Apparatus for estimating state in mold, method for estimating state in mold, and program

Abstract

To provide a device for estimating a state in a mold which is hardly affected by noise and can detect abnormality during casting with high accuracy.SOLUTION: The in-mold state estimation device (1) includes a plurality of temperature measurement points in the casting direction in the copper plate of the mold, and includes a position-fixed temperature component whose position is fixed in the casting direction and a moving temperature component moving along the casting direction, and includes a position-fixed temperature component prediction unit (12), a moving temperature component prediction unit (13), a temperature prediction unit (14) which sets a sum of the predicted values of the position-fixed temperature component and the moving temperature component as a predicted value of a temperature measurement value, and a temperature component optimization unit (15) which calculates an optimal predicted value of each temperature component from an estimation error of the temperature measurement value and the predicted values of the position-fixed temperature component and the moving temperature component.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to an in-mold condition estimation device, an in-mold condition estimation method, and a program. [Background technology]

[0002] In continuous casting of steel materials, molten steel solidifies where it comes into contact with the inner surface of the mold, forming a solidified shell as the casting progresses. If the thickness of the solidified shell deviates from the standard state due to, for example, the inclusion of foreign matter, it can cause serious defects such as breakouts. Therefore, technologies have been proposed that measure the temperature inside the mold during casting to detect abnormalities early.

[0003] For example, Patent Document 1 discloses a method in which thermocouples are embedded in two predetermined locations in a continuous casting mold, and the measured temperature values ​​are used to determine whether a breakout has occurred. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2012-218039 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the temperature readings of the thermocouples may contain noise specific to the location where the thermocouples are buried. The technology described in Patent Document 1 focuses on the amount of temperature change in the temperature readings of each of the two thermocouples, but if the noise acts to bring the temperature readings closer to the average value, the amount of temperature change will be underestimated. In such cases, there is a problem that the intended breakout detection, etc. cannot be performed.

[0006] One aspect of the present invention aims to provide an in-mold condition estimation device, an in-mold condition estimation method, and a program that are less susceptible to noise and capable of detecting abnormalities during casting with high accuracy. [Means for solving the problem]

[0007] In order to solve the above-mentioned problems, an in-mold condition estimating device according to one aspect of the present invention is an in-mold condition estimating device that estimates a condition inside the mold based on temperature measurements obtained at a plurality of temperature measurement points set along the casting direction of a mold of a continuous casting machine, wherein the estimated value of the temperature measurement value is a value obtained by correcting a predicted value of the temperature measurement value using a variable, and when an optimal estimate of the temperature measurement value, which is the estimated value of the temperature measurement value that minimizes an estimation error between the actual value of the temperature measurement value and the estimated value of the temperature measurement value, is expressed as the sum of an optimal estimate of a position-fixed temperature component, which is a component whose value does not move with time relative to the mold, and an optimal estimate of a moving temperature component, which is a component whose value moves with time relative to the mold, the device uses a first model to calculate a predicted value of the position-fixed temperature component of the temperature measurement value after a predetermined time from the previous estimation of the optimal estimate, for each of the temperature measurement points, based on the optimal estimate of the position-fixed temperature component in a previous estimation. a temperature prediction unit that calculates, for each of the temperature measurement points, a predicted value of the moving temperature component of the temperature measurement value a predetermined time after the previous estimation based on an optimal estimated value of the moving temperature component in the previous estimation using a second model; a temperature prediction unit that calculates, for each of the temperature measurement points, the sum of the predicted value of the position-fixed temperature component a predetermined time after the previous estimation and the predicted value of the moving temperature component a predetermined time after the previous estimation as a predicted value of the temperature measurement value a predetermined time after the previous estimation; and a temperature component optimization unit that estimates, for each of the temperature measurement points, the estimated values ​​of the temperature measurement value for each of the position-fixed temperature component and the moving temperature component, which are obtained by minimizing the estimation error between the actual value of the current temperature measurement value and the estimated value of the temperature measurement value, to be the optimal estimated value of the position-fixed temperature component and the optimal estimated value of the moving temperature component, respectively, based on the predicted value of the temperature measurement value a predetermined time after the previous estimation.

[0008] In order to solve the above-mentioned problems, one aspect of the present invention provides a method for estimating a state inside a mold, which estimates a state inside the mold based on temperature measurements obtained at a plurality of temperature measurement points set along the casting direction of a mold of a continuous casting machine, and includes the steps of: correcting a predicted value of the temperature measurement value using a variable to set the estimated value of the temperature measurement value; and, when an optimal estimate of the temperature measurement value, which is the estimated value of the temperature measurement value that minimizes an estimation error between the actual value of the temperature measurement value and the estimated value of the temperature measurement value, is expressed as the sum of an optimal estimate of a position-fixed temperature component, the value of which is a component whose value does not move with respect to the mold over time, and an optimal estimate of a moving temperature component, the value of which is a component whose value moves with respect to the mold over time, using a first model for each of the temperature measurement points, calculating a predicted value of the position-fixed temperature component of the temperature measurement value a predetermined time after the previous estimation of the optimal estimate, based on the optimal estimate of the position-fixed temperature component in a previous estimation. a moving temperature component prediction step for calculating, for each of the temperature measurement points, a predicted value of the moving temperature component of the temperature measurement value a predetermined time after the previous estimation based on an optimal estimated value of the moving temperature component in the previous estimation using a second model; a temperature prediction step for calculating, for each of the temperature measurement points, the sum of the predicted value of the position-fixed temperature component a predetermined time after the previous estimation and the predicted value of the moving temperature component a predetermined time after the previous estimation as a predicted value of the temperature measurement value a predetermined time after the previous estimation; and a temperature component optimization step for estimating, for each of the temperature measurement points, the estimated values ​​of the temperature measurement value for each of the position-fixed temperature component and the moving temperature component, which are obtained by minimizing an estimation error between the actual value of the current temperature measurement value and the estimated value of the temperature measurement value, to be the optimal estimated value of the position-fixed temperature component and the optimal estimated value of the moving temperature component, respectively, based on the predicted value of the temperature measurement value a predetermined time after the previous estimation. [Effects of the Invention]

[0009] According to one aspect of the present invention, it is possible to realize an in-mold condition estimation device, an in-mold condition estimation method, and a program that are less susceptible to noise and capable of detecting abnormalities during casting with high accuracy. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 2 is a schematic diagram showing the state inside a mold during casting by a continuous casting machine. [Figure 2] FIG. 1 is a graph illustrating a relationship between the progress of casting time, the shape of the solidified shell, the surface temperature of the mold copper plate, and the temperature components of the surface temperature. [Figure 3] 1 is a block diagram showing a schematic configuration of an in-mold state estimation device according to an embodiment of the present invention. [Figure 4] FIG. 2 is a perspective view of a mold, and is a schematic diagram showing an example of how temperature measuring points are installed on a mold copper plate. [Figure 5] FIG. 10 is a diagram showing the positions of temperature measurement points on the p-axis set on the mold copper plate. [Figure 6] FIG. 10 is a diagram showing the positions of temperature calculation points on the z-axis set on the slab. [Figure 7] FIG. 1 is a flow chart showing an example of a method for estimating a state inside a mold according to an embodiment of the present invention. [Figure 8] FIG. 1 is a graph plotting measured temperatures caused by foreign matter getting caught on the surface of a cast slab in one embodiment. [Figure 9] FIG. 10 is a graph plotting changes in temperature values ​​at each temperature measurement point over a predetermined time period in one embodiment. [Figure 10] FIG. 10 is a graph plotting actual values ​​and optimal estimated values ​​at each temperature measurement point and each temperature calculation point at time t=6 in one example. [Figure 11] FIG. 10 is a graph plotting actual values ​​and optimal estimated values ​​at each temperature measurement point and each temperature calculation point at time t=9 in one example. [Figure 12] FIG. 10 is a graph plotting actual values ​​and optimal estimated values ​​at each temperature measurement point and each temperature calculation point at time t=12 in one embodiment. [Figure 13] FIG. 10 is a graph plotting actual values ​​and optimal estimated values ​​at each temperature measurement point and each temperature calculation point at time t=25 in one embodiment. [Figure 14]FIG. 10 is a graph showing the average root mean square deviation between the actual value and the optimal estimated value of the moving temperature component and the number of temperature measurement points passed by the temperature drop portion in one embodiment. [Figure 15] FIG. 10 is a graph plotting the optimal estimated values ​​of each temperature component and the actual values ​​of the position-fixed temperature components at each temperature measurement point against time in one embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, one embodiment of the present invention will be described in detail.

[0012] [Summary of the Invention] An in-mold condition estimation device according to one embodiment of the present invention estimates the casting condition inside a mold with high accuracy based on temperature measurements at temperature measurement points provided on a mold copper plate of a continuous steel casting machine.

[0013] FIG. 1 is a schematic diagram showing the state inside a mold M during casting by a continuous casting machine; the upper diagram is a top view of the mold M, and the lower diagram is a cross-sectional view taken along line A-A' in the upper diagram, viewed from the side. During casting by a continuous casting machine, molten steel L1 is discharged into the mold M from a submerged entry nozzle M2 inserted into the mold M. The molten steel L1 discharged into the mold M spreads as a discharge flow in the direction indicated by the black arrow in the lower diagram of FIG. 1, filling the mold M. A powder film F, which is a layer of mold powder, is formed in the mold M at the position that will become the surface of the molten steel L1.

