Device Control Device, Device Control Method, and Computer-Readable Recording Medium

Abstract

An apparatus control device, an apparatus control method, and a program that predict the occurrence of an operation abnormality in a controlled device, appropriately operate a control operation unit so as not to cause an operation abnormality, and improve control effects and operation efficiency. A first operation process having a predetermined response speed to an operation and a second operation process having a slower response speed to the operation than the first operation process are performed on the controlled device. The first operation process is executed by a first operation unit, and the second operation process is executed by a second operation unit. A safety operation range determination unit that determines a safety operation range of the first operation process of the first operation unit based on the performance of the first operation unit is provided. In the determination of the safety operation range determination unit, when the first operation process of the first operation unit is not within the safety operation range, an instruction for the second operation process is corrected or changed so that the actual position of the first operation process of the first operation unit is moved to an actual position that is not presumed to cause an operation abnormality in terms of production material units.

Description

Technical Field

[0001] The present invention relates to a device control apparatus, a device control method, and a program. Background Art

[0002] When performing mill control, which is one of device controls, in shape control for controlling the undulating state of a plate, fuzzy control and neuro-fuzzy control have been conventionally applied. Fuzzy control is applicable to shape control using a coolant. In addition, neuro-fuzzy control is applicable to shape control of Sendzimir mills.

[0003] In the shape control to which neuro-fuzzy control is applied, as shown in Patent Document 1, processing is performed to obtain the difference between the actual shape pattern detected by a shape detector and a target shape pattern, and the similarity ratio with a preset reference shape pattern. And, in the shape control to which neuro-fuzzy control is applied, based on the obtained similarity ratio, a control output amount for an operation end is obtained through a control rule expressed by an operation amount of a control operation end for a preset reference shape pattern.

[0004] Shape control has a plurality of control operation ends, and control is executed by the difference in the characteristics of these plurality of control operation ends. The shape is the undulating state of the plate in the plate width direction, and the control operation end can change the shape of a specific region in the plate width direction. For example, the AS-U roll can change the shape near the saddle position being operated, and the intermediate roll movement can change the shape of the plate end. When performing shape control, according to the actual shape, each control operation end is combined to perform an action so as to suppress the shape deviation.

[0005] When the mill performs rolling, a cooling material (hereinafter referred to as a coolant) is required for lubrication between the rolled material and the rolls of the mill and cooling of the heat generated during rolling. This cooling material becomes an operation end of shape control, and by adjusting the injection amount of the coolant in the plate width direction, the shape can be changed over the entire region in the plate width direction. In a six-high mill, as shown in Patent Document 2, there is a coolant injection amount adjustment mechanism in the plate width direction, and shape control is implemented by changing the injection amount using the actual shape. However, in a Sendzimir mill, during rolling, the rolls of the mill are in a state of being immersed in the coolant, and it takes time until the effect of the coolant on the shape appears compared to the AS-U or intermediate roll movement. In addition, since the coolant injection amount cannot be automatically adjusted, there is also a case where an operator needs to perform operations such as operating a flow rate adjustment valve.

[0006] In addition, depending on the mechanical structure of the rolling mill, the flow rate of the coolant in the plate width direction may not be adjusted, but even in this case, the actual injection amount in the plate width direction can be changed by replacing the coolant injection nozzle. In order to circulate the coolant near the work roll, the Sendzimir rolling mill has a mechanism in which a mechanical device with multiple coolant injection nozzles in the plate width direction is set in the rolling mill. Therefore, the Sendzimir rolling mill can change the coolant flow rate in the plate width direction by preparing multiple mechanical devices with different coolant injection nozzles before rolling.

[0007] During the rolling process in a rolling mill, an abnormal operation occurs in which the material being rolled breaks. The material being rolled breaks mostly because of the material being rolled, but there are also cases where the material is broken because of meandering. Meaning that the material being rolled deviates to one side of the rolling mill, and rolling is usually performed in the center of the rolling mill.

[0008] Such meandering is expected to occur at the position where the AS-U roll or the intermediate roll moves. In shape control, the AS-U or the intermediate roll is operated to move in order to maintain the shape of the rolled material at the target shape, but as a result, the machine state may become a state where the rolled material is prone to meandering.

[0009] In the past, in common rolling mills such as 4-stage rolling mills and 6-stage rolling mills, in addition to the plate bending machine and the leveling machine as the mechanical operating unit, there is also an operating unit that changes the coolant injection amount in the plate width direction for shape control. Unlike the Sendzimir mill, in a common rolling mill, the rollers of the rolling mill are not immersed in the coolant, and the effect of the coolant on the shape can be obtained in the same time as the mechanical operating unit. Therefore, the shape control in a common rolling mill treats the mechanical operating unit and the coolant equally, and uses the coolant as a control operating end. In this case, the coolant also has an effect on the entire area in the plate width direction, and therefore conflicts with the mechanical operating unit. Even if the actual shape of the rolled material is the same, the actual position of the mechanical operating unit is mostly different.

[0010] In addition, in 4-stage and 6-stage rolling mills, the coolant flow rate in the plate width direction may not be adjusted during rolling due to the mechanical structure of the coolant device. In such a case, the operator may manually change the coolant flow rate in the plate width direction before rolling begins.

[0011] Figure 19 A schematic structure of a conventional control device for a Sendzimir rolling mill is shown.

[0012] First, the difference between the target shape d1 and the actual shape d2 of the rolled material obtained by the rolling mill 990 is obtained in the arithmetic unit 901, and this difference is provided to the first shape control unit 902. The first shape control unit 902 controls the mechanical operation end 904 as a mechanical operation unit.

[0013] The rolling mill 990 performs the mechanical operation process of the mechanical operation end 904 and the operation process of the coolant injection amount of the coolant operation end 905, and performs the rolling process of the rolled material to obtain the actual shape d2 and the rolling actual result d3 of the rolled material. In this case, in the mechanical operation process of the mechanical operation end 904, the shape of the rolled material is changed with a relatively high speed response by controlling the operation amount.

[0014] In this Figure 19 In the case of the structure shown, there is a problem that it is difficult to appropriately control the mechanical operation process of the mechanical operation end 904. For example, when in a state close to the upper limit of the operation range of the mechanical operation end 904 with a high speed response, it is sometimes impossible to stably control the actual shape d2 and the rolling actual result d3, resulting in vibration, and thus the rolling mill 990 as the controlled device cannot operate stably.

[0015] As an existing technique for appropriately performing the shape control of such a Sendzimir rolling mill, there is a technique, for example, as described in Patent Document 3, in which the relationship between the actual shape deviation of the rolled material and the operation amount of the control operation end is learned using machine learning based on the actual result data and control is performed. In the technique described in Patent Document 3, shape control is performed by issuing a control output based on the shape deviation and operating the control operation end.

[0016] The technique described in Patent Document 3 performs shape control by learning the control operation method for the shape deviation and does not consider the position of the control operation end. In particular, in the existing shape control, since control is performed based on the shape deviation, there is a problem that the actual result position of the mechanical operation end associated with operation abnormalities such as plate breakage cannot be restricted.

[0017] In addition, in the description so far, the problems of the shape control of the Sendzimir rolling mill have been described, but in the case where various equipment control devices perform a control operation with a fast response and a control operation with a slow response simultaneously, there are the same problems when appropriately performing both control operations simultaneously.

[0018] Patent Document 1: Japanese Patent Publication No. 2804161

[0019] Patent Document 2: Japanese Patent Publication No. 2515028

[0020] Patent Document 3: Japanese Unexamined Patent Application Publication No. 2018 - 005544 Summary of the Invention

[0021] An object of the present invention is to provide an equipment control device, an equipment control method, and a program. By predicting the occurrence of operation abnormalities in a controlled equipment, the control operation terminal is appropriately operated in a manner that does not cause operation abnormalities, thereby improving the control effect and operation efficiency.

