided from the delivery side of the continuous annealing furnace via the furnace nose to the zinc plating bath is immersed in the hot dip galvanizing bath formed in the zinc plating bath 1, whereby a hot dip galvanizing layer is formed on the surface of the steel strip S to become a hot dip galvanizing steel sheet G.

[0078] At a position of the continuous hot dip galvanizing apparatus 100 located downstream of the zinc plating bath 1, a reheating apparatus, a temper rolling apparatus, and a chemical conversion coating device, not shown, are provided as needed. The hot dip galvanizing steel sheet pulled up from the zinc plating bath is passed through the gas purging apparatus 70, and then passed through the reheating apparatus, the temper rolling apparatus, and the chemical conversion coating device as needed. Then, the surface defects of the hot dip galvanizing steel sheet G are inspected by a defect detection device 80 provided downstream. The defect detection device 80 is an apparatus for inspecting defects on the front and back surfaces of the steel sheet, and can identify various surface defects. As for the detection of the dross defect X among the various surface defects, for example, a method of combining image capturing and image processing of the surface of the steel sheet can be employed. By the defect detection device 80, the dross having a size of a certain size or more attached to the surface of the steel sheet is defined as the dross defect X. Figure 2 As described above, a method of identifying by combining image capturing and image processing of the surface of the steel sheet can be employed. By the defect detection device 80, the dross having a size of a certain size or more attached to the surface of the steel sheet is defined as the dross defect X.

[0079] In the heating zone 20, the steel strip S can be indirectly heated using a radiant tube (RT) or a heating zone heating device such as an electric heater. A reducing gas or a non-oxidizing gas can be additionally supplied to the heating zone 20 at the same time as the gas from the soaking zone 21, the first cooling zone 22, the second cooling zone 23, and the furnace nose 2 flows into the heating zone 20. As the reducing gas, an H2-N2mixed gas is generally used. As such an H2-N2mixed gas, for example, a gas having a composition of 1 to 20 vol% of H2and the remainder consisting of N2and unavoidable impurities (dew point: about -60°C) can be cited. In addition, as the non-oxidizing gas, a gas having a composition consisting of N2and unavoidable impurities (dew point: about -60°C) is used. The method of gas supply to the heating zone 20 is not particularly limited, but the gas is preferably supplied from two or more points in the height direction and one or more points in the length direction of the inlet so as to uniformly introduce the gas into the heating zone 20.

[0080] In the soaking zone 21, the steel strip S can be indirectly heated using a radiant tube (not shown) as a heating unit. The average temperature inside the soaking zone 21 is preferably 700 to 900°C. A reducing gas or a non-oxidizing gas is supplied to the soaking zone 21. As the reducing gas, an H2-N2mixed gas is generally used, for example, a gas having a composition of 1 to 20 vol% of H2and the remainder consisting of N2and unavoidable impurities (dew point: about -60°C) can be cited. In addition, as the non-oxidizing gas, a gas having a composition consisting of N2and unavoidable impurities (dew point: about -60°C) can be cited.

[0081] Figure 4 is a schematic view of a gas supply system of a humidified gas or a dry gas to the soaking zone 21. In Figure 4 , the above gas is supplied to the inside of the soaking zone 21 through the gas piping 40 via the gas supply ports 41A, 41B, 41C provided at the upper portion of the soaking zone 21 and the gas supply ports 42A, 42B, 42C provided at the lower portion of the soaking zone 21. By using a humidified gas or a dry gas as the gas supplied by the gas supply system, it is possible to control the dew point inside the soaking zone 21. The dew point in the furnace affects the generation of oxides on the surface of the steel strip S, thereby affecting the dross generation in the galvanizing bath 1, and thus becomes an object to be controlled as an operating parameter of the continuous annealing furnace 30.

[0082] The cooling zones 22, 23 are provided with a cooling device to cool the steel strip S during the passage of the steel strip S in the cooling zones 22, 23. As with the soaking zone, the above gas can also be supplied to the cooling zones 22, 23. The gas is preferably supplied from two or more points in the height direction and two or more points in the length direction of the inlet of the cooling zones 22, 23 so as to uniformly introduce the gas into the cooling zones 22, 23.

[0083] <Galvanizing bath>

[0084] The peripheral equipment of the galvanizing bath 1 in the present embodiment is shown in detail in Figure 1 The furnace nose 2 is connected to the galvanizing bath 1, and the galvanizing bath 1 is provided with a sink roll 3, a support roll 4, an ingot feeding device 7, and a heating device 11. The galvanizing bath 1 is arbitrary, and can be provided with at least one of a bath thermometer 9 and a bath analysis device 10.

[0085] The furnace nose 2 divides a space through which the steel strip S passes, in order to block the steel strip S from the atmosphere and enable the steel strip S to pass in a reducing atmosphere during a period from after the steel strip S comes out of the continuous annealing furnace 30 to before the steel strip S enters the hot-dip galvanizing bath. The furnace nose 2 is a member whose cross section perpendicular to the advancing direction of the steel strip S is rectangular, and as described above, its upper end is connected to the outlet side of the continuous annealing furnace, and its lower end (front end portion) is immersed in the hot-dip galvanizing bath accumulated in the galvanizing bath 1.

[0086] In the present embodiment, the steel strip S annealed in the continuous annealing furnace 30 in a reducing atmosphere is continuously introduced into the galvanizing bath formed in the galvanizing bath 1 by passing through the furnace nose 2. Then, the steel strip S (hot-dip galvanizing steel sheet G) is pulled up toward the upper side of the galvanizing bath via the sink roll 3 and the support roll 4.

[0087] The ingot feeding device 7 is a device for feeding an ingot 8 in which the composition has been adjusted to the galvanizing bath 1. The ingot feeding device 7 replenishes the composition (Zn, Al, etc.) consumed by stripping from the galvanizing bath by plating to the steel strip S, monitors the bath level (height from the bottom of the galvanizing bath 1 to the liquid surface) of the plating bath by a not-shown bath level gauge, and adjusts the amount of the ingot 8 to be fed in accordance with the value thereof. Thus, it is possible to keep the bath level of the galvanizing bath 1 constant. In addition, in order to suppress a drastic change in the composition of the galvanizing bath in the vicinity of the steel strip S, as shown in Figure 1 , the ingot feeding device 7 is generally provided on the back side (back side) of the furnace nose 2.

[0088] The bath thermometer 9 is a device capable of measuring the temperature at a specific position of the galvanizing bath 1. Since the bath temperature is about 460°C, as the bath thermometer 9, a K-type thermocouple is generally used.

[0089] The bath analysis device 10 is a device capable of measuring the composition (Al, etc.) in the galvanizing bath 1, and can monitor the bath composition at all times. The bath analysis device 10 can perform constant measurement of the bath composition, for example, by an optical measurement method using laser-induced breakdown spectroscopy. For the analysis of the bath composition, it is possible to measure at a sampling period of about 1 second, and it is possible to perform continuous analysis of the bath composition.

[0090] The setting position of the bath analysis device 10 in the galvanizing bath 1 is not particularly limited. For the bath analysis, it is important to measure the bath composition in the vicinity of the steel strip S, and therefore it is preferable to set the bath analysis device 10 at a position as close as possible to the steel strip S.

[0091] It is also possible to collect a sample from the galvanizing bath without providing the bath analysis device 10 in the galvanizing bath 1, and to analyze the sample by an algorithm such as inductivity coupled plasma (ICP) analysis, thereby performing the bath analysis.

[0092] In Figure 6 (a) of FIG. 10 is a schematic plan view of the galvanizing bath 1, Figure 6 (b) of FIG. 10 is a schematic front view of the galvanizing bath 1. It is inferred from numerical analysis that the molten zinc in the galvanizing bath flows as a broken line shown in (a) of FIG. 10. The molten zinc in the vicinity of the steel strip S moves to the side wall portion of the furnace nose 2 in about 1 minute, and therefore by setting the bath analysis device 10 at the side wall side of the furnace nose 2, it is possible to analyze the bath composition in the vicinity of the steel strip S practically without a large delay time. However, if the bath analysis device 10 is excessively close to the side wall of the galvanizing bath 1, the bath composition is measured with stagnation due to the influence of the side wall, and therefore it is preferable to set the bath analysis device 10 at a distance dl of 200 mm or more from the side wall. Further, in the example of (a) of FIG. 10, the bath analysis device 10 is set at a position separated by 250 mm from the side wall of the galvanizing bath 1. Figure 6 Figure 6

[0093] The setting position of the bath temperature gauge 9 in the galvanizing bath 1 is not particularly limited, but since the bath temperature and the bath composition are closely related to each other, it is preferable to set the bath temperature gauge 9 at a position close to the bath analysis device 10. Specifically, it is preferable to set the bath temperature gauge 9 at a position in the galvanizing bath 1 within a distance d2 of 500 mm from the bath analysis device 10 in the plan view of the galvanizing bath 1 of (a) of FIG. 11. Further, in the example of (a) of FIG. 11, the bath temperature gauge 9 is set at a position separated by 100 mm from the bath analysis device 10 in the galvanizing bath 1. Figure 6 Figure 6

[0094] The setting height of the bath analysis device 10 and the bath temperature gauge 9 in the bath is not particularly limited, but if excessively lower from the bath surface, the bath composition and the bath temperature are measured with stagnation, and therefore it is preferable to measure in a range where the front end portion is within a distance d3 of 1000 mm from the bath surface. Further, in the example of (b) of FIG. 12, the bath analysis device 10 and the bath temperature gauge 9 are both set so that the front end portions are at a position of 250 mm from the bath surface. Figure 6

[0095] ​​​​​The heating device 11 is a device capable of heating the temperature of the plating bath to a prescribed temperature. Generally, the bath temperature is adjusted to about 460°C by the heating device 11.