[0014] The mold copper plate M1 that forms the outer wall of the mold M is water-cooled from the outer surface of the mold M, and the molten steel L1 filled into the mold M solidifies from the position where it comes into contact with the mold copper plate M1, forming a solidified shell L2. As casting progresses, the solidified shell L2 moves vertically downward, which is the casting direction, while increasing in thickness.

[0015] In casting using such a continuous casting machine, for example, the inflow and lubrication of the mold powder may be impaired due to the precipitation of high-melting-point crystals in the powder film F. In such cases, it is known that the temperature of the mold M fluctuates greatly, and the temperature difference between the opposing surfaces of the mold M becomes large.

[0016] In a conventional method for determining the occurrence of a breakout, such as that disclosed in Patent Document 1, the occurrence of a breakout is determined based on the temperature fluctuation of the mold copper plate M1 or the temperature deviation between the opposing surfaces of the mold copper plate M1. With such a conventional method, there is a risk that a breakout may be erroneously determined to have occurred, depending on, for example, the combination of the casting steel type and the mold powder composition. Such an erroneous determination hinders casting.

[0017] It is also known that the thickness of the solidified shell L2 in the mold M of a continuous casting machine typically increases in proportion to the 1 / 2 power of the casting time from the start of solidification. However, there are several reasons why the thickness may deviate from the standard thickness. The main cause of this is the occurrence of a phenomenon in which the state of the heat flux passing from the molten steel L1 through the solidified shell L2 in its thickness direction deviates from the standard state (steady state). Such phenomena can be divided into two types: one in which the unsteady state of the heat flux deviating from the standard state moves along with the solidified shell L2, and one in which the position of the heat flux in the casting direction within the mold M is fixed.

[0018] A typical example of the phenomenon in which the position of the unsteady state of heat flux moves along with the solidified shell L2 is a phenomenon caused by a foreign object trapped between the solidified shell L2 and the mold copper plate M1. Specifically, an air gap is created around the foreign object trapped in the solidified shell L2, reducing heat transfer from the high-temperature molten steel L1 to the mold copper plate M1, which is relatively cooler than the molten steel L1. This creates a region in the solidified shell L2 that is cooler than the surrounding area near the foreign object. The position of this low-temperature region, i.e., the position of the unsteady state of heat flux, moves downward as the foreign object moves with the progress of casting. An example of a foreign object is slag pellets, which are fragments of sintered mold powder.

[0019] On the other hand, a typical example of a phenomenon in which the position of the unsteady heat flux does not move with the solidified shell L2 but remains fixed in the casting direction within the mold M is when the direction of the discharge flow from the SEN M2 deviates from its intended direction due to clogging inside the SEN M2, resulting in collision with the solidified shell L2 at an unusual location. In other words, due to this collision, the heat of the high-temperature molten steel L1 is supplied to the solidified shell L2 at the location directly hit by the discharge flow from the solidified shell L2 in excess of that in a normal state, increasing heat transfer from the solidified shell L2 to the mold copper plate M1. This results in a region of higher temperature in the solidified shell L2 near the collision location compared to the surrounding area. The position of the unsteady heat flux, i.e., the collision location, is roughly determined by the position of the mold copper plate M1 and the position, clogging level, and orientation of the SEN M2, and therefore remains fixed in approximately the same position regardless of the progress of casting.

[0020] Another example of a phenomenon in which the position in the casting direction is fixed in an unsteady state of heat flux is a decrease in the thickness of the powder flux between the mold copper plate M1 and the solidified shell L2 or a break in the film, etc. This reduces the buffering effect of the mold powder and increases the heat transfer from the solidified shell L2 to the mold copper plate M1, resulting in the mold copper plate M1 always being fixed in approximately the same position regardless of the progress of casting.

[0021] (Definition of temperature components) FIG. 2 is a graph showing the relationship between the progression of casting time, the shape of the solidified shell L2, the surface temperature of the mold copper plate M1, and the temperature components of the surface temperature. In FIG. 2, "foreign matter such as slag bearing" is an example of a phenomenon that moves with the solidified shell L2, and "molten steel flow collision" is an example of a phenomenon whose position in the casting direction is fixed. Note that temperature measurements at each temperature measurement point are obtained by installing multiple thermometers at each of multiple temperature measurement points set in the mold M (especially inside the mold copper plate M1). When the term "temperature component" is used for the temperature measurement value at each temperature measurement point or the surface temperature of the mold copper plate, it means that the temperature measurement value at each temperature measurement point or the surface temperature of the mold copper plate corresponds to the sum of the temperature values ​​indicated as the temperature component. Furthermore, the temperature component of the temperature measurement value at each temperature measurement point, which occurs due to the phenomenon that the position of the unsteady heat flux state moves with the solidified shell L2, is referred to as the "moving temperature component." Furthermore, among the temperature values ​​measured at each temperature measurement point, the sum of the temperature component corresponding to the temperature of the mold copper plate in a steady state and the temperature component resulting from the phenomenon in which the position in an unsteady state does not move along with the solidified shell L2 but is fixed at approximately the same position is referred to as the "position-fixed temperature component."

[0022] (Surface temperature of the mold copper plate on the solidified shell side) Figure 2 shows the state at time t and the state at time t + Δt, a predetermined time after time t. Figure 2(1) schematically illustrates the shape of the solidified shell L2 affected by the heat flux due to the impingement of the molten steel flow, as well as the shape of a foreign object, such as a slag ball, sandwiched between the solidified shell L2 and the mold copper plate M1 and the surrounding air gap. As shown in Figure 2(2), the trapped foreign object moves along with the solidified shell L2, and the resulting decrease in the surface temperature of the mold copper plate M1 on the solidified shell side also moves in the casting direction over time (from time t to time t + Δt). This surface temperature decrease, which moves in the casting direction over time, can be expressed as a temperature component (moving temperature component) that moves along with the solidified shell L2, as shown by the black line in Figure 2(3). On the other hand, as shown in Figure 2(2), the temperature increase due to the impingement of the molten steel flow does not change its position in the casting direction over time. This can be expressed as a temperature component whose position in the casting direction is fixed (position-fixed temperature component), as shown by the gray line in Figure 2 (3).

[0023] At the surface of the mold copper plate M1, where the heat transfer from the solidified shell L2 increases or decreases, the surface temperature of the mold copper plate M1 rises or falls compared to the normal temperature distribution. If a thermocouple or FBG (Fiber Bragg Grating) thermometer or the like is embedded in the mold copper plate M1 to measure the temperature, the measured temperature will also rise or fall, reflecting the temperature of the surface of the mold copper plate M1.

[0024] (Temperature measurement values ​​at each temperature measurement point on the mold copper plate) In FIG. 2(2), the solid lines indicate the measured temperature values ​​at the temperature measurement points in the mold copper plate M1, and the broken lines indicate the normal temperature distribution expected from the casting conditions, etc.

[0025] In the temperature measurement values ​​in Figure 2 (3), the black lines indicate the temperature components at each temperature measurement point that are caused by foreign matter whose position relative to the mold M moves in the casting direction together with the solidified shell L2, and the gray lines indicate the temperature components that are not caused by foreign matter, whose position relative to the mold M is fixed in the casting direction.

[0026] Each measured temperature value at each temperature measurement point on the mold copper plate M1 can be expressed as the sum of a temperature component that moves with the solidified shell L2 and a temperature component whose position in the casting direction is fixed.

[0027] The area where the temperature rises from the standard state at the location where the solidified shell L2 is remelted by the impingement of the molten steel flow may move in the casting direction as the slab is withdrawn from the mold M. In other words, the temperature change of the mold copper plate M1 caused by the impingement of the molten steel flow is usually expressed as a temperature component whose position in the casting direction is fixed, but it may also be expressed as a temperature component that moves together with the solidified shell L2.

[0028] Although methods for estimating the in-mold casting condition using the temperature deviation between the normal temperature distribution and the measured temperature have been attempted in the past, the method of evaluating the measured temperature by separating it into two components as described above has not been known. For example, as in Patent Document 1, the only known method is to detect an abnormality such as foreign matter getting caught in the solidified shell L2 by using a correlation between measured temperatures at two points that are vertically related in the casting direction, taking into account the time difference due to movement at the casting speed.

[0029] However, when detecting foreign matter or breakage in the solidified shell L2 using the product of the temperature change amounts of two thermocouples aligned in the casting direction, as in Patent Document 1, noise specific to the temperature measurement point may affect the measured temperature. If this noise acts to bring the measured temperature value closer to the average value, the temperature change amount may be underestimated. In such a case, there is a risk that the intended detection of a breakout or the like may not be achieved.

[0030] This problem could be solved by increasing the number of temperature measurement points installed in the casting direction and statistically capturing the characteristics of the product of the temperature change. However, increasing the number of temperature measurement points increases the number of calculations for the product in proportion to the square of the number, and therefore there is a limit to the number of temperature measurement points that can be added due to the processing load.

[0031] To address this issue, an apparatus for estimating the state inside the mold according to one embodiment of the present invention separates the measured temperature into a position-fixed temperature component and a moving temperature component. The estimated values ​​of the position-fixed temperature component and the moving temperature component for each temperature measurement point are then used to estimate the state inside the mold. The position-fixed temperature component is a component of the measured temperature that is fixed in position relative to the mold copper plate M1, and the moving temperature component is a component that moves along with the solidified shell L2.

[0032] In one embodiment of the present invention, by evaluating both the fixed-position temperature component and the moving temperature component, the evaluation results are less susceptible to noise affecting the measured temperature values. Therefore, the condition inside the mold can be estimated with high accuracy. Furthermore, in one embodiment of the present invention, increasing the number of temperature measurement points does not significantly increase the processing load, while improving estimation accuracy. Therefore, by comprehensively setting temperature measurement points over the entire mold copper plate M1, it is possible to comprehensively detect the occurrence of abnormalities regardless of the position of the mold M.