[0022] To solve the above problems, for example, the structure described in the scope of the claimed patent is adopted.

[0023] This application includes multiple means for solving the above problems. If one example is cited, as an equipment control device, it is an equipment control device that performs a first operation process with a predetermined response speed for an operation on a controlled equipment and a second operation process with a response speed slower than the first operation process. Among them,

[0024] This equipment control device includes:

[0025] An operation terminal control unit that acquires a state quantity that is the target of the controlled equipment and gives an instruction for the first operation process;

[0026] A second operation terminal setting unit that gives an instruction for the second operation process;

[0027] A first operation terminal that executes the first operation process of the controlled equipment according to the instruction of the operation terminal control unit;

[0028] A second operation terminal that executes the second operation process of the controlled equipment according to the instruction of the operation terminal setting unit;

[0029] A safe operation range determination unit that determines a safe operation range of the first operation process performed by the first operation terminal based on the performance of the first operation terminal;

[0030] A learning unit that learns the position of the first operation terminal of the first operation process of the controlled equipment in terms of production materials unit and the state quantity that is the target according to the operation performance of the controlled equipment; and

[0031] An operation terminal position prediction unit that uses the learning result of the learning unit to predict the position of the first operation terminal during the production of the next production material,

[0032] The operation terminal setting unit determines the position of the second operation terminal based on the predicted value of the operation range of the position of the first operation terminal during the production of the next production material by the operation terminal position prediction unit and the determination of the safe operation range determination unit, so that the position of the first operation terminal does not deviate from the safe operation range during the operation of the next production material.

[0033] According to the present invention, the operation state in the controlled device can be appropriately controlled, and the actual position of the operation end that causes operation abnormality can be suppressed. As a result, it is possible to expect an improvement in the control accuracy, operation efficiency of the controlled device, and suppression of the occurrence of operation abnormality.

[0034] Subjects, configurations, and effects other than those described above will become clear from the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a block diagram showing a configuration example of a device control device according to an embodiment of the present invention.

[0036] Figure 2 It is a block diagram showing a configuration example when the device control device according to an embodiment of the present invention is applied to a rolling mill.

[0037] Figure 3 It is a structural diagram showing an example of a Sendzimir rolling mill.

[0038] Figure 4 It is a structural diagram showing an example of a rolling equipment of a single-stand rolling mill.

[0039] Figure 5 It is a diagram showing an outline of a mechanical operation end according to an embodiment of the present invention.

[0040] Figure 6 It is a block diagram showing a configuration example of a safety operation range determination unit according to an embodiment of the present invention.

[0041] Figure 7 It is a diagram showing a configuration example of a neural network of a mechanical operation end safety operation range determination unit according to an embodiment of the present invention.

[0042] Figure 8 It is a diagram showing a configuration example of a neural network management table according to an embodiment of the present invention.

[0043] Figure 9 It is a diagram showing a configuration example of a learning database according to an embodiment of the present invention.

[0044] Figure 10 It is a block diagram showing a configuration example of a mechanical operation end position prediction unit according to an embodiment of the present invention.

[0045] Figure 11 It is a diagram showing a configuration example of a neural network of a mechanical operation end position prediction unit according to an embodiment of the present invention.

[0046] Figure 12 It is a diagram showing a configuration example of a neural network management table according to an embodiment of the present invention.

[0047] Figure 13 This is a diagram showing an example of the structure of a learning database according to an embodiment of the present invention.

[0048] Figure 14 This is a diagram showing an example of the structure of a coolant operation end setting unit according to an embodiment of the present invention.

[0049] Figure 15 This is a diagram showing an example of the predicted value of a mechanical operation end according to an embodiment of the present invention.

[0050] Figure 16 This is a diagram showing an example of the determination of a mechanical operation end according to an embodiment of the present invention.

[0051] Figure 17 This is a diagram showing an example of the structure of a coolant operation end control output calculation unit and a coolant operation end control output selection unit according to an embodiment of the present invention.

[0052] Figure 18 This is a block diagram showing an example of the hardware structure when a device control device according to an embodiment of the present invention is constituted by a computer.

[0053] Figure 19 This is a block diagram showing an example of the structure of a control device of an existing rolling mill.

[0054] Explanation of reference numerals

[0055] 11…Shape detection preprocessing unit, 12…Pattern recognition unit, 13…Control arithmetic unit, 14…Shape detector, 50…Control device, 100…Equipment control device (computer), 101…Arithmetic unit, 102…Third control unit, 103…High-speed operation terminal, 104…Low-speed operation terminal, 105…Safety operation range determination unit, 110…Control unit, 111…High-speed operation terminal control unit, 112…Low-speed operation terminal setting unit, 190…Controlled object device, 201…Arithmetic unit, 202…Mechanical operation terminal position suppression control unit, 203…Mechanical operation terminal, 204…Coolant operation terminal, 205…Mechanical operation terminal safety operation range determination unit, 210…Mechanical operation terminal position prediction unit, 211…Shape control unit, 212…Coolant operation terminal setting unit, 220…Control unit, 300…Workpiece to be rolled, 301…Rolling mill, 302…Input side tension reel (input side TR), 303…Output side tension reel (output side TR), 304…Rolling speed control unit, 305…Input side TR control unit, 306…Output side tension reel control unit, 307…Roll gap control unit, 308…Input side tensiometer, 309…Output side tensiometer, 310…Rolling speed setting unit, 311…Input side tension setting unit, 312…Output side tension setting unit, 313…Input side tension control unit, 314…Output side tension control unit, 315…Input side tension current conversion unit, 316…Output side tension current conversion unit, 317…Output side sheet thickness gauge, 318…Output side sheet thickness control unit, 319…Roll gap setting unit, 401…Work roll, 402…First intermediate roll, 403…Second intermediate roll, 404…AS-U roll, 405…Dividing roll, 406…Saddle, 501…Input data generation unit, 502…Neural network, 503…Neural network learning control unit, 504…Neural network selection unit, 505…Supervision data generation unit, 506…Operation abnormality determination unit, 511…Learning data database, 512…Control rule database, 610…Mechanical operation terminal position abnormality area determination unit, 611…Mechanical operation terminal position abnormality area search unit, 612…Input data generation unit, 613…Output data determination unit, 620…Mechanical operation terminal position abnormality suppression control unit, 621…Coolant operation terminal control output arithmetic unit, 622…Coolant operation terminal control output selection unit, 623…Coolant control rule database, 701…Input data generation unit, 702…Neural network, 703…Neural network learning control unit, 704…Neural network selection unit, 705…Supervision data generation unit, 706…Manufacturing unit data generation unit, 711…Learning data database, 712…Control rule database, 901…Arithmetic unit, 902…First shape control unit, 903…Second shape control unit, 904…Mechanical operation terminal, 905…Coolant operation terminal, 990…Rolling mill, d11…First state variable target, d12…First state variable, d13…Second state variable,d14…Safety operation range, d21…Target shape, d22…Actual position of mechanical operation end, d23…Actual shape, d24…Rolling actual performance, d25…Safety operation range of mechanical operation end, d26…Position change amount of mechanical operation end, d27…Actual position of coolant operation end, d28…Predicted position of coolant operation end, d30…Coolant operation output, d31…Estimated position, d32…Operation abnormality determination value, d33…Neural network for control, d42…Abnormality suppression output, d51…Determination value for operation end in actual position abnormality area, d52…Operation abnormality evaluation value., Detailed implementation manner

[0056] Hereinafter, with reference to Figures 1 to 18 An equipment control device according to an embodiment of the present invention (hereinafter referred to as "this example") will be described.