[0096] < Furnace Nozzle >

[0097] Figure 5 is a view for explaining a gas supply method to the furnace nozzle 2. Referring to Figure 5 (a), the gas supply portion and the dew point measuring portion to the furnace nozzle 2 are explained. Further, in Figure 5 (a), a view of the furnace nozzle 2 along which the steel strip S is passing is shown, which is cut along the width direction of the steel strip S and the axis of symmetry in the long direction. As shown in Figure 5 (a), the furnace nozzle 2 has a gas supply portion 50. The gas supply portion 50 includes: a first pipe 51A through which hydrogen gas is passed; a second pipe 51B through which nitrogen gas is passed; a third pipe 51C through which water vapor as an oxidizing gas is passed; a valve 52 for flow rate adjustment, which is installed to these pipes; a fourth pipe 51D through which mixed gas mixed from the gases supplied from these pipes is passed; and a fifth pipe 51E which is connected to the fourth pipe 51D and whose front end is located inside the furnace nozzle 2. The first pipe 51A and the third pipe 51C are connected to the second pipe 51B, and hydrogen, nitrogen, and water vapor can be mixed at an arbitrary flow rate ratio by adjusting the valve 52. A dew point measuring hole for measuring the dew point of the gas supplied to the furnace nozzle 2 can be provided to the fourth pipe 51D through which the mixed gas is passed, and the dew point of the gas can be measured by a not-shown dew point meter connected to the dew point measuring hole. Here, in the present embodiment, the flow rate ratio of the mixed gas supplied to the furnace nozzle 2 refers to the flow rate ratio of hydrogen, nitrogen, and water vapor obtained by adjusting the valve 52.

[0098] The dew point of the atmosphere inside the furnace nozzle 2 can be measured by a not-shown dew point meter connected to the dew point measuring hole 53B. The dew point of the atmosphere inside the furnace nozzle 2 can be changed to an arbitrary value by adjusting the gas supplied to the furnace nozzle 2. In addition, the dew point measuring hole 53B can also be connected to at least one of an oxygen concentration meter and a hydrogen concentration meter. If the dew point measuring hole 53B is also connected to at least one of the oxygen concentration meter and the hydrogen concentration meter, at least one of the oxygen concentration and the hydrogen concentration of the atmosphere inside the furnace nozzle 2 can be measured.

[0099] The temperature of the steel strip S within the nose 2 can be measured using a radiation thermometer 54. The radiation thermometer 54 is preferably located directly above the bath surface so that the temperature of the steel strip S when immersed in the galvanizing bath (immersion plate temperature) can be measured. However, if the dew point of the atmosphere within the nose 2 is low and zinc vapor is generated, the radiation thermometer 54 may be contaminated by the zinc vapor, making it impossible to install the radiation thermometer 54 below the nose 2. In such cases, the temperature of the steel strip S when immersed in the zinc bath can be continuously calculated using the atmospheric temperature within the nose measured by a thermocouple 57 installed within the nose 2 (described later) through numerical analysis that takes into account radiative heat conduction and convective heat transfer within the nose 2, thereby estimating the immersion plate temperature.

[0100] like Figure 5 As shown in (a), the oxidizing gas is preferably supplied into the furnace nose 2 from both ends of the furnace nose 2 in the width direction of the steel strip S. The reason why the 5th pipe 51E having a gas inlet at the front end is arranged on the side of the furnace nose 2 is that: since there is a tendency for the temperature near the side inside the furnace nose 2 to become lower, a downward flow is usually generated near the side, so that the oxidizing gas efficiently reaches the vicinity of the bath surface. The height of the gas inlet from the bath surface can be more than 100 mm and less than 3000 mm. By making the height of the gas inlet from the bath surface more than 100 mm, the gas can be prevented from directly reaching the bath surface, and the oxidizing gas can be prevented from being enriched near the bath surface. In addition, by making the height of the gas inlet from the bath surface less than 3000 mm, the gas concentration reaching the bath surface can be prevented from decreasing, thereby saving gas. Figure 5 In the example of FIG. 5 , the height h1 of the gas inlet is 500 mm above the bath surface.

[0101] The dew point of the atmosphere in the furnace nose 2 affects the oxidation of the bath surface (described later), so the dew point meter is preferably located directly above the bath surface. However, if the dew point in the furnace nose 2 is low and zinc vapor is generated, there is a concern that the dew point meter will be contaminated by the zinc vapor, making it impossible to place the dew point meter directly above the bath surface. In such cases, flow analysis can be used, for example, to estimate the dew point of the atmosphere directly above the bath surface in the furnace nose 2 based on the dew point of the atmosphere at a certain height above the bath surface. Specifically, flow analysis is used to determine in advance the relationship between the dew point directly above the bath surface and the dew point at a position x (mm) above the bath surface. For example, if the output of a dew point meter located 500 mm above the bath surface is -35°C, and if the dew point at the dew point meter's location (x = 500 mm) is 5°C higher than the bath surface based on the above relationship, the dew point directly above the bath surface can be estimated to be -30°C, and this value can be used as the dew point of the atmosphere in the furnace nose 2.

[0102] Next, refer to Figure 5(b) will now describe the atmosphere control unit for controlling the atmosphere inside the furnace nose 2. Figure 5 (b) is a cross-sectional view of the steel strip S, taken at the center of the steel strip S in the width direction, of the furnace nose 2 as it passes through. Multiple heaters 55 are arranged in zones on the outer surface of the iron sheet on the wall of the furnace nose 2. The heaters 55 can be, for example, electric heaters. The heaters 55 control the ambient temperature within the furnace nose 2. Multiple thermocouples 57 are arranged within the furnace nose 2, and the output of the heaters 55 is controlled by a temperature control unit 58 so that the measured ambient temperature within the furnace nose falls within a predetermined range. The temperatures measured by the multiple thermocouples 57 within the furnace nose 2 can be an average value or a temperature measured at a predetermined representative position as the ambient temperature within the furnace nose 2, which can be used as an operating parameter of the furnace nose 2.

[0103] On the other hand, in the furnace nose 2, in order to suppress the smoke generated by the evaporation and solidification of zinc, the atmosphere temperature in the furnace nose 2 is controlled, and the smoke discharge amount from the inside of the furnace nose 2 is adjusted by adjusting the opening of the valve 60 of the furnace nose diffusion pipe 59, thereby adjusting the atmosphere in the furnace nose 2.

[0104] Gas purge equipment

[0105] The structure of the gas purge device 70 is as follows Figure 7 As shown in FIG. A gas purge device 70 blows purge gas from purge nozzles 71 disposed on both the front and back sides of the hot-dip galvanized steel sheet G, which is pulled above the galvanizing bath. This purge gas scrapes off excess molten zinc adhering to the surface of the hot-dip galvanized steel sheet G, thereby adjusting the weight per unit area of ​​the molten zinc.

[0106] The gas purge system 70 consists of a purge nozzle 71, a header 72, a pressure gauge 73, a thermometer 74, a flexible hose 75, a gas heater 76, an air compressor 77, and a nozzle height adjuster, a nozzle-strip distance adjuster, and a nozzle angle adjuster (not shown). The purge nozzle 71 is mounted on the header 72, to which the flexible hose 75 is attached. A gas heater 76 is installed upstream of the piping system as needed, and an air compressor 77 is also located upstream of the heater.

[0107] Air compressed by an air compressor 77 is fed into a manifold 72 via a flexible hose 75. The air is then rectified and contracted in a purge nozzle 71, and discharged from the nozzle outlet at a velocity of several tens to several hundred m / s. A pressure gauge 73 and, if necessary, a thermometer 74 are installed on the manifold 72 to monitor the gas pressure and temperature.

[0108] The output of the air compressor 77 is adjusted so that the pressure measured by the pressure gauge 73 becomes 2 kPa or more and 70 kPa or less. The nozzle height H (distance from the center of the slit of the blow nozzle to the surface of the zinc bath) can be adjusted to 50 to 700 mm by a nozzle height adjustment section not shown. In addition, the nozzle offset O (difference in nozzle height between the two nozzles) can be adjusted within 0 to 5 mm. The distance D between the tip of the nozzle and the steel sheet is adjusted to 5 to 30 mm by a nozzle-steel sheet distance adjustment section not shown. The nozzle angle Θ (angle formed by the plane parallel to the surface of the bath and the tip of the nozzle) can be adjusted within a range of 0 to 75° by a nozzle angle adjustment section not shown.

[0109] Further, a gas heating device 76 that heats the blow gas emitted from the nozzles can be provided as needed depending on the kind of plating, etc. The output of the gas heating device can be adjusted so that the temperature of the blow gas measured at the nozzle header by a thermometer becomes 500 to 700°C. In the case where the gas heating device 76 is provided, it is preferable to provide the above-mentioned thermometer 74 as well.

[0110] <Defect detection device>

[0111] The outline of the defect detection device 80 is shown in Figure 8 The defect detection device 80 includes a light projector 81 and a camera 82. The light projector 81 is a device that irradiates white light or monochromatic light at a constant angle to the surface of the hot-dip galvanized steel sheet G with respect to the direction of travel of the steel sheet. It is preferable that the light projector 81 project parallel light onto the surface of the steel sheet. It is preferable that the camera 82 be provided in multiple units (for example, about 20 units) in the width direction of the hot-dip galvanized steel sheet G, and acquire images from a prescribed angle with respect to the direction of travel of the hot-dip galvanized steel sheet G. When there is no defect on the surface of the steel sheet, the light irradiated from the light projector 81 is specularly reflected on the surface of the steel sheet. On the other hand, when there is a defect, the irradiated light is diffusely reflected on the surface of the steel sheet. This diffusely reflected light is received by the camera 82, and thus it is possible to detect the dross defect X. In the example shown in Figure 8 , the cameras 82 arranged in the width direction of the hot-dip galvanized steel sheet G detect the dross defect X on the inspection line L. It is preferable to use a camera that is capable of detecting a size of about 100 to 200 μm as the size of the dross defect X.

[0112] The dross defect information of the steel strip S that is an output variable when the dross defect prediction model is caused to learn by the machine learning section described later is information obtained by the defect detection device 80. As the dross defect information, it is possible to use the number, density, size of the largest dross defect per unit length, etc. of the dross defects X of a certain size or more detected on the surface or back surface of the hot-dip galvanized steel sheet G, or any information related to the dross defects X that can be a problem in the quality of the hot-dip galvanized steel sheet G.

[0113] Next, an example of the processing in the defect detection device 80 will be described. In the defect detection device 80, images are acquired at a pitch of 0.1 seconds by the camera, the repetitive portions of each image are removed by image processing, and converted into continuous images of a predetermined reference length (for example, 1 m). From the thus obtained images of the reference length, the hot-dip galvanized steel sheet G and the background are separated, and only the image of the hot-dip galvanized steel sheet G is extracted. Also, the dross defect X appears black in appearance. For example, if the extracted image is a color image, it is possible to acquire a black and white image by performing binarization after converting into a gray scale image. Then, the number of pixels of the position at which the black points corresponding to the dross defect X are concentrated is counted, and the area of the dross defect X is converted from the number of pixels. As the size of the dross defect X, for example, the diameter in the case where the measured area of the dross defect X is approximated to a circular shape of equal area can be taken as the size of the dross defect X. The black points on the surface of the hot-dip galvanized steel sheet G existing in the continuous images of the reference length obtained are processed in this way, and the number of dross defects X larger than a predetermined size is calculated. The size of the dross defect X counted as the dross defect X is not particularly limited, but for example, it can be 100 pm or more. Such processing is performed on each of the images of the reference length obtained.