[0033] This configuration reduces the risk of overlooking abnormalities such as breakouts, allowing for efficient steel production without waste. This significantly improves energy efficiency in steel production. This effect will also contribute to achieving Goal 12 of the United Nations' Sustainable Development Goals (SDGs), "Ensure sustainable consumption and production patterns."

[0034] [Device for estimating state inside the mold] An estimation device 1 for the state inside a mold (estimation device 1 for the state inside a mold, estimation device 1) according to one embodiment of the present invention is a device that estimates the state inside a mold M based on temperature measurements obtained at a plurality of temperature measurement points 51 set along the casting direction on a mold M of a continuous casting machine. FIG. 3 is a block diagram showing a schematic configuration of the estimation device 1. As shown in FIG. 3, the estimation device 1 includes a control unit 10, a memory unit 20, an input unit 30, and an output unit 40.

[0035] The storage unit 20 is configured to store information processed by the estimation device 1, and may be, for example, an HDD (Hard Disk Drive) or an SSD (Solid State Drive). The storage unit 20 stores information transmitted from the control unit 10, and the stored information may be read by the control unit 10. Examples of such information include casting condition information 21, model information 22, and temperature measurement value information 24, which will be described later.

[0036] The input unit 30 is configured to accept information input to the estimation apparatus 1. The input unit 30 may be, for example, an input device such as a keyboard, a mouse, or a touchpad. The input unit 30 may also be a communication device or an interface that accepts information input from an external device. The estimation apparatus 1 and such an external device may be connected by wire or wirelessly. Examples of external devices include a copper plate temperature measuring device 50 and a casting speed measuring device 60 shown in FIG. 3.

[0037] The copper plate temperature measuring device 50 is provided at each of a plurality of temperature measuring points 51 set at a plurality of positions inside (the inner surface of) the mold copper plate M1, and is a device for measuring the temperature at the temperature measuring points 51, and further outputs the measurement results obtained by the measurement to the estimation device 1. The copper plate temperature measuring device 50 may be any device that can measure the temperature of the mold copper plate M1 at each temperature measuring point 51 in the high-temperature environment inside the mold copper plate M1, and for example, a thermocouple or an FBG (Fiber Bragg Grating) thermometer can be used.

[0038] From the viewpoint of accurately estimating the state inside the mold, the copper plate temperature measuring device 50 is preferably provided in a position close to the inner surface of the mold copper plate M1. Furthermore, if a cooling water slit is provided in the mold copper plate M1, the copper plate temperature measuring device 50 is preferably provided in a position away from the cooling water slit so as to be less affected by the temperature of the cooling water slit.

[0039] FIG. 4 is a perspective view of the mold M and a schematic diagram showing an example of the placement of temperature measurement points 51 on the mold copper plate M1. FIG. 5 is a diagram showing the positions of the temperature measurement points 51 on the p-axis, which is a virtual one-dimensional coordinate axis set on the mold copper plate M1. In FIG. 5, the upper diagram is a top view of the mold copper plate M1, and the lower diagram is an enlarged view of a portion of the cross-sectional view taken along line A-A' in FIG. 1 that corresponds to the top view. The p-axis and z-axis shown in FIG. 5 represent virtual coordinate axes.

[0040] As shown in Fig. 4, a plurality of temperature measurement points 51 are provided at predetermined intervals on the mold copper plate M1 along the casting direction of the mold M. The predetermined intervals are preferably constant. In a row R of a plurality of temperature measurement points 51 lined up along the casting direction, the uppermost temperature measurement point 51 is preferably located at the same position as the molten steel surface L3, which is the surface of the molten steel L1, or at a position below the molten steel surface L3 and close to the molten steel surface L3.

[0041] 4, it is preferable that the row R of the temperature measurement points 51 arranged along the casting direction be provided in multiple rows in the circumferential direction of the mold M. With this configuration, the range in which the temperature of the mold copper plate M1 can be measured by the copper plate temperature measuring device 50 provided at the temperature measurement point 51 can be expanded, thereby improving the accuracy of detecting casting abnormalities.

[0042] As shown in FIG. 5, the estimation device 1 sets a p-axis (first coordinate axis; an axis extending in the casting direction on the mold M) which is a coordinate axis extending in the casting direction on the mold copper plate M1 along the row R on which the temperature measurement points 51 are provided. The p-axis may be an imaginary linear axis connecting the plurality of temperature measurement points 51. Regarding the position on the p-axis, the target position of the molten steel surface L3 level at the start of casting is set to p0. Furthermore, when the number of temperature measurement points 51 on the p-axis is set to N, the temperature measurement points 51 on the p-axis are assigned p1, p2, ..., p3 in order of closest to p0. N-1 , p N Let's say.

[0043] Furthermore, in addition to the p-axis set for the mold M, the estimation device 1 sets a z-axis (second coordinate axis; an axis extending in the casting direction on the slab produced in the mold M), which is a virtual one-dimensional coordinate axis that is set not on the mold copper plate M1 but on the slab in the mold M. The z-axis is preferably parallel to the p-axis and set at a position on the slab close to the mold copper plate M1.

[0044] The z-axis extends in the opposite direction to the casting direction on the slab, with z0 being the position on the slab that corresponds to the target position of the molten steel surface level L3 at the start of casting. In other words, the length of the z-axis is equal to the casting length, which is the value obtained by integrating the casting speed with respect to time from the start of casting at the target position of the molten steel surface level L3. The relationship between the p-axis and the z-axis satisfies the following equation (1). p=z0-z (1)

[0045] When a plurality of rows R of temperature measurement points 51 are provided on the mold copper plate M1, the estimation device 1 may set the p-axis and z-axis for each row R and perform estimation processing for each row R.

[0046] The casting speed measuring device 60 is a device that measures the casting speed of the continuous casting machine and may be a device that outputs the measurement results of the casting speed to the estimation device 1. The casting speed measuring device 60 may be, for example, a component such as an encoder that is located below the mold M and rotates in contact with the surface of the slab delivered from the mold M. The casting speed measuring device 60 may calculate the casting speed from the speed of this rotation.

[0047] The output unit 40 is configured to output information from the estimation device 1. The output unit 40 may be, for example, a display device such as a liquid crystal display. Alternatively, the output unit 40 may be a speaker that outputs a notification by sound or the like when the estimation device 1 outputs a notification. The notification output by the estimation device 1 may be, for example, a warning indicating that an abnormality related to the occurrence of a breakout or the like in casting has been detected. The estimation device 1 may be equipped with both a display unit and a speaker as the output unit 40.

[0048] The control unit 10 is a control device that controls each unit of the estimation device 1. The control unit 10 may be, for example, a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). The control unit 10 includes a position-fixed temperature component prediction unit 12, a moving temperature component prediction unit 13, a temperature prediction unit 14, and a temperature component optimization unit 15. The control unit 10 may also include a casting state abnormality estimation unit 16.

[0049] The control unit 10 may read a control program, which is software stored in the storage unit 20, and develop it in a memory such as a RAM (Random Access Memory), and execute the functions of each unit.

[0050] The predetermined time after the start of casting can be represented as Δt. Furthermore, a discrete time that increases by 1 at Δt second intervals can be represented as time t. If the time point estimated by the estimation device 1 is time t, the time a predetermined time before time t is time t-1, and the time a predetermined time after time t is time t+1.

[0051] (Temperature measurement model) Next, the position-fixed temperature component prediction unit 12 and the moving temperature component prediction unit 13 will be described. First, an example of a model that defines each temperature component used by the position-fixed temperature component prediction unit 12 and the moving temperature component prediction unit 13 will be described. Each model described here may be stored in the storage unit 20 as part of the model information 22. This also applies to the models described in the following sections for each functional block.

[0052] In this specification, the actual temperature value measured at the temperature measurement point 51 may be referred to as the actual temperature value. Also, the value obtained by correcting the predicted temperature value at the temperature measurement point 51 using a variable will be referred to as the estimated temperature value.

[0053] The position-fixed temperature component is a component of the temperature measurement value at each temperature measurement point 51 whose value does not move over time relative to the mold M, and whose position is fixed with respect to the p-axis regardless of the progress of casting. On the other hand, the moving temperature component is a component of the temperature measurement value at each temperature measurement point 51 whose value moves over time relative to the mold M, and which moves along the z-axis together with the solidified shell L2 of the slab as casting progresses. In this way, the temperature measurement value at the temperature measurement point 51 is expressed as a value including a position-fixed temperature component and a moving temperature component.

[0054] Here, the measured temperature is represented as T, the fixed-position temperature component as T1, and the moving temperature component as T2. Hereinafter, in this specification, when a symbol such as T is followed by "1," it indicates a component related to a fixed-position temperature, and when a symbol such as T is followed by "2," it indicates a component related to a moving temperature.

[0055] The temperature measurement value T of the nth temperature measurement point 51 at time t t n is assumed to be expressed by the following equation (2): In the following equation (2), ν represents measurement noise due to the temperature measurement point 51. T t n =T1 t n +T2 t n +ν t n (2)

[0056] It is also assumed that the temperature measured at each temperature measurement point 51 changes every predetermined time. At this time, it is assumed that a disturbance is added to the position-fixed temperature component every predetermined time, and that the temperature changes sequentially as shown in the following equation (3). The disturbance added to the position-fixed temperature component every predetermined time is represented as ΔT1.