[0057] [Overall structure of equipment control device]

[0058] Figure 1 An example showing the overall structure of the equipment control device of this example.

[0059] Figure 1 The shown equipment control device controls the controlled equipment 190, and as the control of the controlled equipment 190, it executes operation processing based on the high-speed operation end (first operation end) 103 and operation processing based on the low-speed operation end (second operation end) 104.

[0060] The equipment control device of this example obtains the first state quantity target d11, and obtains the difference from the first state quantity d12 through the arithmetic unit 101. As a result of the operation processing based on the high-speed operation end 103 and the low-speed operation end 104, the first state quantity d12 is obtained from the controlled equipment 190. In addition, the high-speed operation end 103 and the low-speed operation end 104 obtain the second state quantity d13 as a result of the operation processing of other operation ends applied to the controlled equipment 190.

[0061] The control unit 110 of the equipment control device has a high-speed operation end control unit 111 and a low-speed operation end setting unit 112. The high-speed operation end control unit 111 controls the operation processing of the high-speed operation end 103. The low-speed operation end setting unit 112 controls the operation processing of the low-speed operation end 104.

[0062] The difference between the first state quantity target d11 obtained by the arithmetic unit 101 and the first state quantity d12 is supplied to the high-speed operation end control unit 111, and control is performed to make the first state quantity d12 close to the difference.

[0063] The control output of the high-speed operation end control unit 111 is directly supplied to the high-speed operation end 103 to control the operation processing of the high-speed operation end 103.

[0064] In addition, the equipment control device of this example includes a safe operation range determination unit 105. The safe operation range determination unit 105 obtains the operation performance of the high-speed operation end 103, determines whether the obtained operation performance has a margin within the safe operation range, and provides the data of the safe operation range d14 as the determination result to the low-speed operation end setting unit 112.

[0065] Moreover, the safe operation range determination unit 105 obtains the first state quantity d12 and the second state quantity d13 of the controlled equipment 190. Then, the safe operation range determination unit 105 refers to the first state quantity d12 and the second state quantity d13, and determines whether the current operation performance of the high-speed operation end 103 has a margin within the safe operation range.

[0066] Here, the safe operation range determination unit 105 detects the occurrence of operation abnormalities, and learns the first state quantity d12 and the second state quantity d13 at that time, and the actual position of the high-speed operation end 103, so as to determine the operable range, that is, the safe operation range d14, in such a way that no operation abnormalities occur at the high-speed operation end 103. Since operation abnormalities do not occur frequently, the safe operation range determination unit 105 preferably continuously collects actual performance data, performs learning using machine learning, etc., and obtains the safe operation range d14.

[0067] Then, the safe operation range determination unit 105 supplies the data of the determined safe operation range d14 to the low-speed operation end setting unit 112.

[0068] The operation end position prediction unit 102 learns the movement range of the high-speed operation end 103 in units of production materials based on past operation performance data, and based on the result, predicts the movement amount of the high-speed operation end 103 when the next production material is produced.

[0069] The low-speed operation end setting unit 112 determines the setting value of the low-speed operation end 104 based on the movement amount of the high-speed operation end 103 predicted by the operation end position prediction unit 102 when the next production material is produced, so that the high-speed operation end 103 is within the safe operation range d14, and sets the low-speed operation end 104. The setting of the low-speed operation end 104 is performed mechanically according to the instruction of the low-speed operation end setting unit 112 if automatic setting is possible, and in the case where automatic setting is not possible, the operator is notified by methods such as displaying the setting value of the low-speed operation end 104 on the screen, and the operator performs the setting manually.

[0070] According to Figure 1 The equipment control device with the structure shown can enable the high-speed operation end 103 to operate efficiently within the range where no operation abnormalities occur. In addition to improving the operation efficiency, an improvement in control accuracy can also be expected.

[0071] [Overall Structure When Applied to the Control Device of Sendzimir Mill]

[0072] Next, the overall structure in the case where the equipment control device of this example is applied to a Sendzimir mill will be described.

[0073] Figure 2 It shows the structure of the equipment control device of this example in the case of being applied to a Sendzimir mill.

[0074] Figure 2 The shown equipment control device obtains the target shape d21 of the material to be rolled, and obtains the difference from the actual shape d23 after rolling through the arithmetic unit 201.

[0075] The equipment control device of this example, as the control of the rolling mill 301, executes the operation process based on the mechanical operation end 203 and the setting operation process based on the coolant operation end 204. The operation process based on the mechanical operation end 203 mechanically changes the roll gap for rolling process, etc., and the response manifested in the actual shape d23 of the material to be rolled is high speed. On the other hand, the setting operation process based on the coolant operation end 204 changes the coolant injection amount, and the response manifested in the rolling actual result d24 of the material to be rolled is slower than the operation based on the mechanical operation end 203.

[0076] The control unit 220 of the equipment control device has a first shape control unit 211 that controls the operation process of the mechanical operation end 203 and a coolant operation end setting unit 212 that sets the coolant operation end 204.

[0077] The first shape control unit 211 performs feedback control to make the material to be rolled into the target shape d21. The target shape d21 is preset according to the characteristics of the material to be rolled, etc.

[0078] The coolant operation end setting unit 212 implements the setting of the coolant operation end 204. Here, implementing the setting means that before the manufacturing operation of each production material unit, which is the manufacturing unit of the material to be rolled (production material) manufactured by the rolling mill, the flow distribution in the plate width direction of the coolant operation end 204 is determined in advance. During the manufacturing operation of the material to be rolled, the change of the flow distribution in the plate width direction is not performed. The manufacturing unit of the material to be rolled refers to one rolling operation, and is also called the production material unit. If described in more detail, in the Sendzimir mill, the following operations are performed: while changing the rolling direction, multiple rolling operations with different output side plate thicknesses (decreasing the output side plate thickness in sequence) are performed on the same material to be rolled, and the target product plate thickness is obtained. Each one of the multiple rolling operations is the above-mentioned manufacturing unit or production material unit.

[0079] The control output of the first shape control unit 211 is directly supplied to the mechanical operation end 203 to control the operation process of the mechanical operation end 203.

[0080] The coolant operation end setting unit 212 performs the setting of the coolant operation end 204. However, in the case where direct operation from the control unit 220 of the equipment control device cannot be performed, a setting guide for the coolant operation end 204 is displayed to the operator, and the operator performs the setting operation. Here, the case where direct operation from the control unit 220 cannot be performed means, for example, when the control unit 220 is implemented by a computer, the case where the computer cannot directly operate the coolant operation end 204 via the input / output unit.

[0081] In addition, the equipment control device of this example includes a mechanical operation end safe operation range determination unit 205.

[0082] The mechanical operation end safe operation range determination unit 205 obtains the operation performance of the mechanical operation end 203, that is, the mechanical operation end position performance d22, and determines whether the obtained mechanical operation end position performance d22 has a margin within the safe operation range. And the mechanical operation end safe operation range determination unit 205 supplies the data of the mechanical operation end safe operation range d25 as the determination result to the coolant operation end setting unit 212.

[0083] Furthermore, the mechanical operation end safe operation range determination unit 205 obtains the shape performance d23 and the rolling performance d24 of the rolling mill 301. Then, the mechanical operation end safe operation range determination unit 205 refers to the shape performance d23 and the rolling performance d24 and determines whether the obtained mechanical operation end position performance d22 has a margin within the safe operation range.