[0114] Figure 3 The continuous hot-dip metal plating apparatus 100 shown can further include a reheating apparatus at a position downstream of the zinc plating bath (further downstream of the gas blowing apparatus 70 in the case where the continuous hot-dip metal plating apparatus 100 described above has the gas blowing apparatus 70) and upstream of the defect detection device 80. The reheating apparatus has an alloying zone, a holding zone, and a final cooling zone, and an induction heating device can be disposed in the alloying zone. The hot-dip galvanized steel sheet after the plating treatment is subjected to an alloying treatment by the reheating apparatus to form an alloyed zinc plating layer having an alloy layer of Zn-Fe alloying reaction on the surface of the steel strip S, thereby forming an alloyed hot-dip galvanized steel sheet. The alloyed hot-dip galvanized steel sheet after the alloying treatment is also subjected to the defect detection device 80 located downstream of the continuous hot-dip metal plating apparatus 100 to acquire dross defect information. This dross defect information is used as an output variable when the dross defect prediction model is learned by the machine learning unit described later, and the dross defect prediction method according to the present disclosure is performed, and if the continuous hot-dip galvanizing apparatus controlled based on the dross defect reduction method is used, an alloyed hot-dip galvanized steel sheet with reduced dross defects can be manufactured. The conditions of the alloying treatment can be in accordance with conventional methods.

[0115] As described above, even in the case where the steel strip is subjected to the alloying treatment, the dross defect information can be acquired by the defect detection device 80 as in the case where the steel strip is not subjected to the alloying treatment. However, the microscopic unevenness (surface roughness) of the surface of the steel strip in the case where the steel strip is subjected to the alloying treatment (the surface of the alloyed hot-dip galvanized steel sheet) is greater than the microscopic unevenness (surface roughness) of the surface of the steel strip in the case where the steel strip is not subjected to the alloying treatment (the surface of the hot-dip galvanized steel sheet). Therefore, in the case where the binarization processing of the captured image is performed to acquire a black-and-white image, the threshold value of the binarization can be made different values in the case where the steel strip is subjected to the alloying treatment and in the case where the steel strip is not subjected to the alloying treatment.

[0116] <1st parameter related to the galvanizing bath>

[0117] In the present embodiment, as the input data of the dross defect prediction model, the 1st parameter related to the galvanizing bath 1 is used. The 1st parameter can be selected based on the structure of the galvanizing bath 1. The 1st parameter can use one or two or more selected from the operating parameters related to the steel strip S, the bath temperature in the galvanizing bath 1 of the galvanizing bath, the information related to the temperature distribution in the galvanizing bath 1, the Al concentration of the galvanizing bath, the bath surface level, and the cumulative time of the passage of the steel strip S, the above-mentioned operating parameters related to the steel strip S including the sheet width of the steel strip S, the sheet thickness, the sheet passage speed (line speed) when the steel strip S passes through the galvanizing bath 1, and the temperature of the steel strip S when immersed into the galvanizing bath. In addition, one or two or more selected from the bath temperature control output based on the heating device 11, the press-in amount of the support roll 4, the billet input amount to the galvanizing bath, the billet composition, and the like can be used. Here, the cumulative time of the passage of the steel strip S can be based on the time point when the galvanizing bath in the galvanizing bath 1 is built, for example. Alternatively, the cumulative time of the passage of the steel strip S can be based on the time point when the device in a part of the galvanizing bath 1 is replaced, or the start of the operation cycle such as the time point when the operation is temporarily stopped and restarted. Furthermore, in the case where the galvanizing bath contains additive elements such as Mg, Ni, Ti, and Si, the concentration of these additive elements in the galvanizing bath can be included as the 1st parameter.

[0118] The dross in the galvanizing bath 1 is generated by the reaction of Fe dissolved from the steel strip S into the galvanizing bath 1 with Al and Zn in the galvanizing bath to form a reaction product. Therefore, the temperature of the galvanizing bath and the Al concentration in the galvanizing bath, which affect the amount of Fe dissolution, are factors that affect the generation of dross. If the dross in the galvanizing bath increases, the dross adhering to the steel sheet together with the molten zinc also increases. Therefore, it is preferable that the first parameter include one or two parameters selected from the temperature of the above galvanizing bath and the Al concentration in the above galvanizing bath. At this time, it is preferable to use either or both of the temperature information in the bath obtained from the bath thermometer 9 and the Al concentration in the plating solution obtained from the bath analysis device 10 as the first parameter. It is also possible to collect a sample from the galvanizing bath and perform ICP analysis on the sample to obtain the Al concentration in the plating solution.

[0119] In addition, the amount of press-in of the backup roll 4 with respect to the steel strip S (or the hot-dip galvanized steel sheet G) also affects the generation of the dross defect X, and therefore can be included in the first parameter. This is because: by press-in of the backup roll 4 toward the steel strip S side, the winding angle of the steel strip S with respect to the backup roll 4 becomes large, and therefore the molten zinc is suppressed from passing between the backup roll 4 and the steel strip S, and as a result, the dross is suppressed from passing between the backup roll 4 and the steel strip S, and the amount of adhesion of the dross to the surface of the steel strip S is reduced.

[0120] <Second Parameter Related to Furnace Nose>

[0121] In the present embodiment, the second parameter related to the furnace nose 2 is used as the input data of the dross defect prediction model. The second parameter related to the furnace nose 2 can be selected based on the structure of the furnace nose 2 shown in FIG. 2. As the second parameter, one or more than two can be selected from any of the parameters indicating the operating state in the furnace nose 2, such as the temperature in the furnace nose 2, the dew point of the atmosphere in the furnace nose 2, the gas flow rate, the gas composition, and the like. Figure 5

[0122] As the second parameter, it is preferable to use one or more than two parameters selected from the dew point in the furnace nose 2, the hydrogen concentration in the furnace nose 2, the oxygen concentration in the furnace nose 2, the temperature of the steel strip S in the furnace nose 2, and the atmosphere temperature in the furnace nose 2. This is because: these parameters are associated with the oxidation behavior of the steel strip S, and therefore have an influence on the amount of Fe dissolution into the galvanizing bath, and are associated with the generation behavior of the dross.

[0123] Here, in the present embodiment, both the first parameter related to the galvanizing bath 1 and the second parameter related to the furnace nose 2 are used as the input data of the dross defect prediction model. The reason for this will be described below.

[0124] Figure 9 is a graph showing the relationship between the dross defect X and the dew point in the furnace nose 2. Figure 9 is a graph showing the relationship between the dross defect X and the dew point in the furnace nose 2. Figure 5 ​The gas supply portion 50 shown in Fig. 1 was used to adjust the dew point of the atmosphere in the furnace nose 2 (the estimated dew point on the plating bath) to a range of -45°C to -20°C, and the number of dross defects X per unit length of the steel strip was measured by the above-described method using the defect detection device 80. Further, the following conditions were set: the bath temperature was 452 to 458°C, the oxygen concentration of the atmosphere in the furnace nose 2 was 10 to 50 ppm, and the hydrogen concentration of the atmosphere in the furnace nose 2 was 1 to 5%. Here, as to the bath composition, Condition A: the Al concentration was 4.3 to 4.9%, the Mg concentration was 0.5 to 0.7%, and the remainder was Zn and Fe, and Condition B: the Al concentration was 0.135 to 0.139%, and the remainder was Zn and Fe. According to Figure 9 It was found that the dew point in the furnace nose 2 as the second parameter had a certain correlation with the number of occurrences of dross defects X. Further, the number of occurrences of dross defects X was different between the bath compositions A and B, and it was considered that there was a certain correlation between the bath composition as the first parameter and the number of occurrences of dross defects X.

[0125] If the dew point in the furnace nose 2 is high, the atmosphere in the furnace nose 2 becomes an atmosphere with a high oxygen potential, and oxidation of the steel strip S is promoted when passing through the furnace nose 2, and an iron oxide film is formed on the surface of the steel strip S. Here, the iron oxide film hinders the diffusion of iron from the base iron to the galvanizing bath, and thus has an effect of suppressing the elution of iron into the galvanizing bath. Dross is a compound of the eluted iron from the steel strip S and Al and Zn in the plating bath, and thus under the high dew point condition where the amount of eluted iron is small, the dross defects X are reduced. On the other hand, under the low dew point condition, the oxidation on the surface of the steel strip S is suppressed, and thus the amount of elution of iron from the surface of the steel strip S to the galvanizing bath increases, and the amount of generation of dross defects X increases. It is considered that the above is the reason for the correlation between the dew point of the atmosphere in the furnace nose 2 and the generation behavior of dross defects X.

[0126] On the other hand, it was considered that the tendency of generation of dross defects X with respect to the dew point of the atmosphere in the furnace nose 2 was different depending on the bath composition based on the following reason. There is a limit amount of saturation of the amounts of eluted Fe and Al with respect to the temperature of the galvanizing bath. Therefore, there is a relationship in which if the amount of elution of one of Fe and Al increases, the amount of elution of the other decreases. That is, if the Al concentration in the galvanizing bath increases, the Fe concentration decreases, and thus under the above-described Condition A, the amount of elution of Fe to the galvanizing bath relatively decreases. Thus, the dross in the galvanizing bath decreases, and the generation of dross defects X can be suppressed.

[0127] Further, in the above-described Condition B, the Al concentration in the galvanizing bath was low, and thus the amount of elution of Fe to the galvanizing bath was large, and the amount of generation of dross defects X was large. In contrast, in the above-described Condition A, the Al concentration in the galvanizing bath was high, and thus the amount of elution of Fe to the galvanizing bath was small, and the amount of generation of dross defects X was small. Figure 9In the case where the determination criterion for the product eligibility is selected as the number of dross generation per unit length of 5, the dew point of the atmosphere in the furnace nose 2 is preferably -40°C or higher under the condition A and -35°C or higher under the condition B, and the appropriate range of the dew point of the atmosphere in the furnace nose 2 varies depending on the bath composition. That is, under the condition A where the Al concentration in the plating bath is high, in order to suppress the generation of dross defects, the dew point in the furnace nose is made lower than that under the condition B as the operation parameter of the furnace nose.