[0057] T1 t+1 n =T1 t n +ΔT1 t n (3) The moving temperature component is V on the p axis at each specified time.t The moving temperature component T2(p n -V t-1 Δt), the disturbance ΔT2(p n -V t-1 Δt) is added and is assumed to change sequentially as shown in the following equation (4). Here, V t represents the casting speed at time t, and V t Δt represents the distance traveled over a given time period at the casting speed at time t. ΔT2 represents the disturbance applied to the travel temperature component over a given time period. n indicates the p-axis coordinate of the position of the n-th temperature measurement point 51 on the p-axis of the mold copper plate M1. T2 t n =T2(p n -V t-1 Δt)+ΔT2(p n -V t-1 Δt) (4)

[0058] Here, the actual value of the temperature measured at the temperature measuring point 51 is represented as Y. In this case, the actual value Y of the temperature measured at the temperature measuring point 51 at time t t n In the formula (2), Y t n =T t n Then, the actual temperature measurement value Y t n can be expressed using a position-fixed temperature component and a moving temperature component as shown in the following equation (5). Y t n =T1 t n +T2 t n +ν t n (5)

[0059] (Position-fixed temperature component prediction unit 12) The position-fixed temperature component prediction unit 12 is a functional unit that uses a first model to calculate, for each temperature measurement point 51, a predicted value of the position-fixed temperature component at the temperature measurement point a predetermined time after the previous estimation of the optimal estimated value, based on the optimal estimated value of the position-fixed temperature component in the previous estimation. Here, the predicted value of the position-fixed temperature component estimated by the position-fixed temperature component prediction unit 12 is represented as x1. Furthermore, the optimal estimated value of the position-fixed temperature component, which is the value obtained by optimizing the predicted value by the temperature component optimization unit 15 described below, is represented as X1. Furthermore, the disturbance of the position-fixed temperature component optimized by the temperature component optimization unit 15 is represented as ΔX1.

[0060] The position-fixed temperature component prediction unit 12 calculates a predicted value x1 of the position-fixed temperature component at the n-th temperature measurement point 51 at time t, a predetermined time after the state estimation at time t−1. t n is the optimal estimate of the position-fixed temperature component at time t-1, X1 t-1 n Specifically, the position-fixed temperature component prediction unit 12 calculates the optimum estimated value X1 of the position-fixed temperature component as shown in the following equation (6). t-1 n The optimal estimate of the disturbance ΔX1 t-1 n The value added is the predicted value of the position-fixed temperature component x1 t n It is calculated as follows. x1 t n =X1 t-1 n +ΔX1 t-1 n (6)

[0061] In this way, the position fixed temperature component prediction unit 12 may estimate the value of the position fixed temperature component using a model (first model) that assumes that a predetermined disturbance is applied to the position fixed temperature component.

[0062] Here, the predetermined disturbance applied to the position-fixed temperature component may be modeled based on knowledge of its generation mechanism. An example of the generation mechanism of the disturbance is collision of the flow of molten steel L1 with the solidified shell L2. The model obtained based on knowledge of the generation mechanism may be obtained by, for example, analyzing the correlation between the flow of molten steel L1 and its effect on the temperature of the mold copper plate M1. The model may be stored in the storage unit 20 as model information 22.

[0063] (Moving temperature component prediction unit 13) The moving temperature component prediction unit 13 is a functional unit that uses a second model to calculate, for each temperature measurement point 51, a predicted value of the moving temperature component of the temperature measurement value a predetermined time after the previous optimal estimated value, based on the optimal estimated value of the moving temperature component in the previous estimation. First, the temperature calculation points located on the z-axis and used by the moving temperature component prediction unit 13 will be described. Figure 6 is a diagram showing the positions of the temperature calculation points on the z-axis.

[0064] As shown in FIG. 6, the temperature calculation points are arranged on the z-axis at predetermined intervals on the solidified shell L2 generated in the mold M, and are multiple virtual points that move together with the solidified shell L2. The moving temperature component prediction unit 13 estimates the moving temperature component for each temperature calculation point. That is, The position-fixed temperature component prediction unit 12 estimates the position-fixed temperature component on the p-axis, and the moving temperature component prediction unit 13 estimates the moving temperature component on the z-axis.

[0065] The temperature calculation points on the z axis are preferably arranged at regular intervals Δz. The z-axis coordinate indicating the position of the m-th temperature calculation point on the z axis is expressed as z m In this case, the time when the mth temperature calculation point first appears on the z-axis of the solidified shell L2 is defined as t0 m The coordinate z of the temperature calculation point on the z axis is m is expressed by the following equation (7) using the z-axis coordinate of the (m-1)th temperature calculation point at the same time.

[0066]

number

[0067] In addition, when m=1, t=t0 1 The mth temperature calculation point is V t-1 It can be assumed that the temperature advances in the z-axis direction by Δt. Then, the coordinate z on the z-axis at the m-th temperature calculation point at time t is t m is the z-axis coordinate z a given time ago. t-1 m can be expressed as the following equation (8).

[0068]

number

[0069] z t-1 m From z t m The moving temperature component of the mth temperature calculation point moving to T2 t-1 m , the disturbance applied during the movement is ΔT2 t m Then, from the above formula (4) and formula (8), the moving temperature component at the m-th temperature calculation point at time t can be modeled by the following formula (9). T2 t m =T2 t-1 m +ΔT2 t-1 m (9)

[0070] Here, the predicted value of the moving temperature component calculated by the moving temperature component prediction unit 13 is represented as x2. Also, the optimal estimated value of the moving temperature component, which is the value obtained by optimizing the predicted value by the temperature component optimization unit 15 described later, is represented as X2. Also, the disturbance of the moving temperature component optimized by the temperature component optimization unit 15 is represented as ΔX2.

[0071] The moving temperature component prediction unit 13 calculates the moving temperature component at the m-th temperature calculation point at time t by calculating the predicted value x2 t m The optimal estimate of the moving temperature component at time t-1 when the temperature was measured is X2 t-1 m Specifically, the moving temperature component prediction unit 13 calculates the optimal estimated value X2 of the moving temperature component as shown in the following equation (10). t-1 m The optimal estimate of the disturbance ΔX2 t-1 m The value added is the predicted value of the moving temperature component x2 t m It is calculated as follows. x2 t m =X2 t-1 m +ΔX2 t-1 m (10)

[0072] In this way, the moving temperature component prediction unit 13 may estimate the value of the moving temperature component using a model (second model) that assumes that a predetermined disturbance is applied to the moving temperature component at time t, a predetermined time after the state estimation at time t-1. In this case, the moving temperature component prediction unit 13 may calculate the predicted value of the moving temperature component using, as the second model, a model that assumes that the value of the moving temperature component at the temperature calculation point is the same as the value at a point in time a predetermined time before.

[0073] That is, the moving temperature component prediction unit 13 may estimate the value of the moving temperature component using a model that assumes that the predetermined disturbance is 0 (zero). Disturbances applied to the moving temperature component have the property of suddenly occurring due to the intrusion of foreign matter, etc., and therefore estimation using a model that uses, for example, a statistical method is difficult. This is in contrast to values ​​that can be estimated based on flow analysis data, such as disturbances of position-fixed temperature components, for example, molten steel L1. Therefore, considering the balance between the processing load of the estimation device 1 required to estimate the disturbance of the moving temperature component and the impact on the estimation accuracy of the moving temperature component, it is preferable from the perspective of efficient estimation to assume that the predetermined disturbance applied to the moving temperature component is 0.

[0074] In this way, the moving temperature component prediction unit 13 calculates the predicted value of the moving temperature component for each temperature calculation point. Then, for each temperature measurement point 51, the moving temperature component prediction unit 13 calculates the predicted value of the moving temperature component by proportionally dividing the predicted values ​​of the moving temperature component at the nearest temperature calculation point and the next nearest temperature calculation point. The moving temperature component prediction unit 13 calculates the predicted value of the moving temperature component by dividing the predicted values ​​of the moving temperature component at the nearest temperature calculation point and the next nearest temperature calculation point. t m and the coordinate p of the temperature measuring point 51 on the p axis. n Two temperature calculation points that satisfy the following equation (11) are selected for z t L -z t m+1 <p n ≦z t L -z t m (11)

[0075] The selected temperature calculation points are the closest temperature calculation point to the n-th temperature measurement point 51 and the next closest temperature calculation point, one of which is represented as the m-th temperature calculation point and the other as the m+1-th temperature calculation point. The m+1-th temperature calculation point is located closer to the molten steel surface L3 than the n-th temperature measurement point 51, and the m-th temperature calculation point is located on the opposite side, closer to the casting direction than the temperature measurement point 51. t L indicates the coordinate on the z-axis corresponding to the coordinate p0 on the p-axis at time t.

[0076] The moving temperature component prediction unit 13 calculates a pro rata value of the predicted value of the moving temperature component at the n-th temperature measurement point 51 according to the distance from the m-th and (m+1)-th temperature calculation points to the n-th temperature measurement point 51. The pro rata value can be calculated by the following equation (12). The moving temperature component prediction unit 13 sets the calculated pro rata value as the predicted value of the moving temperature component at the n-th temperature measurement point 51. x2 t n =α t n x2 t m+1 +(1-α t n )x2 tm (12)

[0077] where α t n is a value obtained by dividing the distance from the n-th temperature measurement point 51 to the m-th temperature calculation point by the interval Δz between the temperature calculation points, and is expressed by the following equation (13). α t n =(z t L -z t m -p n ) / Δz (13)

[0078] The position-fixed temperature component prediction unit 12 and the moving temperature component prediction unit 13 may store the predicted values ​​of each temperature component as temperature measurement value information 24 in the storage unit 20 at predetermined time intervals.