[0084] The mechanical operation end safe operation range determination unit 205 detects the occurrence of an operation abnormality and performs the processing as a learning unit that learns the shape performance d23, the rolling performance d24, and the mechanical operation end position performance d22 at this time point. Through this learning, the mechanical operation end safe operation range determination unit 205 determines the operable range, that is, the mechanical operation end safe operation range d25, in such a way that no operation abnormality occurs at the mechanical operation end 203. Here, the mechanical operation end safe operation range determination unit 205 continuously collects performance data, performs learning using machine learning, etc., and obtains the mechanical operation end safe operation range d25.

[0085] Then, the mechanical operation end safe operation range determination unit 205 supplies the data of the determined mechanical operation end safe operation range d25 to the coolant operation end setting unit 212.

[0086] In addition, regarding the detailed structure of the mechanical operation end safe operation range determination unit 205 for learning operation abnormalities, etc., will be described in Figure 6As described later.

[0087] The mechanical operation end position prediction unit 210 learns the mechanical operation end position variation d26 for each manufacturing unit of the rolled material based on the rolling performance d24, the mechanical operation end position performance d22, and the coolant operation end position performance d27. The mechanical operation end position variation d26 is a quantity obtained by classifying the manufacturing units according to the product specifications and learning within which range the performance value of the mechanical operation end 203 varies for each manufacturing unit.

[0088] Here, classifying according to the product specifications means presetting criteria in such a way that the performance values of the mechanical operation end 203 are approximately the same based on the steel type, plate width, plate thickness, etc. of the rolled material, and classifying according to this criterion.

[0089] If classified as the same product specifications, the movement range of the performance value of the mechanical operation end 203 is the same. Therefore, before starting the manufacturing operation for each manufacturing unit of the rolled material, the product specifications of the manufacturing unit starting the manufacturing operation are determined. Thus, the mechanical operation end position prediction unit 210 can obtain the learning result of the mechanical operation end position variation d26 for the classification with the same product specifications. At the same time, the mechanical operation end position prediction unit 210 can also predict the position of the coolant operation end 204 in the case of the mechanical operation end position variation d25, and output the coolant operation end position prediction d28 to the coolant operation end setting unit 212.

[0090] The coolant operation end setting unit 212 obtains the mechanical operation end position variation d26 of the next manufacturing unit and the coolant operation end position prediction d28 at this time from the mechanical operation end position prediction unit 210. Moreover, the coolant operation end setting unit 212 compares the mechanical operation end safety operation range d25 obtained from the mechanical operation end safety operation range determination unit 205 with the mechanical operation end position variation d26. Thus, the coolant operation end setting unit 212 determines whether the position performance of the mechanical operation end 203 is within the mechanical operation end safety operation range d25 when the manufacturing operation of the next manufacturing unit is carried out.

[0091] When the result of this determination is that it is within the mechanical operation end safety operation range d25, the coolant operation end setting unit 212 outputs a coolant setting instruction to the coolant operation end 204 to set the coolant operation end position prediction d28 as the coolant flow rate.

[0092] In addition, when the determination result indicates that the mechanical operation end has left the safe operation range d25, the coolant operation end setting unit 212 outputs a coolant setting instruction to the coolant operation end 204, which corrects the predicted coolant operation end position d28 by setting the coolant flow rate in the plate width direction of the mechanical operation end safe operation range d25 based on the actual position of the mechanical operation end 203.

[0093] As the output instruction to the coolant operation end 204 in this case, in addition to directly outputting from the coolant operation end setting unit 212 to the coolant operation end 204, it also includes the case where the set value of the coolant operation end 204 is displayed to the operator, and the operator performs operations to actually change the coolant flow rate.

[0094] [Structure of Sendzimir Mill]

[0095] Here, an example of the structure of a Sendzimir mill will be described.

[0096] Figure 3 It shows a schematic structure during shape control in a Sendzimir mill.

[0097] The actual shape of the rolled material after rolling is detected by the shape detector 14 in the Sendzimir mill.

[0098] The actual shape detected by the shape detector 14, after undergoing preprocessing of pattern recognition by the shape detection preprocessing unit 11 of the control device 10, is calculated by the pattern recognition unit 12 to determine which one of the pre-set reference shape patterns it is closest to. Then, based on the calculated reference shape pattern, the control operation unit 13 determines the operation end and the operation amount to be operated, and executes the process of controlling the Sendzimir mill with the operation end and the operation amount to be operated accordingly.

[0099] Figure 4 It shows an example of a rolling equipment for a single-stand mill. The Sendzimir mill is a type of single-stand mill.

[0100] Figure 4 The shown rolling equipment consists of a mill 301, an input side tension reel (hereinafter referred to as "TR") 302, and an output side TR 303. The rolled material 300 drawn from the input side TR 302 is wound by the output side TR 303 after passing through the mill 301.

[0101] The mill 301 rolls the rolled material 300. The rolling here is a process of thinning the plate thickness of the rolled material 300 to a predetermined plate thickness.

[0102] In the rolling mill 301, there are provided a rolling speed control unit 304 for adjusting the rolling speed and a roll gap control unit 307 for adjusting the roll gap of the rolling mill 301. In addition, an input side TR control unit 305 and an output side TR control unit 306 for adjusting the generated tension are provided in the input side TR 302 and the output side TR 303, respectively.

[0103] By adjusting the upper and lower roll intervals of the rolling mill 301 using the roll gap control unit 307 to apply a pressure for flattening the material to be rolled 300, and feeding the material to be rolled 300 to the output side by the rolling speed control unit 304, the rolling process is carried out. At this time, the process of applying tension to the material to be rolled 300 using the input side TR 302 and the output side TR 303 is also performed on the input side and the output side of the rolling mill 301.

[0104] For the rolling operation, it is important that the thickness of the material to be rolled 300 that becomes the product (the thickness of the output side plate of the rolling mill). The roll gap and the input side tension and the output side tension are preset so that the material to be rolled 300 becomes a predetermined thickness.

[0105] The input side tension current conversion unit 315 uses the input side tension set by the input side tension setting unit 311 to obtain the current required to obtain the set input side tension, and gives it to the input side TR 302 via the input side TR control unit 305, thereby obtaining the input side tension.

[0106] Similarly, the output side tension current conversion unit 316 uses the output side tension set by the output side tension setting unit 312 to obtain the current required to obtain the set output side tension, and gives it to the output side TR 303 via the output side TR control unit 306, thereby obtaining the output side tension.

[0107] The roll gap set by the roll gap setting unit 319 is provided to the roll gap control unit 307, and the roll gap is set by the roll gap control unit 307.

[0108] The rolling speed setting unit 310 determines the speed of the rolling mill 301 according to the instructions of the operator of the rolling mill, and sets the speed of the rolling mill 301 through the rolling speed control unit 304.

[0109] On the input side and the output side of the rolling mill 301, an input side tension meter 308 and an output side tension meter 309 are provided, and the input side tension control unit 313 and the output side tension control unit 314 perform control so that the actual tension measured by them is consistent with the set tension. In addition, an output side plate thickness meter 317 is provided on the output side of the rolling mill 301, and the output side plate thickness control unit 318 performs control so that the actual plate thickness measured there is consistent with the set plate thickness.

[0110] Based on the above structure, as already describedFigure 3 As shown, a shape detector 14 for detecting the shape of the material being rolled is provided on the output side of the rolling mill, and shape control is performed in such a way that the detected shape conforms to a preset target shape.