[0128] Also, Figure 9 The relationship between the dew point of the atmosphere in the furnace nose 2 and the generation behavior of the dross defects X also varies depending on the temperature of the plating bath. This is because the amount of components such as Fe and Al dissolved with respect to the molten zinc varies depending on the temperature of the plating bath.

[0129] According to the above, it is important to use both the first parameter related to the galvanizing bath 1 and the second parameter related to the furnace nose 2 as the input data of the dross defect prediction model. In this way, by combining a plurality of operation parameters, it is possible to first set appropriate operation conditions for sufficiently reducing the dross defects X.

[0130] The various operation conditions of the continuous hot dip galvanizing apparatus 100 described above are set by a host computer including a process control computer, and the host computer can acquire these operation data.

[0131] The above describes the continuous hot dip galvanizing apparatus 100, and the following describes the dross defect prediction device, the dross defect prediction model generation device, and the operation performance database (DB) that constitute the dross defect prediction system.

[0132] <Dross Defect Prediction Model Generation Device>

[0133] In the present embodiment, the dross defect prediction model generation device refers to the operation data related to the galvanizing bath, the operation data related to the furnace nose 2, and the performance data of the dross defect information, and generates the dross defect prediction model by machine learning. Figure 10 is a functional block diagram showing the structure of one example of the dross defect prediction device. As Figure 10 shown, the dross defect prediction model generation device 83 has an acquisition section 831, a storage section 832, an output section 833, and a machine learning section 834.

[0134] The acquisition section 831 acquires the first parameter related to the galvanizing bath 1 and the second parameter related to the furnace nose, and delivers them to the machine learning section 834. The acquisition section 831 can include any interface capable of acquiring the operation data from the operation performance DB 85. For example, the acquisition section 831 can also include a communication interface for acquiring the operation data from the operation performance DB 85. In this case, the acquisition section 831 can receive the operation data from the operation performance DB 85 in accordance with a prescribed communication protocol.

[0135] The storage section 832 includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or a combination of at least two of these memories. The semiconductor memory is, for example, a RAM (random access memory) or a ROM (read only memory). The RAM is, for example, an SRAM (static random access memory) or a DRAM (dynamic random access memory). The ROM is, for example, an EEPROM (electrically erasable programmable read only memory). The storage section 832 functions as, for example, a main storage device, an auxiliary storage device, or a cache memory. The storage section 832 stores any information used in the operation of the dross defect prediction model generation device. The storage section 832 stores, for example, the operation data acquired by the acquisition section from the operation history DB 85 and the dross defect prediction model generated by the machine learning section. The storage section 832 can also store, for example, system programs and application programs, and the like.

[0136] The machine learning section 834 performs machine learning with the operation data supplied from the acquisition section 831 as an input variable and the dross defect information of the steel strip S observed on the downstream side of the galvanizing bath 1 as an output variable. The machine learning section 834 performs machine learning, for example. The machine learning section 834 generates a dross defect prediction model by performing machine learning. The dross defect prediction model is a machine learning model for calculating a predicted value of the dross defect information of the steel strip S from an operation condition acquired by the acquisition section 841 of the dross defect prediction device 84 described later. The machine learning section 834 supplies the generated dross defect prediction model to the output section 833.

[0137] The machine learning section 834 includes one or more processors. In one embodiment, the "processor" is a general-purpose processor or a dedicated processor customized for a specific processing, but is not limited to these.

[0138] The machine learning section 834 can be, for example, any general-purpose electronic device such as a PC (Personal Computer) or a smartphone. The machine learning section 834 is not limited to these, but can be one or a plurality of server devices communicable with each other, or other electronic devices dedicated to the dross defect prediction system.

[0139] As the method of machine learning, a publicly known machine learning algorithm can be applied. The machine learning section 834 can execute, for example, machine learning selected from a neural network (including deep learning, a convolutional neural network, and a recurrent neural network, etc.), decision tree learning, random forest, support vector regression, Gaussian process, and k-nearest neighbor method to generate the dross defect prediction model. The machine learning section 834 can use these machine learning algorithms individually, or an ensemble model obtained by combining these machine learning algorithms. In addition, the machine learning section 834 can appropriately update the dross defect prediction model using the latest learning data.

[0140] The output section 833 supplies the dross defect prediction model supplied from the machine learning section 834 to the dross defect prediction device 84. The output section 833 can include any interface that can supply the dross defect prediction model to the dross defect prediction device 84. For example, the output section 833 can include a communication interface for supplying the dross defect prediction model to the dross defect prediction device 84. In this case, the output section 833 can transmit the dross defect prediction model to the dross defect prediction device 84 in accordance with a prescribed communication protocol.

[0141] Figure 13 is a flowchart showing one example of the operation of the dross defect prediction model generation device 83. The generation method of the dross defect prediction model based on the dross defect prediction model generation device 83 will be described with reference to Figure 13

[0142] In step S101, the machine learning section 834 of the dross defect prediction model generation device 83 acquires the operation data including the first parameter and the second parameter from the operation history DB 85 via the acquisition section 831.

[0143] In step S102, the machine learning section 834 of the dross defect prediction model generation device 83 executes machine learning with the operation data acquired in step S101 as the input variable and the dross defect information as the output variable. Thereby, the machine learning section 834 generates the dross defect prediction model.

[0144] When the operation data further includes the third parameter, the machine learning section 834 can execute machine learning with the operation data further including the third parameter as the input variable in step S102, with the dross defect information of the steel strip detected on the downstream side of the galvanizing bath 1 as the output variable.

[0145] <Operation History Database>

[0146] ​The operation history DB 85 includes a first parameter related to the galvanizing bath 1, a second parameter related to the furnace nose 2, and dross defect information related to a dross defect of the steel strip S detected on the downstream side of the galvanizing bath 1. The operation history DB 85 can arbitrarily include a third parameter related to the gas purging device 70. Further, the operation history DB 85 can be provided separately from the dross defect prediction model generation device 83, or can constitute a single device together with the dross defect prediction model generation device 83.

[0147] An upper computer including a process control computer can select the first parameter from various information used for setting and controlling the operation parameter of the galvanizing bath 1 and supply to the operation history DB 85. The upper computer sets, for example, the plate width, the plate thickness, the passing speed, the temperature of the steel strip S at the time of immersion into the galvanizing bath, and the like, which are used for controlling the operation parameter of the continuous hot dip galvanizing apparatus 100, and the upper computer can acquire these operation data. In addition, the upper computer can also collect the bath temperature obtained from the bath thermometer 9, the Al concentration in the bath obtained from the bath analysis device 10, and the like. However, it is preferable that the upper computer transfers to the operation history DB 85 after appropriately processing the continuous operation data obtained from various measurers arranged in the galvanizing bath 1. For example, the upper computer can transfer the operation data after performing an averaging process every certain length of the steel strip S (for example, 10 m of the length of the steel strip S) or every certain time (for example, 2 seconds) to the operation history DB 85.

[0148] An upper computer including a process control computer can select the second parameter from various information used for setting and controlling the operation parameter of the furnace nose 2 and supply to the operation history DB 85. As the second parameter, for example, the dew point in the furnace nose 2, the hydrogen concentration in the furnace nose 2, the oxygen concentration in the furnace nose 2, the temperature of the steel strip S in the furnace nose 2, the atmosphere temperature in the furnace nose, the gas flow rate, the gas composition, and the like can be cited. It is preferable that the upper computer transfers to the operation history DB 85 after appropriately processing the continuous operation data obtained from various measurers arranged in the furnace nose 2. As described above, for example, the upper computer can transfer the operation data after performing an averaging process every certain length of the steel strip S (for example, 10 m of the length of the steel strip S) or every certain time (for example, 2 seconds) to the operation history DB 85.

[0149] In addition, the upper computer including the process control computer can acquire the dross defect information based on the image data acquired from the defect detection device 80 and transfer to the operation history DB 85 by the above-described method.

[0150] The host computer preferably associates the dross defect information with the position information in the longitudinal direction of the hot-dip galvanized steel sheet G to deliver to the operation history DB 85. In one example, the host computer determines the position within the steel strip S from the distance from the leading end of the steel strip S (the welded portion at the leading end), traces the position of the welded portion from the feeding side of the continuous hot-dip galvanizing apparatus 100, and associates the second parameter relating to the furnace snout 2 with the position information in the longitudinal direction of the steel strip S at the stage where the steel strip S passes through the furnace snout 2. In addition, the host computer associates the second parameter relating to the galvanizing bath 1 with the position information in the longitudinal direction of the steel strip S at the stage where the steel strip S passes through the galvanizing bath 1. Also, the host computer associates the dross defect information with the position information in the longitudinal direction of the steel strip S at the stage where the steel strip S passes through the defect detection device 80. In this way, the host computer calculates the representative value of the operation data at the position in the longitudinal direction of the steel strip S from the second parameter relating to the furnace snout 2, the first parameter relating to the galvanizing bath 1, and the dross defect information associated with the position information within the steel strip S, for example, by averaging treatment every certain length of the steel strip S (for example, every 10 m). Also, the calculated representative value is aggregated for each position in the longitudinal direction of the steel strip S, and accumulated in the operation history DB 85 as a set of operation data. At this time, as the length of the steel strip S, for example, a length arbitrarily decided within the range of 1.0 to 100 m can be used. Furthermore, it can be judged that the time difference from when the specific portion of the steel strip S passes through the furnace snout 2 to when it passes through the galvanizing bath 1 is sufficiently small, and therefore the host computer can also use the information acquired at the same timing of the furnace snout 2 and the galvanizing bath 1 when the steel strip S passes through the galvanizing bath 1 as a set of operation data.

[0151] The operation history DB 85 stores a plurality of data sets obtained by associating the input variables and the output variables collected as described above. In addition, the operation history DB 85 preferably contains information relating to the steel grade and the size (the sheet thickness and the sheet width of the steel strip S) as the manufacturing object. In addition, each data set can also contain the position information in the longitudinal direction of the steel strip S. As the number of data in the operation history DB 85, it is preferable that there are 750 or more for each steel strip S. In addition, in the operation history DB 85, data relating to at least 5 or more coils of the steel strip S are accumulated, preferably data relating to 20 or more coils of the steel strip S are accumulated, and more preferably data relating to 100 or more coils of the steel strip S are accumulated.

[0152] The machine learning unit 834 uses the database thus created to generate a dross defect prediction model by machine learning using, as input variables, one or more operation data selected from at least one of the first parameter relating to the galvanizing bath and the second parameter relating to the furnace snout 2, and, as an output variable, the dross defect information of the steel strip S observed on the downstream side of the galvanizing bath corresponding to the input variable.