[0079] (Temperature prediction unit 14) The temperature prediction unit 14 is a functional unit that calculates, for each temperature measurement point 51, the sum of the predicted value of the position-fixed temperature component a predetermined time after the previous estimation and the predicted value of the moving temperature component a predetermined time after the previous estimation as the predicted value of the measured temperature a predetermined time after the previous estimation. That is, the temperature prediction unit 14 adds up the predicted value of the position-fixed temperature component calculated by the position-fixed temperature component prediction unit 12 and the predicted value of the moving temperature component calculated by the moving temperature component prediction unit 13, and calculates the predicted value of the measured temperature as an overall temperature independent of the type of component.

[0080] If the predicted value of the temperature measurement is represented as "y", the predicted value y of the temperature measurement at the nth temperature measurement point 51 at time t is t n is expressed by the following equation (14). y t n =x1 t n +x2 t n (14)

[0081] The temperature prediction unit 14 may store the predicted temperature value in the storage unit 20 as temperature value information 24 at predetermined time intervals.

[0082] (Temperature Component Optimization Unit 15) The temperature component optimization unit 15 is a functional unit that estimates the estimated temperature values ​​for each temperature measurement point 51 by minimizing the estimation error between the actual temperature measurement value and the estimated temperature measurement value based on the predicted temperature measurement value a predetermined time after the previous estimation. The estimated temperature values ​​are the optimal estimated value of the fixed temperature component and the optimal estimated value of the moving temperature component. That is, the temperature component optimization unit 15 optimizes the estimated values ​​of the fixed temperature component and the moving temperature component using a statistical method. Specifically, the temperature component optimization unit 15 calculates an estimation error, which is the difference between the actual temperature measurement value and the estimated value at each temperature measurement point 51, every predetermined time. Then, using the estimated error and the predicted value of the fixed temperature component and the predicted value of the moving temperature component at each temperature measurement point 51, the temperature component optimization unit 15 calculates the optimal estimated value of the fixed temperature component and the optimal estimated value of the moving temperature component after obtaining the actual temperature measurement value, so as to estimate the actual temperature measurement value with the minimum error.

[0083] For example, the temperature component optimization unit 15 calculates the temperature measurement value estimation error Y t n -y t n The temperature component optimization unit 15 uses the temperature measurement value estimation error, the predicted value of the moving temperature component, and the predicted value of the position-fixed temperature component to calculate an optimal solution to an optimization problem for estimating the actual temperature measurement value with minimal error, with the estimated value of the position-fixed temperature component and the estimated value of the moving temperature component as decision variables. The temperature component optimization unit 15 may calculate the optimal estimated value of each temperature measurement value component, for example, so as to minimize the expected value of the variance of the temperature measurement value estimation error, which is the deviation between the actual temperature measurement value and the sum of the estimated value of the position-fixed temperature component and the estimated value of the moving temperature component.

[0084] The temperature component optimization unit 15 may statistically correct the estimated values ​​calculated by the position-fixed temperature component prediction unit 12 and the moving temperature component prediction unit 13, for example, using the model shown below, and calculate the optimal estimated values ​​for each.

[0085] First, a model for estimating each disturbance will be described. t n is a Gaussian noise signal w1 with a mean of 0, as shown in the following equation (15). t n We assume that it can be modeled by ΔT1 t n =w1 t n (15)

[0086] Also, the disturbance ΔT2 of the moving temperature component at the mth temperature calculation point at time t t m is a Gaussian noise signal w2 with a mean of 0, as shown in the following equation (16). t m We assume that it can be modeled by ΔT2 t m =w2 t m (16)

[0087] The optimal estimate of the Gaussian noise signal is equal to the average value of the Gaussian noise signal. t-1 n The optimal estimate of ΔX1 t-1 n The predicted value x1 of the position-fixed temperature component at time t is set to 0. t n From the above formula (6), the value is obtained by the following formula (17) at every predetermined time. x1 t n =X1 t-1 n (17)

[0088] In addition, the disturbance ΔT2 calculated by the moving temperature component prediction unit 13 t-1 m The optimal estimate of ΔX2 t-1 m is set to 0, and the predicted value of the moving temperature component at time t is x2 t mFrom the above formula (10), the value is obtained by the following formula (18) at every predetermined time. x2 t m =X2 t-1 m (18)

[0089] The estimation device 1 acquires the temperature measurement value at the temperature measurement point 51 and can calculate the actual temperature measurement value at time t. As shown below, the time changes of the fixed-position temperature component, the moving temperature component, and the respective disturbances can be described by a state space model. Therefore, the temperature component optimization unit 15 can obtain an optimal solution for the optimization problem at predetermined time intervals using, for example, a Kalman filter algorithm.

[0090] The following describes an example in which the temperature component optimization unit 15 calculates an optimal solution to the optimization problem using a Kalman filter algorithm. Here, the magnitude of fluctuation in the value of the position-fixed temperature component is represented as a first constant, and the magnitude of fluctuation in the value of the moving temperature component is represented as a second constant. The temperature component optimization unit 15 corrects the predicted values ​​of the position-fixed temperature component and the moving temperature component using a Kalman filter algorithm that uses the first and second constants, and calculates the optimal estimated values ​​of the position-fixed temperature component and the moving temperature component after obtaining the actual value of the current temperature measurement.

[0091] The method used by the temperature component optimization unit 15 to calculate the optimal solution to the optimization problem is not limited to the method using the Kalman filter algorithm. For example, when it is assumed that the disturbance or the like has a non-Gaussian distribution, the temperature component optimization unit 15 may use a method using the particle filter algorithm.

[0092] As shown in FIG. 6, the m-th temperature calculation point and the (m+1)-th temperature calculation point are located at the coordinate p on the p-axis where the N-th temperature measurement point 51 is located. N It is also assumed that the (m+M−2)th temperature calculation point and the (m+M−1)th temperature calculation point satisfy the relationship of formula (11) with respect to the coordinate p1 on the p axis where the first temperature measurement point 51 is located.

[0093] (State space model showing each component of the temperature measurement) Regarding the above formulas (3), (9), (15), and (16), if the variables are treated as vectors and the coefficients are treated as matrices, the following formula (19) can be obtained. Note that "T" indicates the transpose of a matrix or vector only when it is attached to the right superscript of a matrix. T t =[T1 t T T2 t T ] T (19)

[0094] Here, matrix T1 t and matrix T2 t are expressed by the following formulas (19-1) and (19-2), respectively. T1 t =[T1 t 1 T1 t 2 … T1 t N ] T (19-1) T2 t =[T2 t m T2 t m+1 … T2 t m+M-1 ] T (19-2)

[0095] The disturbance vector in the model of each component of the temperature measurement value is expressed by the following equation (20). w t =[w1 t T w2 t T ] T (20)

[0096] Here, the matrix w1 t and matrix w2 t are expressed by the following formulas (20-1) and (20-2), respectively. w1 t =[w1 t1 w1 t 2 … w1 t N ] T (20-1) w2 t =[w2 t m w2 t m+1 … w2 t m+M-1 ] T (20-2)

[0097] Then, the state space model of each component of the temperature measurement value is expressed by the following equation (21). T t+1 =FT t +Gw t (twenty one)

[0098] Here, F is a matrix with (N+M) rows and (N+M) columns as shown in the following formula (22), and G is a matrix with (N+M) rows and (M+N) columns as shown in the following formula (23).

[0099]

number

[0100] In addition, "0 R×C " indicates a zero matrix with R rows and C columns.

[0101] The matrix F1 is an N-row, N-column matrix, and each element is F1 n,n =1 (where n=1,...,N), and the other components are 0.

[0102] The matrix F2 is a matrix with M rows and M columns, and each element is m,m = 1 (where m = 1, ..., M), and the other components are 0.

[0103] The matrix G1 is an N-row, N-column matrix, and each element is G1 n,n =1 (where n=1,...,N), and the other components are 0.

[0104] The matrix G2 is an M-row, M-column matrix, and each element is G2 m,m = 1 (where m = 1, ..., M), and the other components are 0.

[0105] In this embodiment, the disturbance w1 in the fixed-position temperature component included in the above equation (15) is t n With mean 0 and standard deviation σ1 n Also, the disturbance in the moving temperature component included in the equation (16) is assumed to be Gaussian noise. t m With mean 0 and standard deviation σ2 m When these disturbances are grouped together as an N+M-dimensional vector, the variance-covariance matrix Q has a mean of 0, and we define Q=diag(σ11 2 ,σ12 2 ,…,σ1 N 2 ,σ2 m 2 ,σ2 m+1 2 ,…,σ2 m+M-1 2 ) is expressed as "diag(a1,...,a K )" is a1,...,a K Here, we denote a diagonal matrix with K rows and K columns.