[0111] As described above, the shape is the degree of undulation of the metal plate of the material being rolled. Therefore, according to the workability in the next process of the rolling mill and the efficiency of the rolling operation in the rolling mill, the target shape, that is, the target shape, is preset. Generally, since tension is applied to the material being rolled, if there are damages such as cracks at the plate end, it is likely that cracks will occur from here and the material being rolled will break in the plate width direction (plate fracture). Therefore, in order not to concentrate the tension, the plate end is mostly in an undulated state.

[0112] The undulation of the material being rolled actually applies tension to the material being rolled. Therefore, although it is not manifested and there is no undulation in appearance, the tension distribution changes in the plate width direction.

[0113] Here, Figure 3 The shown shape detector 14 estimates the undulation of the plate by measuring the tension distribution in the plate width direction and detects it as the actual shape.

[0114] [Structure and Processing of the Mechanical Operation End of the Shape Control Machine]

[0115] Figure 5 (a) in shows the structure when operating and processing through the mechanical operation end 203 of the Sendzimir rolling mill. In Figure 5 shows the cross-section in the plate width direction of the material being rolled 300, and only the structure on the upper side of the material being rolled 300 is shown, and the structure on the lower side is omitted.

[0116] In addition, Figure 5 (b) and (c) in respectively show the action waveforms when changing the shape of the material being rolled 300.

[0117] As Figure 5 shown in (a) in, the Sendzimir rolling mill is composed of a work roll 401, a first intermediate roll 402, a second intermediate roll 403, and an AS-U roll 404 with the material being rolled 300 in between.

[0118] The first intermediate roll 402 is provided with cones on opposite upper and lower sides, and by moving in the plate width direction, it can affect the shape of the plate end of the material being rolled 300.

[0119] The AS-U roll 404 has a structure in which a saddle 406 enters between a plurality of split rolls 405, and by changing the position of the saddle 406 ( Figure 5 the longitudinal position), the deflection of the AS-U roll 404 can be changed in the plate width direction.

[0120] For example, as shown in (b) of Figure 5 , when the saddle 406 at the center is lowered, it can affect the shape of the central part of the material to be rolled 300.

[0121] Here, Figure 5 The operation waveforms shown in the lowermost rows of (b) and (c) of represent the changes in the plate thickness distribution of the material to be rolled 300 when the saddle 406 or the first intermediate roll 402 is moved. The shape change is opposite to the plate thickness distribution.

[0122] The shape is the distribution of the undulation degree in the plate width direction. A large undulation means that the material to be rolled 300 elongates. This is because "the plate thickness of the output side plate becomes thinner", "the elongation of the material to be rolled in the thinner part is large", and "the shape of the material to be rolled becomes larger" are equivalent.

[0123] Regarding operation abnormalities, the breakage of the plate of the material to be rolled 300 is a major problem. If the plate breaks, the broken material to be rolled 300 will damage the work roll 401 and the first intermediate roll 402 of the rolling mill. In addition, depending on the situation, the second intermediate roll 403 and the AS-U roll 404 may also be damaged when the plate breaks. If these damages occur, these rolls need to be replaced, and the removal process of the material to be rolled 300 remaining in the rolling mill takes time, and the operation efficiency is extremely reduced.

[0124] The AS-U roll 404 presses the splitting roll 405 against the second intermediate roll 403 in a form pressed by the saddle 406. Therefore, depending on the position of the saddle 406, the splitting roll 405 may not contact the second intermediate roll 403 sometimes. If such a state occurs, the force applied to the material to be rolled 300 from the work roll 401 at this part decreases sharply, the material to be rolled 300 no longer elongates, and the tension applied to the material to be rolled 300 at this part increases.

[0125] When such a state occurs at the plate end of the material to be rolled 300, the plate breaks from the plate end. In addition, due to the change in the tension at both ends in the plate width direction of the material to be rolled 300, a phenomenon occurs in which the center of the plate width of the material to be rolled 300 deviates from the center of the plate width of the rolling mill, and sometimes it may collide with the mechanical equipment before and after the rolling mill and cause the plate to break. In this way, depending on the actual position of the mechanical operation end 203, operation abnormalities may sometimes occur.

[0126] The actual position where operation abnormalities occur is not calculated based on the mechanical structure, but changes according to the plate thickness distribution in the plate width direction of the material to be rolled 300, the plate thickness of the input and output side plates, the tension, the rolling load and other rolling states, and the positional relationship with other shape control mechanical operation ends. Therefore, it is difficult to predict in advance.

[0127] Therefore, in this example, the mechanical operation end safety operation range determination unit 205 saves these conditions during rolling abnormalities as actual performance data, compares them with the actual performance data during normal times, and thereby obtains the actual performance position of the shape control mechanical operation end where rolling abnormalities are likely to occur.

[0128] The mechanical operation end safety operation range determination unit 205 in this example uses machine learning to determine the mechanical operation end safety operation range. Among the actual performance data during machine learning, the rolling states such as the input / output side plate thickness, tension, and rolling load, and the actual performance position of the mechanical operation end 203 are used, and the rolling abnormality occurrence information is used in the supervised data.

[0129] As the rolling abnormality occurrence information, the information on plate breakage and the emergency stop of the rolling mill is used. Plate breakage can be determined by a decrease in the input / output side tension. In the case of an emergency stop when some abnormality occurs in the rolling state and the operation is stopped, the information of the operation switch operated by the operator is used. The information of the operation switch can be detected by the computer constituting the control device of the rolling mill and can be used as one of the actual performance information. The mechanical operation end safety operation range determination unit 205 uses these actual performance data and supervised data to generate a neural network (N.N.) for determining whether a job abnormality has occurred.

[0130] [Structure of the mechanical operation end safety operation range determination unit and structure of the neural network]

[0131] Figure 6 The structure showing the case where the mechanical operation end safety operation range determination unit 205 is implemented by machine learning.

[0132] In addition, Figure 7 The structure of the neural network 502 included in the mechanical operation end safety operation range determination unit 205 is shown.

[0133] As Figure 7 shown, the neural network 502 receives the rolling actual performance d24 and the mechanical operation end position actual performance d22 from the input data generation unit 501 at the input end 502a, and outputs the job abnormality determination value d32 from the output end 502b. The job abnormality determination value d32 is the information on plate breakage and the information on emergency stop as the rolling abnormality occurrence information. The neural network 502 performs learning based on the combination of these input data and output data.

[0134] If Figure 6Regarding the mechanical operation end safety operation range determination unit 205 shown, the input data generation unit 501 collects the actual mechanical operation end position d22 and the actual shape d23. In addition, the supervision data generation unit 505 collects the operation abnormality determination value d32 determined by the operation abnormality determination unit 506. The data collection in these input data generation unit 501 and supervision data generation unit 505 is performed at a constant time period under the control of the neural network learning control unit 503, and a set of learning data is obtained for each operation cycle. The obtained learning data is sequentially stored in the learning data database 511.

[0135] The operation abnormality determination unit 506 determines whether there is a plate breakage as an operation abnormality and an emergency stop of the rolling mill based on the rolling actual result d24. The operation abnormality determination value d32 as the determination result is information on the plate breakage and the emergency stop.

[0136] However, the rolling mill rolls various rolled materials 300 according to the specifications to obtain products. Therefore, the rolling mill usually changes the specifications of the work roll 401 (diameter distribution in the plate width direction), the tapered specifications of the first intermediate roll 402, and the combination of the split rolls 405 of the AS-U roll 404 as the mechanical structure to cope with the rolled material 300. In addition, for the rolled material 300, the plate width and the material are also different. Therefore, the neural network 502 can be divided according to the mechanical structure and the specifications of the rolled material 300 to enable efficient learning.