[0153] Figure 14 is a schematic diagram showing the structure of the dross defect prediction model. The structure of the dross defect prediction model generated by the machine learning section will be described with reference to Figure 14

[0154] Among the input items of the dross defect prediction model generated by the machine learning section 834, for example, the operating conditions including the first operating conditions related to the galvanizing bath 1 and the second operating conditions related to the furnace snout 2, which become the prediction targets of the dross defect information, are input via the acquisition section 841 of the dross defect prediction device 84. The dross defect information is predicted based on the operating conditions that have been input to the input items, and the dross defect information is output at the output items.

[0155] Dross Defect Prediction Device

[0156] Figure 12 is a functional block diagram showing the structure of the dross defect prediction device. One example of the dross defect prediction device included in the dross defect prediction system will be described with reference to Figure 12

[0157] As shown in Figure 12 , the dross defect prediction device 84 has an acquisition section 841, an output section 842, a storage section 843, and a control section 844.

[0158] The acquisition section 841, for example, includes any interface capable of acquiring the dross defect prediction model generated by the machine learning section 834 from the dross defect prediction model generation device 83. For example, the acquisition section 841 can include a communication interface for acquiring the dross defect prediction model from the dross defect prediction model generation device 83. In this case, the acquisition section 841 can receive the dross defect prediction model from the machine learning section 834 in accordance with a prescribed communication protocol.

[0159] In addition, the acquisition section 841, for example, acquires the operating conditions from the upper computer that controls the continuous hot dip galvanizing apparatus 100. For example, the acquisition section 841 can include a communication interface for acquiring the operating conditions from the upper computer. In this case, the acquisition section 841 can receive the operating conditions from the upper computer in accordance with a prescribed communication protocol.

[0160] ​​The acquisition unit 841 can also acquire input information based on the user's operation. In this case, the dross defect prediction device 84 further has an input unit including one or more input interfaces that detect a user input and acquire input information based on the user's operation. For example, the input unit is a physical key, an electrostatic capacity key, a touch screen provided integrally with a display of the output unit, or a microphone that accepts a voice input, but is not limited to these. For example, the input unit accepts an input of the operating condition with respect to the dross defect prediction model acquired by the acquisition unit 841 from the dross defect prediction model generation device 83.

[0161] The storage unit 843 includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or a combination of at least two of these. The semiconductor memory is, for example, a RAM (random access memory) or a ROM (read only memory). The RAM is, for example, a SRAM (static random access memory) or a DRAM (dynamic random access memory). The ROM is, for example, an EEPROM (electrically erasable programmable read only memory). The storage unit 843 functions as, for example, a main storage device, an auxiliary storage device, or a cache memory. The storage unit 843 stores arbitrary information used in the operation of the dross defect prediction device. The storage unit 843 stores, for example, the dross defect prediction model acquired by the acquisition unit 841 from the dross defect prediction model generation device 83, the operating condition acquired by the acquisition unit 841 from the upper computer, and the dross defect information predicted by the control unit 844. For example, the storage unit 843 can also store a system program and an application program, and the like.

[0162] The control unit 844 includes one or more processors. In one embodiment, the "processor" is a general-purpose processor or a dedicated processor customized for a specific process, but is not limited to these. The control unit 844 is communicably connected to each of the structural units that constitute the dross defect prediction device, and controls the operation of the entire dross defect prediction device 84.

[0163] The control unit 844 can be, for example, a PC (Personal Computer) or an arbitrary general-purpose electronic device such as a smart phone. The control unit 844 is not limited to these, and can be one or a plurality of server devices that are communicable with each other, or other electronic devices dedicated to the dross defect prediction system.

[0164] The control section 844 calculates the prediction of the dross defect information by the dross defect prediction model acquired from the dross defect prediction model generation device 83, based on the operation condition acquired via the acquisition section 841.

[0165] The output section 842 supplies the prediction of the dross defect information calculated by the control section 844 to the operation condition setting device 90 described later. The output section 842 can include any interface capable of supplying the prediction of the dross defect information to the operation condition setting device 90. For example, the output section 842 can include a communication interface for supplying the prediction of the dross defect information to the operation condition setting device 90. In this case, the output section 842 can transmit the prediction of the dross defect information to the operation condition setting device 90 in accordance with a prescribed communication protocol.

[0166] In addition, the output section 842 can also include one or more output interfaces that output information to notify a user. The output interface is, for example, a display. The display is, for example, an LCD (liquid crystal display) or an organic EL (electro luminescence) display. The output section 842 outputs data obtained by the operation of the dross defect prediction device. The output section 842 can be connected to the dross defect prediction device 84 as an external output device instead of being equipped in the dross defect prediction device 84. As the connection method, for example, any of USB, HDMI (registered trademark), Bluetooth (registered trademark), or the like can be used. The output section 842 is, for example, a display that outputs information as an image or a speaker that outputs information as sound, but is not limited to these. For example, the output section 842 prompts the user with the prediction of the dross defect information predicted by the control section 844. The user can appropriately set the operation condition of the continuous hot dip galvanizing apparatus 100 based on the prediction of the dross defect information prompted by the output section 842.

[0167] Figure 15 is a flowchart showing an example of the operation of the dross defect prediction device 84 of Figure 12 The dross defect prediction method of calculating the prediction of the dross defect information will be mainly described with reference to Figure 15

[0168] In step S201, the control section 844 of the dross defect prediction device 84 acquires the dross defect prediction model from the dross defect prediction model generation device 83 via the acquisition section 841, the dross defect prediction model being a model of machine learning performed based on operation data including the first parameter related to the galvanizing bath 1 and the second parameter related to the furnace nose 2 as input variables and the dross defect information of the steel strip S as an output variable. ​

[0169] In step S202, the control section 844 of the dross defect prediction device 84 receives the operation conditions via the acquisition section 841.

[0170] In step S203, the control section 844 of the dross defect prediction device 84 calculates the predicted value of the dross defect information based on the operation conditions input in step S202 and by the dross defect prediction model acquired in step S201.

[0171] In step S204, the control section 844 of the dross defect prediction device 84 outputs the predicted value of the dross defect information calculated in step S203 as needed via the output section 842.

[0172] In the above embodiment, the dross defect prediction device 84 is provided with the acquisition section 841 that acquires the operation conditions and the output section 842 that outputs the predicted value of the dross defect information, but in addition to the acquisition section 841 and the output section 842 provided in the dross defect prediction device 84, an input section that performs acquisition of the operation conditions and an output section that performs output of the predicted value of the dross defect information can be provided separately from the dross defect prediction device 84. The dross defect prediction terminal system according to an embodiment is provided with: the dross defect prediction device 84 described above; and a terminal device that receives a user input related to a change in the operation conditions described above and transmits user input information based on the user input to the dross defect prediction device 84 described above, and the dross defect prediction device 84 can include a control section 844 that changes at least a part of the operation conditions based on the user input information as changed operation conditions and calculates the predicted value of the dross defect information of the steel strip S based on the changed operation conditions and by the dross defect prediction model.

[0173] According to the dross defect prediction terminal system, the user input is received by the terminal device and the predicted value of the dross defect information is calculated based on the operation conditions changed based on the user input information, so the user can predict the change in the dross defect information accompanying the change in the operation conditions and can rapidly change the continuous hot dip galvanizing apparatus 100 to the appropriate operation conditions.

[0174] The terminal device is an information processing terminal, and in one example, has an input / output section, a communication section, a storage section, and a control section.

[0175] The input / output section has an input interface that detects a user input related to a change in the operating conditions of the continuous hot-dip galvanizing apparatus 100 and delivers the user input information to the control section. The input interface can be, for example, any input interface including a physical key, an electrostatic capacitance key, a touch panel provided integrally with a panel display, various click devices, a microphone that accepts a voice input, a camera that acquires a captured image or an image code, and the like. In addition, the input / output section can also have an output interface that outputs the dross defect prediction information acquired from the dross defect prediction device to the user. The output interface can be, for example, any output interface including an external or internal display that outputs information as an image, a video, a speaker that outputs information as a sound, or a connection interface with an external output device.

[0176] The storage section includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or a combination of at least two of these. The semiconductor memory is, for example, a RAM (random access memory) or a ROM (read only memory). The RAM is, for example, an SRAM (static random access memory) or a DRAM (dynamic random access memory). The ROM is, for example, an EEPROM (electrically erasable programmable read only memory). The storage section functions, for example, as a main storage device, an auxiliary storage device, or a cache memory. The storage section stores any information used in the operation of the terminal device. The storage section stores, for example, the user input information and the dross defect prediction information acquired from the dross defect prediction device 84. The storage section can also store, for example, system programs and application programs, and the like.

[0177] The control section includes one or more processors. In one embodiment, the "processor" is a general-purpose processor or a dedicated processor customized for a specific process, but is not limited to these. The control section is communicably connected to each of the structural sections that constitute the terminal device and controls the operation of the entire terminal device.

[0178] The communication section supplies the user input information based on the user input to the above-described dross defect prediction device 84. The communication section can include any interface that can supply the user input information to the dross defect prediction device 84. For example, the communication section can include a communication interface for supplying the user input information to the dross defect prediction device 84. In this case, the communication section can transmit the user input information to the dross defect prediction device 84 in accordance with a prescribed communication protocol.

[0179] As the terminal device, a tablet terminal of a touch panel type, a smartphone, a personal computer, and the like can be given. The user input is accepted by the terminal device, whereby the user can obtain the predicted value of the dross defect information in the case where the operating conditions of the continuous hot-dip galvanizing apparatus 100 are virtually changed, regardless of the place. The user can, for example, operate the terminal device in an office of a factory equipped with the continuous hot-dip galvanizing apparatus 100 to calculate the predicted value of the dross defect information, which contributes to the decision of the operating conditions of the continuous hot-dip galvanizing apparatus 100.

[0180] An example of the operation of the dross defect prediction terminal system will be described. First, the input / output section of the terminal device accepts the user input relating to the change of the operating conditions and supplies it to the transmission section. The user input relating to the change of the operating conditions includes information for changing at least a part of the operating conditions. As the user input, for example, a change value of the first operating conditions relating to the galvanizing bath 1, a change value of the second operating conditions relating to the furnace nose 2, or a change value of the third operating conditions relating to the gas purging apparatus 70 can be given. Next, the transmission section transmits the user input information based on the user input to the dross defect prediction device 84.