[0106] At this time, the standard deviation of the Gaussian noise of the disturbance, especially the Gaussian noise related to the position-fixed temperature component, is σ1 n can express the magnitude of the surface temperature variation due to the frequency of air gaps occurring for each cast steel type and the type of mold powder used. Therefore, for each cast steel type, the standard deviation σ n It is preferable to optimize each of

[0107] Next, for a model of the actual temperature measurement values ​​at each temperature measurement point 51, first, the left side of the equation (5) and the observation noise for n=1, . . . , N are organized into vectors as shown in the following equations (24) and (25). Y t =[Yt 1 … Y t N ] T (twenty four) v t =[ν t 1 … ν t N ] T (twenty five)

[0108] And, α in the above formula (13) t n Matrix H using t is a matrix with N rows and (N+M) columns, and the actual temperature measurement values ​​are modeled as shown in the following equation (26). t For n=1,…,N, m=1,…,M, the (n,n) component is set to 1, and the (n,N+m) component is set to 1-α t n and the (n, N+m+1) component is α t n and other components are set to 0. Y t =H t T t +v t (26)

[0109] Here, the observed noise in the above formula (2) has an average of 0 and a standard deviation of σ ν n The variance-covariance matrix R of the entire observation noise is expressed as a diagonal matrix with N rows and N columns shown in the following equation (27). R=diag((σ ν 1 ) 2 ,…,(σ ν N ) 2 ) (27)

[0110] The predicted values ​​and optimal estimated values ​​of the fixed-position temperature component and the moving temperature component are summarized as vectors using the above equations (6) and (10). The predicted value vector for each component of the temperature measurement value is expressed as the following equation (28). x t =[x1 t T x2t T ] T (28)

[0111] Here, matrix x1 t and matrix x2 t are expressed by the following formulas (28-1) and (28-2), respectively. x1 t =[x1 t 1 x1 t 2 … x1 t N ] T (28-1) x2 t =[x2 t m x2 t m+1 … x2 t m+M-1 ] T (28-2)

[0112] The optimal estimated value vector for each component of the temperature measurement value is expressed as the following equation (29). X t =[X1 t T X2 t T ] T (29)

[0113] Here, matrix X1 t and matrix X2 t are expressed by the following formulas (29-1) and (29-2), respectively. X1 t =[X1 t 1 X1 t 2 … X1 t N ] T (29-1) X2 t =[X2 t m X2 t m+1 …X2 t m+M-1 ] T (29-2)

[0114] Then, the predicted values ​​of the position-fixed temperature component and the moving temperature component are expressed as the following equation (30) using the matrix F of equation (22) and the definitions of equations (28) and (29) from equations (6) and (10). x t =FX t-1 (30)

[0115] Further, the predicted value of the temperature measurement value at each temperature measurement point 51 is calculated from the above equation (11) for m=1, . . . , M by the matrix H t Using the above formula (14), it is expressed as the following formula (31). y t =H t x t (31)

[0116] The temperature component optimization unit 15 optimizes the Kalman gain matrix K t Calculate the Kalman gain matrix K t is a matrix with (N+M) rows and N columns, and may be calculated as shown below:

[0117] The estimated value of the estimation error covariance matrix in the predicted value vector of each component of the temperature measurement value shown in the above equation (28) is expressed as a matrix P t The estimation error covariance matrix in the optimal estimation value vector of each component of the temperature measurement value shown in the above equation (29) is expressed as the matrix S t It is expressed as:

[0118] matrix P t Each component indicating the estimation error covariance in the predicted value vector of the position-fixed temperature component included in the matrix P t Each component of the matrix S, which indicates the estimation error covariance in the predicted value vector of the moving temperature component, is included in the second constant, which indicates the magnitude of the fluctuation of the value of the moving temperature component. t Each component indicating the estimation error covariance in the optimal predicted value vector of the position-fixed temperature component included in the matrix S is included in the first constant indicating the magnitude of fluctuation in the value of the position-fixed temperature component.t Each component indicating the estimation error covariance in the optimal predicted value vector of the moving temperature component, included in the second constant indicating the magnitude of fluctuation in the value of the moving temperature component, is included in the second constant.

[0119] Thus, the matrix P t and matrix P t includes a first constant indicating the magnitude of fluctuation in the value of the position-fixed temperature component and a second constant indicating the magnitude of fluctuation in the value of the moving temperature component. That is, both the first constant and the second constant may indicate the magnitude of fluctuation in the predicted value of each component of the temperature measurement value as well as the magnitude of fluctuation in the optimal estimate.

[0120] matrix P t and matrix S t are non-negative definite symmetric matrices S -1 is used as an initial value, and is calculated using the following equations (32) and (33). P t =FS t-1 F T +GQG T (32) S t =P t -K t H t P t (33)

[0121] Kalman gain matrix K t can be calculated using the following formula (34). K t =P t H t T (H t P t H t T +R) -1

[0122] Using the above model, the temperature component optimization unit 15 calculates the optimal estimated value vector for each component of the temperature measurement value using the following equation (35). As a result, the temperature component optimization unit 15 can calculate the optimal estimated value for each component of the temperature measurement value after obtaining the actual value of the current temperature measurement value by correcting the predicted value of the position-fixed temperature component and the predicted value of the moving temperature component at each temperature measurement point 51 at predetermined time intervals. X t =x t +K t (Y t -y t ) (35)

[0123] The temperature component optimization unit 15 may store the optimized estimated values ​​of each component of the temperature measurement value as temperature measurement value information 24 in the storage unit 20 at predetermined time intervals.

[0124] (Casting state abnormality estimation unit 16) The casting state abnormality estimation unit 16 is a functional unit that detects deviations from the standard state of the slab surface or solidified shell L2 based on the values ​​of the moving temperature components calculated using the optimal estimated values ​​of the moving temperature components at each position on the z-axis, which is the second coordinate axis extending in the casting direction, on the slab produced in the mold M, calculated by the temperature component optimization unit 15. Furthermore, the casting state abnormality estimation unit 16 detects deviations from the standard state caused by anything other than the slab surface or solidified shell L2 at each position on the p-axis, which is the first coordinate axis, based on the values ​​of the optimal estimated values ​​of the position-fixed temperature components optimized by the temperature component optimization unit 15.

[0125] The casting state abnormality estimation unit 16 may detect each abnormality, which is a deviation from the standard state, by, for example, using a method of detecting statistical outliers within a predetermined time range for the optimally estimated temperature component vector estimated at predetermined time intervals.The casting state abnormality estimation unit 16 can detect abnormalities, such as foreign matter entrapment and vertical cracks on the slab surface on the p-axis and poor growth of the solidified shell L2, as deviations from the standard state from the optimally estimated value of the moving temperature component.In addition, the casting state abnormality estimation unit 16 can detect abnormalities, such as mold powder inflow or crystallization abnormalities and local collision of the molten steel discharge flow, at a fixed position within the mold M as deviations from the standard state from the optimally estimated value of the position-fixed temperature component.

[0126] The abnormal casting state estimation unit 16 may output a notification indicating that any of the abnormalities has been detected via the output unit 40. The abnormal casting state estimation unit 16 may also output the notification including information indicating the cause of the abnormality. The information indicating the cause of the abnormality may be, for example, information indicating whether the abnormality is a "deviation from the standard state of the surface of the slab or the solidified shell L2" or a "deviation from the standard state caused by something other than the surface of the slab or the solidified shell L2."

[0127] In this way, the abnormal casting state estimation unit 16 can estimate whether the cause of the abnormality is due to the slab surface and the solidified shell L2 or something else. Therefore, the user of the estimation device 1 can take appropriate measures during casting based on the estimation result by the abnormal casting state estimation unit 16.

[0128] Furthermore, the casting state abnormality estimation unit 16 may output position information of the temperature measurement point 51 where the abnormality was detected. This allows the user of the estimation device 1 to identify the position in the longitudinal direction of the slab with respect to the change in the temperature measurement value that moves together with the solidified shell L2, and therefore allows pinpoint adjustment of the slab using, for example, a grinder, etc., thereby improving the casting yield.

[0129] Furthermore, the casting state abnormality estimation unit 16 can also detect the occurrence of an abnormality within the mold, such as a molten steel flow collision occurring at a fixed position within the mold M. Therefore, the user of the estimation device 1 can identify specific effects on the continuous casting machine, such as clogging of the submerged entry nozzle, from the information indicating the cause of the abnormality.

[0130] (Variation) The device 1 for estimating the state inside the mold according to one embodiment of the present invention is not limited to the above-described embodiment, and various modifications are conceivable.

[0131] For example, the calculation of the proportional value according to the equation (12) executed by the moving temperature component prediction unit 13 may be executed not by the moving temperature component prediction unit 13 but by, for example, a separately set functional block (not shown). In this case, the moving temperature component prediction unit 13 calculates a predicted value of the moving temperature component at each temperature calculation point. Then, the functional block may calculate a predicted value of the moving temperature component at the n-th temperature measurement point 51 using the predicted value of the moving temperature component at each temperature calculation point.

[0132] Furthermore, each function of the control unit 10 of the estimation device 1 may be realized by multiple devices. For example, the estimation device 1 may be configured with a temperature measurement value prediction device and a temperature measurement value optimization device. In this case, the temperature measurement value prediction device has a position-fixed temperature component prediction unit 12, a moving temperature component prediction unit 13, and a temperature prediction unit 14. The temperature measurement value optimization device has a temperature component optimization unit 15.

[0133] In this way, when the estimation device 1 is configured by multiple devices, the functions of each device are not limited to those described above and may be set arbitrarily. Furthermore, each device may be directly connected to each other via a wired or wireless connection, or may be indirectly connected via a network such as the Internet.

[0134] [Method for estimating the state inside the mold] A method for estimating the state inside a mold according to one embodiment of the present invention will be described below. Hereinafter, the method for estimating the state inside a mold according to one embodiment of the present invention may be simply referred to as "this estimation method." This estimation method may be executed by an estimation device 1. Regarding the details of the processing in each step shown below, matters that have already been described in the description of the estimation device 1 will not be described again.