[0137] Therefore, the mechanical operation end safety operation range determination unit 205 in this example is provided with a control rule database 512 and a neural network selection unit 504 so as to be able to have multiple types of neural networks 502 and switch between them for use.

[0138] Figure 8 Shows a structural example of the control rule database 512.

[0139] As Figure 8 shown in (a) in, a plurality of neural networks that have been learned using the learning data composed of the combination of input data and supervision data are stored in the control rule database 512.

[0140] Then, the neural network learning control unit 503 designates the neural network number that needs to be learned. The neural network selection unit 504 receives the designation of the neural network number required for learning by the neural network learning control unit 503, retrieves the neural network from the control rule database 512, and sets it as the neural network 502.

[0141] The neural network selection unit 504 takes out the neural network with the corresponding neural network number from the control rule database 512 according to the current rolling performance d24 combined with the rolling conditions and the mechanical structure, and sets it in the mechanical operating end position suppression control unit 202 as the control neural network d33.

[0142] Figure 8 (b) in the figure represents the structure of the neural network management table stored in the control rule database 512. The management table is divided according to the (B1) plate width, the (B2) steel type, and the mechanical structure (A). As the (B1) plate width, for example, four categories of 3-foot width, meter width, 4-foot width, and 5-foot width are used. As the (B2) steel type, about 10 categories of steel types (1) to (10) are used. For (A), for example, according to the length of the tapered portion of the tapered specification of the first intermediate roller 402, it is divided into (A1) and (A2).

[0143] The above table differences are only examples and need to be set appropriately according to the rolling equipment and the types of rolled materials produced.

[0144] The machine operating end safe operation range determination unit 205 uses these neural networks separately according to the rolling conditions and the machine structure.

[0145] The neural network learning control unit 503 follows Figure 8 The neural network management table shown in (b) will Figure 8 The combination of input data and supervisory data shown in (a), i.e., learning data, is stored in the learning data database 511 in association with the corresponding neural network number.

[0146] Figure 9 This shows an example of the learning data stored in the learning data database 511.

[0147] like Figure 9 As shown, the learning data database 511 stores the learning data corresponding to each neural network number.

[0148] The neural network learning control unit 503 instructs the input data generation unit 501 and the supervisory data generation unit 505 to fetch the input data and supervisory data corresponding to the neural network from the management table from the learning data database 511. The neural network 502 performs learning using these. Various neural network learning methods have been proposed in the past, and any learning method can be used.

[0149] A large number of sets of learning data are required in machine learning, and if a certain amount (eg, 10,000 sets) are stored in the learning data database 511, the neural network 502 performs learning.

[0150] When the learning of the neural network 502 is completed, the neural network learning control unit 503 writes back the neural network 502 as the learning result to the position of the neural network number in the control rule database 512, thereby completing the learning.

[0151] The learned neural network 502 outputs an operation abnormality determination value by inputting the rolling actual results d24 and the mechanical operation end position actual results d22. Therefore, the neural network 502 can predict the occurrence of operation abnormalities and search for the safe operation range d25 of the mechanical operation end by providing the predicted future shape actual results d23 and the mechanical operation end position actual results d22.

[0152] [Structure of the mechanical operation end position prediction unit and structure of the neural network]

[0153] Figure 10 Shows the structure when the mechanical operation end position prediction unit 210 is implemented by machine learning.

[0154] In addition, Figure 11 Shows the structure of the neural network 702 included in the mechanical operation end position prediction unit 210.

[0155] As Figure 11 shown, the neural network 702 obtains the rolling actual results d24 from the input data generation unit 701 at the input end 702a, and outputs the coolant operation end position prediction d28 and the mechanical operation end position change amount d26 from the output end 702b. The neural network 702 performs learning based on the combination of these input data and output data.

[0156] If the mechanical operation end position prediction unit 710 shown in Figure 10 is described, the input data generation unit 701 collects the rolling actual results d24. In addition, the supervised data generation unit 705 collects the mechanical operation end position change amount d26 and the coolant operation end position prediction d28 for each manufacturing unit generated by the manufacturing unit data generation unit 706.

[0157] The manufacturing unit data generation unit 706 determines the manufacturing unit of the material to be rolled based on the rolling speed of the rolling actual results d24. As described above, the manufacturing unit of the material to be rolled refers to the first rolling operation. Therefore, the first rolling operation (manufacturing of the manufacturing unit) of the material to be rolled can be determined to change from the rolling speed = 0 to a state where the rolling speed is not 0, and then until the rolling speed = 0.

[0158] Therefore, the manufacturing unit data generation unit 706 collects the maximum and minimum values of the mechanical operation end position actual value d22 and the coolant operation end position actual value d27 during the period until the rolling speed becomes 0, and generates the mechanical operation end position change amount d26 and the coolant operation end position prediction d28.

[0159] After the first rolling operation (manufacturing unit) of the material to be rolled is completed and the generation of the mechanical operation end position change amount d26 and the coolant operation end position prediction d28 is completed, one piece of learning data required for the learning of the neural network 702 is generated. Therefore, the manufacturing unit data generation unit 706 notifies the neural network learning control unit 703. The neural network learning control unit 703 receives the notification from the manufacturing unit data generation unit 706 and performs data acquisition in the input data generation unit 701 and the supervised data generation unit 705. The obtained learning data is sequentially stored in the learning data database 711.

[0160] However, the rolling mill rolls various materials to be rolled 300 according to specifications to obtain products. Therefore, the rolling mill usually changes the specifications of the work roll 401 (diameter distribution in the plate width direction), the taper specifications of the first intermediate roll 402, and the combination of the split roll 405 of the AS-U roll 404 as the mechanical structure to cope with the material to be rolled 300. In addition, for the material to be rolled 300, the plate width and material are also different. Therefore, the neural network 702 can be efficiently learned according to the mechanical structure and the specifications of the material to be rolled 300.

[0161] Therefore, the mechanical operation end position prediction unit 201 of this example includes a control rule database 712 and a neural network selection unit 704 so as to have multiple types of neural networks 702 and switch between them for use.

[0162] Figure 12 Shows a structural example of the control rule database 712.

[0163] As Figure 8 shown in (a) of, multiple neural networks that have been learned using learning data composed of a combination of input data and supervised data are stored in the control rule database 712.

[0164] Moreover, the neural network learning control unit 703 designates the neural network number that needs to be learned. The neural network selection unit 704 receives the designation of the neural network number required for learning by the neural network learning control unit 703, retrieves the neural network from the control rule database 712, and sets it as the neural network 702.

[0165] The neural network selection unit 704 retrieves the neural network with the corresponding neural network number from the control rule database 712 based on the current rolling performance d24 in combination with the rolling conditions and the mechanical structure, and sets it as the control neural network d33.

[0166] The rolling performance d24 from the input data generation unit 701 is input to the control neural network d33, and the mechanical operation end position variation d26 and the coolant operation end position prediction d28 are output to the coolant operation end setting unit 212. At this time, since the rolling performance d24 contains the manufacturing information of the next manufacturing unit of the rolled material such as the set values of plate thickness, steel type, tension, load, etc., the mechanical operation end position variation d26 and the coolant operation end position prediction d28 are estimated based on this manufacturing information.

[0167] Then, the mechanical operation end position variation d26 and the coolant operation end position prediction d28 obtained by the control neural network d33 are output to the coolant operation end setting unit 212.