[0181] Next, the acquisition section 841 of the dross defect prediction device 84 acquires the user input information from the above-described terminal device. The acquisition section 841 supplies the acquired user input information to the control section 844. The control section 844 changes at least a part of the operating conditions as the changed operating conditions based on the user input information. For example, in the case where the user input information includes a change value of the first operating conditions relating to the galvanizing bath 1, the control section 844 can change the first operating conditions based on the user input information.

[0182] Next, the control section 844 calculates the predicted value of the dross defect information of the steel strip S by the dross defect prediction model with the changed operating conditions as the input variable. The control section 844 of the dross defect prediction device 84 transmits the calculated predicted value of the dross defect information to the communication section of the terminal device through the output section 842. The communication section of the terminal device supplies the acquired predicted value of the dross defect information to the input / output section of the terminal device. The input / output section of the terminal device outputs the predicted value of the dross defect information to the user.

[0183] <Method for Reducing Dross Defects>

[0184] In the dross defect reduction method according to the present embodiment, a step of calculating a predicted value of dross defect information of the steel strip S on the downstream side of the galvanizing bath 1 using the dross defect prediction method described above, and a step of resetting the operation conditions based on the predicted value of the dross defect information described above are included. The dross defect reduction method according to the present embodiment can be implemented, for example, using a dross defect reduction system obtained by combining an operation condition setting device to the dross defect prediction system including the dross defect prediction device 84.

[0185] In Figure 11 a summary of the dross defect reduction method is shown. As shown in Figure 11 , the dross defect prediction device 84 acquires the operation conditions from a host computer 95 including a process control computer. In addition, the dross defect prediction device 84 can also acquire the operation conditions directly from various measuring devices included in the continuous hot dip galvanizing apparatus 100.

[0186] The dross defect prediction device 84 can calculate the predicted value of the dross defect information at a prescribed position in the longitudinal direction of the steel strip S after passing through the galvanizing bath 1 and before the portion reaches the defect detection device 80. Alternatively, the dross defect prediction system can perform prediction of the dross defect information after the steel strip S passes through the gas blowing apparatus at a prescribed position in the longitudinal direction of the steel strip S and before the portion reaches the defect detection device 80. It is preferable that the dross defect prediction device 84 perform the calculation of the predicted value immediately after the steel strip S passes through the galvanizing bath 1 or immediately after the steel strip S passes through the gas blowing apparatus 70. The dross defect prediction device 84 predicts the predicted value of the dross defect information generated in the steel strip S before the defect detection device 80 detects the dross defect information, so that the operation conditions can be changed promptly to minimize the length of the dross defect X generated.

[0187] Here, it is preferable that the dross defect prediction device 84 perform the calculation of the predicted value of the dross defect information continuously in the longitudinal direction of the steel strip S. In one example, the dross defect prediction device 84 predicts the dross defect information at an arbitrary pitch set in the range of 1.0 to 100 m in the traveling direction of the steel strip S and supplies the prediction results to the output section. In this way, if the dross defect prediction device 84 calculates the predicted value of the dross defect information at an arbitrary interval set in the range of 1.0 to 100 m in the traveling direction of the steel strip S, the dross defect X can be predicted with high accuracy even in the case where the first operation conditions related to the galvanizing bath 1 and the second operation conditions related to the furnace nose 2 are changed during the period in which one coil of the steel strip S is processed through the continuous hot dip galvanizing apparatus 100.

[0188] The operating condition setting device 90 resets the operating conditions of the continuous hot-dip galvanizing equipment 100 based on the predicted value of the dross defect information calculated by the dross defect prediction device 84 as described above. Figure 11 As shown, the operating condition setting device 90 obtains the predicted value of dross defect information from the dross defect prediction device 84. Furthermore, in one example, the determination unit 901 of the operating condition setting device 90 obtains a preset dross defect upper limit from a host computer. The determination unit 901 compares the predicted value of the dross defect information with the upper limit. If the predicted value of the dross defect information exceeds the upper limit, the determination unit 901 determines that the determination result is NG. If the predicted value of the dross defect information is below the upper limit, the determination unit 901 determines that the determination result is OK. If the determination result is NG, the determination unit 901 notifies the operating condition resetting unit 902 of the change in operating conditions, which then modifies the operating conditions of the continuous hot-dip galvanizing facility 100. If the determination result is OK, the determination unit 901 does not notify the operating condition resetting unit 902 of the change in operating conditions. If the determination result is OK, the determination unit 901 transmits a control instruction to maintain the operating conditions at that point in time to the operating condition control unit of each device included in the continuous hot-dip galvanizing facility 100. Notified of the change in operating conditions, the operating condition reset unit 902 can reset the operating conditions of, for example, the nose 2, the galvanizing tank 1, and the gas purge system 70.

[0189] Specifically, upon receiving notification of the operating condition changes, the operating condition resetting unit 902 can prioritize changing highly responsive operating conditions. Examples of highly responsive operating conditions include the sheet metal threading speed, the amount of pressure applied by the backup rolls 4 relative to the steel strip S, and the gas pressure of the gas purge system 70. Furthermore, if changes to these operating conditions alone do not sufficiently reduce the dross defect X, the operating condition resetting unit 902 can additionally reset at least one of, for example, the set value of the bath temperature control output of the galvanizing bath 1 and the flow rate ratio of the mixed gas supplied to the snout 2. Furthermore, the re-set operating conditions can be used as input parameters of the dross defect prediction model to re-predict dross defect information. After confirming whether the predicted dross defect information value is less than the dross defect upper limit, the set values ​​of the operating conditions are determined.

[0190] Furthermore, it is preferable that the operating condition resetting unit 902 sets upper and lower limits for various operating conditions when resetting the operating conditions, and resets the operating conditions within the set ranges. By setting upper and lower limits for various operating conditions and resetting the operating conditions within the set ranges, the operating condition resetting unit 902 can appropriately prevent the occurrence of defects other than the dross defect X.

[0191] The upper computer 95 can set a dross defect upper limit value in advance. As the dross defect upper limit value, for example, the number of dross defects X per unit length in the long side direction of the steel strip S can be cited. The upper computer 95 can set, for example, one or less of the dross defects X per 1 km in the long side direction of the steel strip S as the dross defect upper limit value.

[0192] As the operating conditions which the operating condition re-setting section 902 re-sets, the first operating conditions relating to the galvanizing bath 1 and the second operating conditions relating to the furnace nose 2 can be included. As the first operating conditions relating to the galvanizing bath 1, for example, one or more than two selected from the strip passing speed at the time when the steel strip S passes through the galvanizing bath 1, the amount of press-in of the support roll 4 in the galvanizing bath 1 with respect to the steel strip S, the set value of the bath temperature control output of the galvanizing bath 1, and the amount of the billet 8 fed to the galvanizing bath 1 can be included.

[0193] As the second operating conditions relating to the furnace nose 2, for example, one or more than two selected from the flow ratio of the mixed gas supplied to the furnace nose 2, the opening degree of the valve 60 of the furnace nose diffusion pipe 59, and the set value of the heater 55 in the furnace nose 2 can be included.

[0194] The operating condition re-setting section 902 can also change the set condition of the gas pressure of the gas purging device 70 based on the dross defect information. Also, the operating condition re-setting section 902 can re-set the operating conditions of the continuous annealing furnace 30. As the operating conditions of the continuous annealing furnace 30 which the operating condition re-setting section 902 can re-set, the air ratio of the atmosphere in the heating zone 20 in the case where the heating zone 20 is a direct-fired furnace, the kind of gas supplied from the gas supply system of the soaking zone 21, the amount of gas supplied from the gas supply system of the soaking zone 21, the flow rate of the cooling gas as the operating condition of the cooling zone 23, the kind of gas, and the gas concentration, and the like can be cited.

[0195] Among these operating conditions, it is particularly preferable that the operating condition re-setting section 902 re-set the strip passing speed at the time when the steel strip S passes through the galvanizing bath 1 and the amount of press-in of the support roll 4 in the galvanizing bath 1 with respect to the steel strip S. The operating condition re-setting section 902 re-sets the strip passing speed at the time when the steel strip S passes through the galvanizing bath 1 and the amount of press-in of the support roll 4 in the galvanizing bath 1 with respect to the steel strip S, so that the dross defects X can be immediately reduced by the re-setting of the operating conditions. The operating condition re-setting section 902 can re-set the strip passing speed at the time when the steel strip S passes through the galvanizing bath 1 and the amount of press-in of the support roll 4 in the galvanizing bath 1 with respect to the steel strip S with reference to the mechanical response time of various control means. Also, in order to reduce the dross defects X, the operating condition re-setting section 902 can also re-set the gas pressure of the gas purging device 70.

[0196] On the other hand, the operation condition re-setting section 902 re-sets the flow ratio of the mixed gas supplied to the furnace nose 2, thereby being able to change at least one of the dew point in the furnace nose 2, the hydrogen concentration in the furnace nose 2, and the oxygen concentration in the furnace nose 2. The operation condition re-setting section 902 re-sets the flow ratio of the mixed gas supplied to the furnace nose 2, thereby being able to change the atmosphere in the furnace nose 2 in a short time of about 3 minutes. The atmosphere in the furnace nose 2 suppresses the elution of iron with respect to the galvanizing bath through the change in the state of the iron oxide film formed on the surface of the steel strip S, thereby suppressing the dross defect X. The operation condition re-setting section 902 re-sets the flow ratio of the mixed gas supplied to the furnace nose 2, thereby being able to reduce the generation of the dross defect X with a higher effect. It is preferable that the operation condition re-setting section 902 re-sets the sheet passage speed of the steel strip S when passing through the galvanizing bath 1, the press-in amount of the support roll 4 with respect to the steel strip S in the galvanizing bath 1, and the flow ratio of the mixed gas supplied to the furnace nose 2, with reference to the dross defect information.

[0197] Also, the operation condition re-setting section 902 can re-set the set value of the bath temperature control output of the galvanizing bath 1. The operation condition re-setting section 902 re-sets the set value of the bath temperature control output of the galvanizing bath 1 based on the dross defect information, thereby being able to appropriately reduce the dross defect X. This is because the bath temperature has a large influence on the amount of Fe elution from the steel strip S to the galvanizing bath. From the viewpoint of improving responsiveness, it is preferable that the operation condition re-setting section 902 re-sets the set value of the bath temperature control output of the galvanizing bath 1 as the operation condition, in combination with the re-setting of the other operation conditions. In particular, the flow ratio of the mixed gas supplied to the furnace nose 2 and the appropriate value of the gas diffused from the furnace nose 2 to the outside of the furnace for reducing the dross defect X vary depending on the bath temperature of the galvanizing bath 1, and therefore it is preferable that the operation condition re-setting section 902 re-sets in combination the second operation condition related to the furnace nose 2 and the first operation condition related to the galvanizing bath 1.