[0135] 7 is a flow chart showing an example of this estimation method. As shown in FIG. 7, first, the position-fixed temperature component prediction unit 12 calculates a predicted value of the position-fixed temperature component at each temperature measurement point 51 using a first model (S2, position-fixed temperature component prediction step). Furthermore, the moving temperature component prediction unit 13 calculates a predicted value of the moving temperature component using a second model (S3, moving temperature component prediction step). Here, the moving temperature component prediction unit 13 first calculates a predicted value of the moving temperature component at each temperature calculation point on the z-axis (S3). Then, based on the predicted value of the moving temperature component at each temperature calculation point and information on the distance from each temperature calculation point to each temperature measurement point, the moving temperature component prediction unit 13 calculates a predicted value of the moving temperature component at each temperature measurement point 51 (S4).

[0136] Next, the temperature prediction unit 14 calculates the sum of the predicted value of the position-fixed temperature component and the predicted value of the moving temperature component as the predicted value of the measured temperature (S5, temperature prediction step).

[0137] Next, the temperature component optimization unit 15 calculates an estimation error, which is the difference between the actual measurement value and the estimated value (S6).The temperature component optimization unit 15 then optimizes the estimated values ​​of the position-fixed temperature component and the moving temperature component so that the variance estimate of the estimation error, which is the difference between the actual measurement value and the estimated value, is minimized.The temperature component optimization unit 15 uses the estimation error and the estimated values ​​of the position-fixed temperature component and the moving temperature component to calculate the optimal estimated values ​​of the position-fixed temperature component and the moving temperature component (S7, temperature component optimization step).

[0138] The abnormal casting state estimation unit 16 may also detect an abnormality on the surface of the slab or the solidified shell L2 using the optimal estimated value of the moving temperature component at each position on the z axis (S8).The abnormal casting state estimation unit 16 may also detect an abnormality caused by something other than the surface of the slab or the solidified shell L2 at each position on the p axis using the optimal estimated value of the position-fixed temperature component (S9).

[0139] When such an abnormality is detected, the casting state abnormality estimation unit 16 may output a notification indicating the detection of the abnormality (S10). The notification may include an estimation result as to whether the cause of the abnormality is due to the surface of the slab or the solidified shell L2, or due to some other cause.

[0140] The estimation device 1 may repeatedly perform these processes at predetermined time intervals. The estimation device 1 may end the series of processes, for example, when it receives a signal indicating the end of casting. The signal may be input by a user through the input unit 30, or may be input from another device such as the casting speed measuring device 60. Furthermore, the estimation device 1 may determine whether a predetermined planned casting time has elapsed each time a series of processes is completed, and may end the series of processes when it determines that the predetermined planned casting time has elapsed.

[0141] [Software implementation example] The functions of the estimation device 1 (hereinafter referred to as the "device") can be realized by a program that causes a computer to function as the device, and a program that causes a computer to function as each control block of the device (particularly each part included in the control unit 10).

[0142] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The control device and storage device execute the program to realize each function described in each of the above embodiments.

[0143] The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be supplied to the device via any wired or wireless transmission medium.

[0144] Furthermore, some or all of the functions of the control blocks can be realized by logic circuits. For example, an integrated circuit in which a logic circuit that functions as each of the control blocks is formed is also included in the scope of the present invention. In addition, the functions of the control blocks can also be realized by, for example, a quantum computer.

[0145] Furthermore, each process described in each of the above embodiments may be executed by AI (Artificial Intelligence). In this case, the AI ​​may run on the control device or on another device (for example, an edge computer or a cloud server).

[0146] 〔summary〕 A first aspect of the present invention provides an in-mold condition estimation device that estimates a condition inside a mold based on temperature values ​​measured at a plurality of temperature measurement points set along the casting direction of a mold of a continuous casting machine, and includes a position-fixed temperature component prediction unit that uses a first model to calculate, for each temperature measurement point, a predicted value of the position-fixed temperature component of the temperature measured at a predetermined time after the previous estimation of the optimal estimate, based on the optimal estimate of the position-fixed temperature component in the previous estimation, when the estimated value of the temperature measured is a value obtained by correcting a predicted value of the temperature measured at a plurality of temperature measurement points set along the casting direction of the mold, and the estimated value of the temperature measured at a predetermined time after the previous estimation of the optimal estimate, if the optimal estimate of the temperature measured at the temperature measurement point is a value obtained by minimizing the estimation error between the actual value of the temperature measured at the temperature measurement point and the estimated value of the temperature measured at the temperature measurement point, and the position-fixed temperature component prediction unit calculates, for each temperature measurement point, a predicted value of the position-fixed temperature component of the temperature measured at a predetermined time after the previous estimation of the optimal estimate, based on the optimal estimate of the position-fixed temperature component in the previous estimation; The system includes a moving temperature component prediction unit that uses a second model to calculate, for each temperature measurement point, a predicted value of the moving temperature component of the temperature measurement value a predetermined time after the previous estimation based on an optimal estimated value of the moving temperature component in the previous estimation; a temperature prediction unit that calculates, for each temperature measurement point, the sum of the predicted value of the position-fixed temperature component a predetermined time after the previous estimation and the predicted value of the moving temperature component a predetermined time after the previous estimation as a predicted value of the temperature measurement value a predetermined time after the previous estimation; and a temperature component optimization unit that estimates, for each temperature measurement point, the estimated values ​​of the temperature measurement value for each of the position-fixed temperature component and the moving temperature component, which are obtained by minimizing the estimation error between the actual value of the current temperature measurement value and the estimated value of the temperature measurement value, to be the optimal estimated value of the position-fixed temperature component and the optimal estimated value of the moving temperature component, respectively, based on the predicted value of the temperature measurement value a predetermined time after the previous estimation.

[0147] In the in-mold state estimation device according to aspect 2 of the present invention, in aspect 1, the moving temperature component prediction unit may calculate predicted values ​​of the moving temperature component at a plurality of temperature calculation points that are set at predetermined length intervals on the solidified shell generated in the mold and move along with the solidified shell, and may calculate the predicted value of the moving temperature component by proportionally dividing the predicted values ​​of the moving temperature component at the nearest temperature calculation point and the next nearest temperature calculation point for each temperature measurement point.

[0148] In the mold interior condition estimation device according to aspect 3 of the present invention, in aspect 1 or 2, the position-fixed temperature component prediction unit may calculate a predicted value of the position-fixed temperature component using a model that assumes that a predetermined external disturbance is applied to the first model, and the moving temperature component prediction unit may calculate a predicted value of the moving temperature component using a model that assumes that the value of the moving temperature component at the temperature calculation point is the same as the value at a point in time prior to the predetermined time.

[0149] In the mold in-state estimation device according to aspect 4 of the present invention, in any of aspects 1 to 3, the temperature component optimization unit may optimize the first model and the second model using a Kalman filter algorithm that uses a first constant representing the magnitude of fluctuation in the value of the position-fixed temperature component and a second constant representing the magnitude of fluctuation in the value of the moving temperature component.

[0150] The in-mold condition estimation device of aspect 5 of the present invention, in any of aspects 1 to 4, includes a casting condition abnormality estimation unit, and the casting condition abnormality estimation unit may detect deviations from the standard state of the surface of the slab or the solidified shell occurring within the mold using the optimal estimated value of the moving temperature component at each position on a second coordinate axis extending in the casting direction on the slab produced in the mold calculated by the temperature component optimization unit, and may detect deviations from the standard state caused by factors other than the surface of the slab or the solidified shell at each position on a first coordinate axis extending in the casting direction on the mold using the optimal estimated value of the position-fixed temperature component.

[0151] A program according to a sixth aspect of the present invention is a program for causing a computer to function as the in-mold state estimation device according to the first aspect.

[0152] A seventh aspect of the present invention provides a method for estimating a state inside a mold, based on temperature measurements obtained at a plurality of temperature measurement points set along the casting direction of a mold of a continuous casting machine, for estimating a state inside the mold, the method comprising the steps of: correcting a predicted value of the temperature measurement value using a variable to set the estimated value of the temperature measurement value; and, when an optimal estimate of the temperature measurement value, which is the estimated value of the temperature measurement value that minimizes an estimation error between the actual value of the temperature measurement value and the estimated value of the temperature measurement value, is expressed as the sum of an optimal estimate of a position-fixed temperature component, the value of which does not move with respect to the mold, and an optimal estimate of a moving temperature component, the value of which moves with respect to the mold with respect to the mold; and, for each of the temperature measurement points, using a first model, calculating a predicted value of the position-fixed temperature component of the temperature measurement value a predetermined time after the previous estimation of the optimal estimate, based on the optimal estimate of the position-fixed temperature component in the previous estimation; The method includes a moving temperature component prediction step for calculating, for each temperature measurement point, a predicted value of the moving temperature component of the temperature measurement value a predetermined time after the previous estimation based on an optimal estimated value of the moving temperature component in the previous estimation using a second model; a temperature prediction step for calculating, for each temperature measurement point, the sum of the predicted value of the position-fixed temperature component a predetermined time after the previous estimation and the predicted value of the moving temperature component a predetermined time after the previous estimation as the predicted value of the temperature measurement value a predetermined time after the previous estimation; and a temperature component optimization step for estimating, for each temperature measurement point, the estimated values ​​of the temperature measurement value for each of the position-fixed temperature component and the moving temperature component, which are obtained by minimizing the estimation error between the actual value of the current temperature measurement value and the estimated value of the temperature measurement value, to be the optimal estimated value of the position-fixed temperature component and the optimal estimated value of the moving temperature component, respectively, based on the predicted value of the temperature measurement value a predetermined time after the previous estimation.