[0168] Figure 12 (b) in represents the structure of the neural network management table stored in the control rule database 712. The management table is divided according to (B1) plate width, (B2) steel type, and mechanical structure (A). As (B1) plate width, for example, 4 classifications such as 3-foot width, meter width, 4-foot width, and 5-foot width are used. As (B2) steel type, about 10 classifications from steel type (1) to steel type (10) are used. For (A), for example, it is divided into (A1) and (A2) according to the length of the tapered portion which is the tapered specification of the first intermediate roll 402.

[0169] The above table differences are only examples, and it is necessary to set them appropriately according to the rolling equipment and the types of rolled materials produced.

[0170] The mechanical operation end position prediction unit 210 uses these neural networks separately according to the rolling conditions and the mechanical structure.

[0171] The neural network learning control unit 703 stores the learning data, which is the combination of the input data and the supervised data shown in (a) in, associated with the corresponding neural network number in the learning data database 711 according to the neural network management table shown in (b) in. Figure 12 Figure 12

[0172] Figure 13

[0173] Figure 13

[0174] As Figure 13 shown, the learning data database 711 stores the learning data corresponding to each neural network number.

[0174] The neural network learning control unit 703 instructs the input data generation unit 701 and the supervised data generation unit 705 to retrieve the input data and the supervised data corresponding to the neural network from the management table from the learning data database 711. The neural network 702 uses them to perform learning. Various neural network learning methods have been proposed in the past, and any learning method can be used.

[0175] In machine learning, a large number of sets of learning data are required. If a certain degree (for example, 10,000 sets) is stored in the learning data database 711, the neural network 702 performs learning.

[0176] When the learning of the neural network 702 is completed, the neural network learning control unit 703 writes back the neural network 702 as the learning result to the position of the neural network number in the control rule database 712, thereby completing the learning.

[0177] If the rolling actual result d24 is input from the input data generation unit 701, the learned neural network 702 outputs the mechanical operation end position change amount d26 and the coolant operation end position prediction d28 to the neural network selection unit 704. Therefore, the neural network 702 can predict the change range of the mechanical operation end position and the distribution of the coolant injection amount in the plate width direction at this time based on the past actual result data when manufacturing the same or similar manufacturing units as the next production.

[0178] [Structure of coolant operation end setting unit]

[0179] Figure 14 Represents the structure of the coolant operation end setting unit 212.

[0180] The coolant operation end setting unit 212 includes a mechanical operation end position abnormal area determination unit 610 and a mechanical operation end position abnormal suppression control unit 620.

[0181] The mechanical operation end position abnormal area determination unit 610 uses the neural network 502 described in Figure 7 to estimate the mechanical operation end 203 where the occurrence of prediction operation abnormality occurs. The neural network 502 used here is the control neural network d33 received by the mechanical operation end safe operation range determination unit 205 ( Figure 6 ).

[0182] The mechanical operation end position abnormal suppression control unit 620 generates a setting instruction for the coolant operation end 204 based on the determination result in the mechanical operation end position abnormal area determination unit 610.

[0183] The mechanical operation end position prediction unit 210 ( Figure 2)During the rolling operation of the next manufacturing unit, i.e., the rolled material, predict how much the actual position of the mechanical operation end 203 will change, and transmit it as the mechanical operation end position change amount d26 to the coolant operation end setting unit 212.

[0184] Here, the mechanical operation end position change amount d26 is due to the difference between the maximum value and the minimum value of each mechanical operation end 203 within the manufacturing unit, so it becomes the content described below. Figure 15 as shown.

[0185] Regarding whether the combinations of various positions within the operating range of this mechanical operation end will not cause operation abnormalities, it can be learned by dividing the operating range into several parts and inputting them into the neural network d33 that can determine operation abnormalities learned in the mechanical operation end safety range determination unit 205, and judging the possibility of operation abnormality occurrence. However, since there are multiple mechanical operation ends 203 (7 types in the case of this embodiment), the number of cases to be determined is extremely large and not practical.

[0186] The actual position of the mechanical operation end that causes operation abnormality is considered to be near the upper limit or the lower limit of the movement of the mechanical operation end 203, rather than the central part of the movement range. Therefore, the mechanical operation end position prediction unit 210 can simply implement the determination of the safe operation range by obtaining the combination of the maximum value and the minimum value of each mechanical operation end 203 and inputting it into the control neural network d33 (refer to Figure 10 ).

[0187] Figure 15 Shows an example of the mechanical operation end position change amount. Figure 15 The horizontal axis of represents the type of the mechanical operation end 203, and the vertical axis represents the position prediction value.

[0188] In Figure 15 's example, when there are n types (n is an integer) of the mechanical operation end 203, the maximum value and the minimum value of the mechanical operation end position change amount d26 are:

[0189] POSMAX(k), POSMIN(k), k = 1, 2,..., n.

[0190] For example, Figure 15 the movement range of the manufacturing unit of the mechanical operation end (3) shown becomes the range shown by the maximum value POSMAX(3) and the minimum value POSMIN(3).

[0191] Here, the mechanical operation end 203 is n types, for example, the total value of the number of saddles 406 equivalent to the AS-U roll 404 and the number of the first intermediate roll 402 that can move in the plate width direction. For example, in Figure 5In the case of the example shown, the number of saddles is 5 and the number of the first intermediate rolls is 2 (upper and lower), so n = 7. In the following description, when it is expressed as the mechanical operation end 203(k), each of the n types (1 to n) of the mechanical operation end 203 is represented.

[0192] Thus, the actual results of the estimated positions of the respective mechanical operation ends 203 can be generated in 2 n types. For example, when n = 7, 128 types of actual results of the estimated positions can be generated. The actual results of the estimated positions are sequentially output to the input data generation unit 612 (see Figure 7 ).

[0193] The input data generation unit 612 generates input data for the neural network 502 from the rolling actual results d24 and the estimated position d31, and outputs the input data to the neural network 502.

[0194] The neural network 502 outputs Figure 7 the operation abnormality determination value d32 shown. The operation abnormality determination value d32 is the degree of plate breakage and emergency stop. Here, receiving the operation abnormality determination value d32 output from the neural network 502, the output data determination unit 613 weights and adds the degrees of both, and uses the result as the operation abnormality evaluation value d26. Usually, in the case of an operation abnormality, the operator performs an emergency stop, but the operator also performs an emergency stop when there are signs of plate breakage. Here, as a sign of plate breakage, for example, the case where the material to be rolled 300 snakes is considered.

[0195] In the case of plate breakage occurring without an emergency stop, since it occurs without signs of plate breakage, in such a case, the priority of the method for suppressing plate breakage becomes higher. Therefore, the weight of the degree of plate breakage is increased.

[0196] Figure 14 The mechanical operation end position abnormality region search unit 611 shown stores in advance the output estimated position d31 and the returned operation abnormality evaluation value d26, and searches for the estimated value at which the operation abnormality evaluation value d26 becomes the maximum. As a result of the search, when the maximum value of the operation abnormality evaluation value d26 exceeds a preset threshold value, the mechanical operation end position abnormality region search unit 611 stores a combination of either the maximum value POSMAX(i) or the minimum value POSMIN(i) of the mechanical operation end position variation amount d26 at the estimated position d31 in this case asPOSEST(i). Then, in order to confirm the increase or decrease of the operation abnormality evaluation value d26 by correctingPOSEST(i), onlyPOSEST(i) is increased or decreased by a predetermined ΔPOS(i).

[0197] That is, the following operations are performed.

[0198] POSEST(i)+ = POSEST(i) + ΔPOS(i)

[0199] POSEST(i)- = POSEST(i) - ΔPOS(i)

[0200] POSEST(i)0 = POSEST(i) + 0 (ΔPOS(i) = 0)

[0201] Then, the above three mechanical operation end position estimation values of each mechanical operation end 203 are combined with each mechanical operation end 203 again to set the estimated position d31, and output to the input data generation unit 612. The input data generation unit 612 generates input data for the neural network 502 from the rolling actual result d24 and the estimated position d31, and outputs it to the neural network 502.