[0198] The temperature of the steel strip S immersed into the galvanizing bath 1 is a factor that influences the amount of Fe eluted to the galvanizing bath 1. Therefore, the operation condition re-setting section 902 can further reduce the dross defect X by changing the control output set value of the electric heater in the furnace nose 2 provided for heating the furnace nose 2 and the operation condition of the cooling belt 23 in the continuous annealing furnace 30 that is a preceding step of the furnace nose 2.

[0199] The composition of the galvanizing bath is one factor that determines the amount of Fe elution from the steel strip S with respect to the galvanizing bath. Therefore, the operation condition re-setting section 902 can further reduce the dross defect X by re-setting the timing of the billet 8 to be charged with respect to the galvanizing bath and changing the bath composition within a management range.

[0200] The surface properties of the steel strip S are a factor that determines the amount of Fe elution from the steel strip S into the galvanizing bath. As the surface properties of the steel strip S that determine the amount of Fe elution, there are included the degree of oxidation of the surface of the steel strip S and the amount of enrichment of the strengthening elements such as Si and Mn in the surface of the steel strip S. Thus, it is also possible to further reduce the dross defects X by the operation condition re-setting section 902 re-setting the flow rate of the gas supplied to the soaking zone 21 and used for adjusting the dew point in the furnace, or the air ratio of the heating zone 20 in the case of the direct-fired furnace, to further reduce the dross defects X.

[0201] Thus, according to the present dross defect reduction method, in the continuous hot dip galvanizing apparatus 100, it is possible to maintain appropriate operation conditions for reducing the dross defects X within the coil of the steel strip S and between the coils, and thus it is possible to produce a hot dip galvanized steel sheet G that is high in quality and good in yield. That is, the present disclosure also relates to a method of producing a hot dip galvanized steel sheet by forming a zinc plated layer on the surface of the above-described steel strip using the above-described continuous hot dip galvanizing apparatus controlled based on the above-described dross defect reduction method to form a hot dip galvanized steel sheet.

[0202] <Other>

[0203] The dross defect prediction model generation device 83 can also include a fourth parameter related to the continuous annealing furnace 30 in the input variables to create the dross defect prediction model. In this case, the operation condition re-setting section 902 can re-set the fourth operation condition related to the continuous annealing furnace 30. As the fourth parameter related to the continuous annealing furnace 30, it is possible to use one or more than two information obtained by a higher computer represented by a process control computer for controlling the continuous annealing furnace 30 as an operation parameter of annealing from the heating zone 20, the soaking zone 21, and the cooling zone 22, 23 (the first cooling zone 22, the second cooling zone 23) in the continuous annealing furnace 30. Figure 3

[0204] ​As the fourth parameter related to the heating zone 20, at least one selected from the time required for the steel strip S to pass through the heating zone 20 from the steel strip entrance to the steel strip exit, the temperature rise of the steel strip S in the heating zone 20, and the average temperature rise rate of the steel strip S in the heating zone 20 can be used. As the operating parameter related to the soaking zone 21, at least one of the soaking temperature, which is the average temperature of the steel strip S in the soaking zone 21, and the soaking time, which is the time required for the steel strip S to pass through the soaking zone 21 can be used. As the operating parameter related to the first cooling zone 22, at least one selected from the time required for the steel strip S to pass through the first cooling zone 22, the temperature drop of the steel strip S in the first cooling zone 22, and the average cooling rate of the steel strip S in the first cooling zone 22 can be used. Similarly, as the operating parameter related to the second cooling zone, the time required for the steel strip S to pass through the second cooling zone 23, the temperature drop of the steel strip S in the second cooling zone 23, and the average cooling rate of the steel strip S in the second cooling zone 23 can be used. Furthermore, the dross defect prediction device 84 can also use information related to the dew point as the fourth parameter. For example, it is preferable that the dross defect prediction device 84 use the soaking time and the dew point in the soaking zone 21 as the fourth parameter.

[0205] The fourth parameter is not limited to the above-mentioned parameters. The slag defect prediction model generation device 83 can also use the output value of the control system for controlling the operating conditions of the continuous annealing furnace 30, such as the control output value of the heating zone heating device in the heating zone 20 and the control output value of the cooling device in the cooling zones 22 and 23, as the fourth parameter.

[0206] The fourth parameter related to the continuous annealing furnace 30 described above affects the state of oxides on the surface of the steel strip S when the steel strip S is charged into the furnace nose 2. Therefore, the fourth parameter affects the amount of Fe eluted from the steel strip S into the galvanizing bath after the steel strip S is charged into the galvanizing bath 1, and is therefore associated with the generation behavior of dross defects X.

[0207] Examples <Example 1>

[0208] The present disclosure will be described in more detail below based on examples. In the following Example 1, the dross defect reduction method of this embodiment is applied to a case where a galvanized layer is formed on the surface of a thin steel sheet using a continuous hot-dip galvanizing device to form a hot-dip galvanized steel sheet, and then the hot-dip galvanized steel sheet is further alloyed to form an alloyed hot-dip galvanized steel sheet. In this example, Figure 3The continuous hot-dip metal coating equipment shown uses a reheating device located downstream of a gas purge device 70 located on the discharge side of the galvanizing tank and upstream of a defect detection device 80. The reheating device comprises an alloying belt, a heat-retaining belt, and a final cooling belt, with an induction heating device located in the alloying belt. The alloying belt is used to form a zinc coating having an alloy layer formed by a Zn-Fe alloying reaction on the surface of a thin steel sheet to produce alloyed hot-dip galvanized steel sheet. The galvanizing bath is a Zn bath containing Al. Cold-rolled raw coils with a thickness of 0.6 to 1.4 mm, a width of 690 to 1550 mm, and a weight of 12 to 16 tons were used as the steel strip. Twenty raw coils were prepared, and ten of them were converted into alloyed hot-dip galvanized steel sheets using the continuous hot-dip galvanizing equipment, with operational results obtained. The operating results were fed to the dross defect prediction model generation device to generate a dross defect prediction model. The remaining 10 raw material coils were annealed, plated, and alloyed in a continuous hot-dip galvanizing facility, and dross defect prediction was performed. The raw material coil composition (specification) was SGH440, and the target value for the unit weight per single side of the galvanized layer of the raw material coil was 40 g / m 2 .

[0209] Using Figure 8 The defect detection device with the structure shown defines defects with a size of 100 μm or larger as dross defects, and uses the number of dross defects per 50 meters of the steel strip's longitudinal length as the dross defect information as the output variable. Furthermore, the upper limit of the permissible dross defect count is set at 0.5 per 50 meters of the strip's longitudinal length. Furthermore, the output of the dross defect prediction model is based on statistical information, so the actual number of dross defects converted to the aforementioned length is output as the dross defect count, and the upper limit of the dross defect count is also determined by the actual number.

[0210] The dross defect prediction model uses input variables for the steel strip thickness, width, and strip feed rate, as well as the dew point and oxygen concentration in the nostril as second operating conditions related to the nostril. Furthermore, the zinc bath temperature and the Al concentration in the galvanizing bath are used as first operating conditions related to the galvanizing tank. The initial settings for the second parameters, the dew point and oxygen concentration in the nostril, are -30°C and 10 ppm, respectively. The initial settings for the first parameters, the zinc bath temperature and Al concentration in the galvanizing bath, are 460°C and 0.130%, respectively.

[0211] The second parameter relating to the furnace nose and the first parameter relating to the galvanizing bath change over time. In generating the dross defect prediction model, operations in which the second parameter relating to the furnace nose, i.e., the dew point and oxygen concentration in the furnace nose, and the first parameter relating to the galvanizing bath, i.e., the zinc bath temperature and Al concentration, were intentionally changed were performed, running data was collected every 50 m in the long side direction of the steel strip, and running data associated with dross defect information for 10 coils was accumulated in the running performance database. Then, as a machine learning algorithm, a neural network was used with two layers in the intermediate layer. Using a sigmoid function as an activation function, a dross defect prediction model was generated.

[0212] The dross defect prediction model thus generated was applied to the remaining 10 coils, and the predicted values of the dross defect information were compared with the dross defect information based on the defect detection device every 50 m in the long side direction of each coil. As a result, for 200 pieces of dross defect information, the proportion (accuracy) of accurately predicting the presence or absence of dross defects was as good as 94.9%.

[0213] <Example 2>

[0214] In the following example, the dross defect reduction method according to the present embodiment was applied to the case of manufacturing other hot-dip galvanized steel sheets. The continuous hot-dip galvanizing apparatus used in this example was the same as the continuous hot-dip metal plating apparatus shown in Example 1 described above. However, unlike Example 1, in this example, a reheating apparatus was not used, and a hot-dip galvanized steel sheet was manufactured without performing alloying treatment on the hot-dip galvanized steel sheet after the plating treatment. As a raw material coil for the plating treatment, a raw material coil after cold rolling having a sheet thickness of 0.6 to 1.4 mm, a width of 690 to 1550 mm, and a weight of 12 to 16 tons was used. Twenty raw material coils were prepared, 10 of which were made into hot-dip galvanized steel sheets by the continuous hot-dip galvanizing apparatus, and running performance was obtained. The running performance was supplied to the dross defect prediction model generation device to generate a dross defect prediction model. For the remaining 10 raw material coils, annealing and plating treatment were performed by the continuous hot-dip galvanizing apparatus, and prediction of dross defects was performed. The composition group (specification) of the raw material coil was SGH440, and the target value of the weight per unit area of each single side of the zinc plated layer of the raw material coil was 50 g / m 2 .

[0215] As in Example 1, defects having a size of 100 μm or more were taken as dross defects by the defect detection device, and the number of dross defects per 50 m in the long side direction of the steel strip was used as dross defect information that became an output variable. In addition, the upper limit value of the dross defects that could be allowed was set to 0.5 per 50 m in the long side direction of the steel strip.

[0216] As input variables of the dross defect prediction model, in addition to the plate thickness, the plate width, and the passing speed of the steel strip, the dew point in the nozzle and the oxygen concentration in the nozzle as the second operating conditions related to the nozzle were used. In addition, as the first operating conditions related to the galvanizing bath, the zinc bath temperature and the Al concentration of the galvanizing bath were used. Further, the initial set values of the second parameters, that is, the dew point in the nozzle and the oxygen concentration in the nozzle, were -25°C and 15 ppm, respectively, and the initial set values of the first parameters, that is, the zinc bath temperature and the Al concentration of the galvanizing bath, were 450°C and 0.220%, respectively.