[0153] [Additional Notes] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. [Example]

[0154] An embodiment of the present invention will be described below. In the embodiment shown below, the results of an operation simulation of the estimation device 1 are shown, assuming that a foreign object has been caught on the surface of a cast slab.

[0155] On the mold copper plate M1 of the mold M, six temperature measurement points 51 are installed in the casting direction at regular intervals of 50 mm in a row R shown in Figure 4. The temperature calculation points on the z axis are arranged at regular intervals of Δz = 20 mm.

[0156] In this example, it is assumed that a foreign object has been caught on the slab surface, and that on the p-axis coordinate, a temperature waveform with a length of 100 mm has been generated in which the temperature drops from the normal temperature distribution by a maximum of 10°C in a triangular wave shape before returning to normal. The standard deviation of the disturbance in the state space model and the standard deviation of Gaussian noise in the observation noise are shown in Table 1 below. In this example, all measured temperature values ​​are expressed in degrees Celsius (Cel). The predetermined time is set to 1 second, and the estimation device 1 performs estimation processing every second.

[0157] [Table 1]

[0158] FIG. 8 shows a graph plotting measured temperature values ​​caused by the inclusion of foreign matter on the slab surface. FIG. 9 shows a graph plotting changes in measured temperature values ​​at each temperature measurement point over a predetermined time period. In this example, it is assumed that temperature components caused by the inclusion of foreign matter occur at the positions shown in FIG. 8 relative to the casting length of the slab. It is also assumed that changes in measured temperature values ​​over time as shown in FIG. 9 are observed at each temperature measurement point 51. The casting speed in the simulation of this example was 40 mm / s.

[0159] In the simulation of this example, the first temperature calculation point on the z-axis at time 0 was placed at the target position of the molten steel surface L3 level, and thereafter, a new temperature calculation point was placed at the target position of the molten steel surface L3 level each time casting progressed by Δz. The initial value of the moving temperature component was set to 0.

[0160] 10 to 13 are graphs plotting actual values ​​and optimal estimated values ​​at each temperature measurement point 51 and each temperature calculation point at each time. In each of Fig. 10 to 13, the upper graph plots each value against the distance from the molten steel surface L3, and the lower graph plots each value against the casting length. In this example, the "actual value" refers to a value set as a condition in this simulation.

[0161] The results of Figures 10 to 13 show that for temperature calculation points where the actual value of the moving temperature component is other than 0, the optimal estimated value of the moving temperature component generally approaches the actual value while passing through the section where the temperature measurement point 51 in the mold M is located.

[0162] Figure 14 shows a graph summarizing the root mean square error (RMSE) between the actual values ​​and the optimally estimated values ​​of the moving temperature component shown in Figures 10 to 13, and the number of temperature measurement points 51 passed by the temperature drop portion. The temperature drop portion indicates a portion on the slab surface where the temperature has dropped. The results in Figure 14 indicate that the estimation accuracy of the moving temperature component improves as the number of temperature measurement points 51 passed increases.

[0163] Fig. 15 shows a graph plotting the optimal estimated values ​​of each component of the temperature measurement value and the actual value of the fixed-position temperature component against time at each temperature measurement point 51. The results in Fig. 15 show that the moving temperature component and the fixed-position temperature component are estimated separately at each temperature measurement point 51. It also shows that the estimation accuracy of each component of these temperature measurement values ​​improves as the number of temperature measurement points 51 passed by the temperature drop section increases. [Explanation of symbols]

[0164] 1 Estimation device 12 Position-fixed temperature component prediction section 13 Moving temperature component prediction section 14 Temperature prediction section 15 Temperature Component Optimization Section 16 Casting condition abnormality estimation unit 51 Temperature measurement points M mold M1 molded copper plate L2 solidified shell L3 Molten steel surface

Claims

1. A mold in-state estimation device that estimates a state inside a mold based on temperature measurement values ​​obtained at a plurality of temperature measurement points set along the casting direction of a mold of a continuous casting machine, a value obtained by correcting the predicted value of the temperature measurement value using a variable is set as an estimated value of the temperature measurement value; When the optimal estimate of the temperature measurement value, which is the estimate of the temperature measurement value that minimizes the estimation error between the actual value of the temperature measurement value and the estimated value of the temperature measurement value, is expressed as the sum of the optimal estimate of a position-fixed temperature component, which is a component whose value does not move over time relative to the mold, and the optimal estimate of a moving temperature component, which is a component whose value moves over time relative to the mold, a position-fixed temperature component prediction unit that calculates, for each of the temperature measurement points, a predicted value of the position-fixed temperature component of the temperature measurement value after a predetermined time has elapsed since the previous estimation of the optimal estimated value, based on the optimal estimated value of the position-fixed temperature component in the previous estimation, using a first model; a moving temperature component prediction unit that calculates, for each of the temperature measurement points, a predicted value of the moving temperature component of the temperature measurement value after a predetermined time has elapsed since the previous estimation of the optimal estimated value, based on the optimal estimated value of the moving temperature component in the previous estimation, using a second model; a temperature prediction unit that calculates, for each of the temperature measurement points, the sum of a predicted value of the position-fixed temperature component after a predetermined time from the previous estimation and a predicted value of the moving temperature component after a predetermined time from the previous estimation as a predicted value of the temperature measurement value after a predetermined time from the previous estimation; a temperature component optimization unit that estimates, for each of the temperature measurement points, the estimated values ​​of the temperature measurement values ​​for each of the position-fixed temperature component and the moving temperature component, which are obtained by minimizing the estimation error between the actual value of the current temperature measurement value and the estimated value of the temperature measurement value based on the predicted value of the temperature measurement value after a predetermined time from the previous estimation, as the optimal estimated value of the position-fixed temperature component and the optimal estimated value of the moving temperature component, respectively; An apparatus for estimating a state inside a mold, comprising:

2. The moving temperature component prediction unit calculating predicted values ​​of the moving temperature components at a plurality of temperature calculation points that are set at predetermined length intervals on the solidified shell generated in the mold and move together with the solidified shell; 2. The device for estimating a state inside a mold according to claim 1, wherein the predicted value of the moving temperature component at the closest temperature calculation point and the second closest temperature calculation point is divided proportionally for each of the temperature measurement points to calculate the predicted value of the moving temperature component.

3. the position-fixed temperature component prediction unit calculates a predicted value of the position-fixed temperature component using a model that assumes that a predetermined disturbance is added to the first model; 3. The device for estimating an in-mold state according to claim 2, wherein the moving temperature component prediction unit calculates the predicted value of the moving temperature component using, as the second model, a model that assumes that the value of the moving temperature component at the temperature calculation point is the same as the value at the time point the predetermined time prior.

4. 3. The device for estimating an in-mold state according to claim 1, wherein the temperature component optimization unit optimizes the first model and the second model using a Kalman filter algorithm that uses a first constant that represents a magnitude of fluctuation in the value of the position-fixed temperature component and a second constant that represents a magnitude of fluctuation in the value of the moving temperature component.

5. A casting state abnormality estimation unit is provided, The casting state abnormality estimation unit detecting deviations from a standard state of a surface of the slab or a solidified shell occurring in the mold using the optimal estimated values ​​of the moving temperature components at each position on a second coordinate axis extending in the casting direction on a slab produced in the mold calculated by the temperature component optimization unit; 3. The in-mold condition estimation device according to claim 1, wherein deviations from a standard state caused by factors other than the surface of the slab or the solidified shell at each position on the mold along a first coordinate axis extending in the casting direction are detected using the optimal estimated values ​​of the position-fixed temperature components.

6. A program for causing a computer to function as the in-mold state estimation device according to claim 1.

7. A method for estimating a state inside a mold of a continuous casting machine, which estimates a state inside the mold based on temperature measurement values ​​obtained at a plurality of temperature measurement points set along the casting direction of the mold, a value obtained by correcting the predicted value of the temperature measurement value using a variable is set as an estimated value of the temperature measurement value; When the optimal estimate of the temperature measurement value, which is the estimate of the temperature measurement value that minimizes the estimation error between the actual value of the temperature measurement value and the estimated value of the temperature measurement value, is expressed as the sum of the optimal estimate of a position-fixed temperature component, which is a component whose value does not move over time relative to the mold, and the optimal estimate of a moving temperature component, which is a component whose value moves over time relative to the mold, a position-fixed temperature component prediction step of calculating, for each of the temperature measurement points, a predicted value of the position-fixed temperature component of the temperature measurement value after a predetermined time from the previous estimation of the optimal estimated value, based on the optimal estimated value of the position-fixed temperature component in the previous estimation, using a first model; a moving temperature component prediction step of calculating, for each of the temperature measurement points, a predicted value of the moving temperature component of the temperature measurement value after a predetermined time from the previous estimation of the optimal estimated value, based on the optimal estimated value of the moving temperature component in the previous estimation, using a second model; a temperature prediction step of calculating, for each of the temperature measurement points, the sum of a predicted value of the position-fixed temperature component after a predetermined time from the previous estimation and a predicted value of the moving temperature component after a predetermined time from the previous estimation as a predicted value of the temperature measurement value after a predetermined time from the previous estimation; a temperature component optimization step of estimating, for each of the temperature measurement points, the estimated values ​​of the temperature measurement values ​​for each of the position-fixed temperature component and the moving temperature component, which are obtained by minimizing the estimation error between the actual value of the current temperature measurement value and the estimated value of the temperature measurement value based on the predicted value of the temperature measurement value after a predetermined time from the previous estimation, as the optimal estimated value of the position-fixed temperature component and the optimal estimated value of the moving temperature component, respectively; The method for estimating a state inside a mold comprises:

Images