[0217] In generating the dross defect prediction model, an operation of intentionally changing the second parameters related to the nozzle, that is, the dew point in the nozzle and the oxygen concentration in the nozzle, and the first parameters related to the galvanizing bath, that is, the zinc bath temperature and the Al concentration, was performed, the operating data were collected every 50 m in the longitudinal direction of the steel strip, and the operating data associated with the dross defect information of 10 coils were accumulated in the operating performance database. Then, as the machine learning algorithm, a neural network was used, and the intermediate layer was made to be two layers. Using a sigmoid function as an activation function, the dross defect prediction model was generated.

[0218] The dross defect prediction model thus generated was applied to the remaining 10 coils, and the predicted values of the dross defect information were compared with the dross defect information based on the defect detection device every 50 m in the longitudinal direction of each coil. As a result, for 200 pieces of dross defect information, the proportion (accuracy) of accurately predicting the presence or absence of the occurrence of the dross defect was as good as 92.9%.

[0219] The present disclosure has been described based on the drawings and the embodiments, but note that as long as it is a person skilled in the art, various modifications and changes can be easily made based on the present disclosure. Therefore, note that these modifications and changes are included in the scope of the present disclosure. For example, the functions and the like included in each structure or each step and the like can be rearranged in a logically non-contradictory manner, and a plurality of structures or steps and the like can be combined into one or divided. In addition, although the data appearing in the present specification has been described in natural language, more specifically, it can be specified by a simulation language, an instruction, a parameter, a machine language, and the like that can be recognized by a computer.

[0220] For example, the present disclosure can also be implemented as a program in which the processing contents of each function of the above-described dross defect prediction device and the dross defect prediction model generation device are described or a storage medium in which the program is recorded. It is understood that these are also included in the scope of the present disclosure.

[0221] In the present disclosure, the program can be recorded in advance on a recording medium that is readable by a computer. The recording medium that is readable by a computer includes a non-transitory computer-readable medium, such as a magnetic recording device, an optical disc, an optical magnetic recording medium, or a semiconductor memory. Circulation of the program is performed, for example, by selling, transferring, or lending a removable recording medium, such as a DVD (digital versatile disc) or a CD-ROM (compact disc read only memory) on which the program is recorded. In addition, circulation of the program can also be performed by pre-storing the program in a storage of a server and transmitting the program from the server to other computers. In addition, the program can also be provided as a program product.

[0222] For example, at least a part of the processing actions performed by the dross defect prediction device 84 can be performed by the machine learning device. Conversely, at least a part of the processing actions performed by the dross defect prediction model generation device 83 can also be performed by the dross defect prediction device 84.

[0223] In the above-described embodiments, it is assumed that the dross defect prediction device 84 and the dross defect prediction model generation device 83 are independent devices, but the present disclosure is not limited thereto. The processing associated with the dross defect prediction method based on the dross defect prediction device 84 and the generation method of the dross defect prediction model based on the dross defect prediction model generation device 83 can also be performed by a single device. At this time, the single device can also perform the setting of the operating conditions of the continuous hot dip galvanizing apparatus 100 as well.

[0224] Explanation of Reference Signs

[0225] S…steel strip; G…hot-dip galvanized steel sheet; X…dross defect; L…inspection line; 100…continuous hot-dip galvanizing apparatus; 1…galvanizing bath; 2…furnace nose; 3…guide roll; 4…backup roll; 5…top dross; 6…bottom dross; 7…material feeding device; 8…material; 9…bath thermometer; 10…bath analysis device; 11…heating device; 20…heating zone; 21…soaking zone; 22…1st cooling zone (rapid cooling zone); 23…2nd cooling zone (slow cooling zone); 30…continuous annealing furnace; 40…gas piping; 41A, 41B, 41C, 42A, 42B, 42C…gas supply port; 50…gas supply section; 51A, 51B, 51C, 51D, 51E…piping; 52…valve; 53A, 53B…dew point measurement hole; 54…radiation thermometer; 55…heater; 56…heat insulating material; 57…thermocouple; 58…temperature control section; 59…diffusion pipe; 60…valve; 70…gas purging apparatus; 71…purging nozzle; 72…header; 73…pressure gauge; 74…thermometer; 75…flexible hose; 76…gas heating device; 77…air compressor; 80…defect detection device; 81…light projector; 82…high-performance camera; 83…dross defect prediction model generating device; 831…acquisition section; 832…storage section; 833…output section; 834…machine learning section; 84…dross defect prediction device; 841…acquisition section; 842…output section; 843…storage section; 844…control section; 85…operation performance database; 90…operation condition setting device; 95…host computer; 901…determination section; 902…operation condition re-setting section; 200…dross defect prediction system.

Claims

1. A dross defect prediction method of a steel strip detected on a downstream side of a galvanizing bath in a continuous hot dip galvanizing apparatus including an annealing furnace, the galvanizing bath formed with a galvanizing bath, and a furnace snout provided on a delivery side of the annealing furnace and configured so that a front end portion is immersed in the galvanizing bath, the dross defect prediction method characterized by comprising: a step of receiving input of an operation condition including a first operation condition related to the galvanizing bath and a second operation condition related to the furnace snout with respect to a dross defect prediction model that is a model of machine learning performed based on operation data including a first parameter related to the galvanizing bath and a second parameter related to the furnace snout as input variables and dross defect information of the steel strip as an output variable; and a step of calculating a predicted value of the dross defect information of the steel strip based on the operation condition that has been input and by the dross defect prediction model, the first parameter and the first operation condition including one or two parameters selected from a temperature of the galvanizing bath and an Al concentration in the galvanizing bath, the second parameter and the second operation condition including one or more parameters selected from a dew point in the furnace snout, a hydrogen concentration in the furnace snout, an oxygen concentration in the furnace snout, a temperature of the steel strip in the furnace snout, and an atmosphere temperature in the furnace snout.

2. The dross defect prediction method according to claim 1, characterized in that the continuous hot dip galvanizing apparatus has a gas purging apparatus on a delivery side of the galvanizing bath, the dross defect prediction model is a model of the machine learning performed based on further including a third parameter related to the gas purging apparatus as the input variables, and the operation condition further includes a third operation condition related to the gas purging apparatus. comprising: a step of calculating a predicted value of the dross defect information of the steel strip on a downstream side of the galvanizing bath using the dross defect prediction method according to claim 1 or 2; and a step of resetting the operation condition based on the predicted value of the dross defect information.

4. A method of manufacturing a hot dip galvanized steel sheet, wherein the continuous hot dip galvanizing apparatus whose operation condition is controlled using the dross defect reduction method according to claim 3 is used to form a galvanizing layer on a surface of the steel strip to form a hot dip galvanized steel sheet.

5. A method of manufacturing an alloyed hot dip galvanized steel sheet, the continuous hot dip galvanizing apparatus whose operation condition is controlled using the dross defect reduction method according to claim 3 is characterized in that the continuous hot dip galvanizing apparatus further has a reheating apparatus at a position on a downstream side of the galvanizing bath and upstream of a defect detection device, a galvanizing layer is formed on a surface of the steel strip in the galvanizing bath to form a hot dip galvanized steel sheet, and the hot dip galvanized steel sheet is further subjected to an alloying treatment by the reheating apparatus to form an alloyed hot dip galvanized steel sheet. ​ ​ ​ ​ ​ 3. A method of reducing dross defects in a steel strip, characterized by, ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ 6. A method of generating a dross defect prediction model for predicting a dross defect of a steel strip detected on a downstream side of a galvanizing bath in a continuous hot dip galvanizing apparatus including an annealing furnace, the galvanizing bath, and a furnace nose, the annealing furnace annealing the steel strip, the galvanizing bath forming a galvanizing bath, the furnace nose being provided on a delivery side of the annealing furnace and configured to dip an end portion in the galvanizing bath, the method of generating the dross defect prediction model characterized by comprising: a step of acquiring operation data including a first parameter related to the galvanizing bath and a second parameter related to the furnace nose; and a step of performing machine learning using the acquired operation data as input variables and dross defect information of the steel strip detected on the downstream side of the galvanizing bath as output variables, the first parameter including one or two parameters selected from a temperature of the galvanizing bath and an Al concentration in the galvanizing bath, the second parameter including one or more parameters selected from a dew point in the furnace nose, a hydrogen concentration in the furnace nose, an oxygen concentration in the furnace nose, a temperature of the steel strip in the furnace nose, and an atmosphere temperature in the furnace nose.

7. The method of generating the dross defect prediction model according to claim 6, wherein in the step of performing the machine learning, one or more machine learning algorithms selected from a neural network, a decision tree learning, a random forest, a support vector regression, a Gaussian process, and a k-nearest neighbor method are used.

8. A dross defect prediction device for predicting a dross defect of a steel strip detected on a downstream side of a galvanizing bath in a continuous hot dip galvanizing apparatus including an annealing furnace, the galvanizing bath, and a furnace nose, the galvanizing bath forming a galvanizing bath, the furnace nose being provided on a delivery side of the annealing furnace and configured to dip a front end portion in the galvanizing bath, the dross defect prediction device characterized by comprising: an acquisition section that acquires operation conditions including a first operation condition related to the galvanizing bath and a second operation condition related to the furnace nose; and a control section that calculates a predicted value of dross defect information of the steel strip with respect to a dross defect prediction model input with the operation conditions, the dross defect prediction model being a model based on machine learning performed using operation data including a first parameter related to the galvanizing bath and a second parameter related to the furnace nose as input variables and the dross defect information of the steel strip as output variables, the first parameter and the first operation condition including one or two parameters selected from a temperature of the galvanizing bath and an Al concentration in the galvanizing bath, the second parameter and the second operation condition including one or more parameters selected from a dew point in the furnace nose, a hydrogen concentration in the furnace nose, an oxygen concentration in the furnace nose, a temperature of the steel strip in the furnace nose, and an atmosphere temperature in the furnace nose.

9. A dross defect prediction terminal system characterized by comprising: comprising: the dross defect prediction device according to claim 8; and the dross defect prediction device according to claim 8. A terminal device receives a user input related to the change of the operation condition and transmits user input information based on the user input to the dross defect prediction device, The dross defect prediction device includes a control section that changes at least a part of the operation condition based on the user input information as a changed operation condition, calculates a predicted value of the dross defect information of the steel strip based on the changed operation condition and by the dross defect prediction model.

Images