[0202] Here, the mechanical operation end position abnormal area search unit 611 searches for a combination of estimated positions d31 where the operation abnormal evaluation value d26 is the minimum (the degree of abnormality is the minimum), and outputs the result as the actual result position abnormal area operation end determination value d51 (ΔPOS(i), -ΔPOS(i)). In addition, the mechanical operation end position abnormal area search unit 611 also includes the operation abnormal evaluation value d26 at this time as the maximum operation abnormal evaluation value in the actual result position abnormal area operation end determination value d51.

[0203] Figure 16 Indicates the state where the Figure 15 shown mechanical operation end position variation amount is corrected. Figure 16 The range indicated by the marked diagonal line of the action range shown is the corrected part.

[0204] For example, Figure 16 the action range of the manufacturing unit of the mechanical operation end (3) shown is corrected to the range shown by POSEST(3)+ or POSEST(3)- obtained by increasing or decreasing POSEST(3) by a predetermined ΔPOS(3). In addition, there is also a case where the range is not corrected as in the case of POSEST(i)0 = POSEST(i) + 0 (ΔPOS(i) = 0) in the above formula.

[0205] In the example described above, the mechanical operation end position abnormal area search unit 611 uses the maximum and minimum values of each mechanical operation end of the mechanical operation end position variation amount d26 to generate the estimated position d31. In contrast, when the processing power of the computer is sufficient, the estimated position d31 can be further subdivided and generated.

[0206] In addition, the mechanical operation end position abnormal area search unit 611 generates the estimated position d31 by setting the variation amount from the POSEST(i) to three types, but it can also be generated by other processes. For example, the mechanical operation end position abnormal area search unit 611 can also finely control the variation amount of the estimated position d31. In addition, the mechanical operation end position abnormal area search unit 611 can also not perform searches in directions where operation abnormalities clearly do not occur, etc., and change according to the situation in a timely manner. Here, for the direction where operation abnormalities clearly do not occur, consider, for example, the case of moving in the central direction of the actual position of the mechanical operation end.

[0207] The mechanical operation end position abnormal suppression control unit 620 generates the coolant setting d41, which is the set output to the coolant operation end 204, based on the actual position abnormal area operation end determination value d51, which is the output of the mechanical operation end position abnormal area determination unit 610, and the control command of the second shape control unit 212 for the coolant operation end 204.

[0208] When it is determined based on the mechanical operation end position variation amount d26 that operation abnormalities will not occur even if the mechanical operation end position changes, the mechanical operation end position prediction unit 210 can also directly use the coolant operation end position prediction d28 predicted simultaneously with the mechanical operation end position variation amount d26.

[0209] The coolant control rule database 623 presets the correspondence between the mechanical operation end 203 of each manufacturing unit and the affected coolant operation end 204. This correspondence can also be obtained by actually operating the mechanical operation end 203 and the coolant operation end 204 during the rolling operation. Additionally, it can also be obtained by machine learning based on the actual performance data. Here, consider the case where the correspondence is obtained based on the results of actual operation and registered in the coolant control rule database 623.

[0210] [Structure and operation of the coolant operation end control output calculation unit]

[0211] Figure 17 Shows the structure and operation of the coolant operation end control output calculation unit 621.

[0212] In the coolant control rule database 623( Figure 14 ), the necessary amount of coolant flow rate change that can obtain the same effect as operating each mechanical operation end 203(k) is registered. Figure 17 The database retrieval unit 631 shown in (a) in retrieves the necessary amount of coolant flow rate change corresponding to the actual position change amount of the mechanical operation end 203(k) that causes operation abnormalities from the coolant control rule database 623 based on the actual position abnormal area operation end determination value d27 obtained by the mechanical operation end position abnormal area determination unit 610.

[0213] Then, the output synthesis unit 632 adds up the necessary amounts of coolant flow rate changes for each of the extracted mechanical operation ends 203(k) to obtain Figure 17 the abnormal suppression output d42 shown in (b) of

[0214] The coolant operation end control output selection unit 622 adds the abnormal suppression output d42 and the coolant operation end position prediction d28 in the adder 622a, and then performs upper and lower limit processing in the upper and lower limit processing unit 622b, and outputs it as the coolant setting d41.

[0215] In addition to being directly set at the coolant operation end 204, the coolant setting d41 can also be output to the human-machine interface screen as a guidance for the operator, and the operator operates the coolant operation end 204 according to the guidance.

[0216] Based on the above, when it is determined that no operation abnormality caused by the position of the mechanical operation end 203 will occur even if the rolling operation of the next manufacturing unit is performed according to past performance, the coolant operation end position prediction d28 is directly output as the coolant setting d41. In addition, when it is predicted that an operation abnormality will occur, the coolant setting d41 corrected in the direction of suppressing the occurrence of the operation abnormality is output.

[0217] In the mechanical operation end position prediction unit 210, the coolant settings and the position changes of the mechanical operation ends of each manufacturing unit are always learned and used when the same rolling operation of the manufacturing unit is performed next time, so that the coolant setting such as suppressing the occurrence of operation abnormalities can be realized.

[0218] As described above, according to the equipment control device of this example, it is possible to prevent operation abnormalities caused by the actual value d22 of the mechanical operation end position of the mechanical operation end 203 and perform good shape control.

[0219] [Modification Example]

[0220] In addition, the present invention is not limited to the above-described embodiment examples and includes various modification examples. For example, the above-described embodiment examples are examples described in detail for easy understanding of the present invention and are not limited to having all the structures described.

[0221] For example, in the above-described embodiment example, the mechanical operation end safe operation range determination unit 205 is realized by machine learning, but it can also be expressed by a mathematical formula based on the experience of the operator, thereby realizing the mechanical operation end safe operation range determination unit 205. Alternatively, the rolling states at the time of occurrence of operation abnormalities can be databaseized in advance, and the corresponding situation can be determined to realize the mechanical operation end safe operation range determination unit 205.

[0222] In addition, the mechanical operation end safety operation range determination unit 205 can also generate a numerical model and a symbolic logic model based on the knowledge of operators and technicians, and utilize them during machine learning.

[0223] In addition, in the above-described embodiment, the mechanical operation end position abnormality suppression control unit 620 stores and utilizes the results obtained in advance through experiments or the like in the coolant control rule database 623. In contrast, the mechanical operation end position abnormality suppression control unit 620 can also use machine learning to generate a rule base based on actual performance data.

[0224] In addition, in the above-described embodiment, the shape control of a rolling mill is taken as an object, but the present invention can also be applied to general equipment control.

[0225] In addition, in Figure 1 In block diagrams such as etc., the control lines and information lines only represent the control lines and information lines required for explanation, and do not necessarily represent all the control lines and information lines on the product. In fact, it can be considered that almost all the structures are interconnected.

[0226] In addition, the processing units such as the control unit described in the above-described embodiment can be respectively constituted by dedicated hardware, or the functions of the respective processing units described in the above-described embodiment can be realized by installing a program (application program) in a computer.

[0227] Figure 18 This shows an example of the hardware structure when the device control device is constituted by a computer.

[0228] Figure 18 The shown device control device (computer) 100 includes a CPU (Central Processing Unit) 100a, a ROM (Read Only Memory) 100b, and a RAM (Random Access Memory) 100c, which are respectively connected to a bus. Moreover, the device control device 100 includes a non-volatile memory 100d, a network interface 100e, an input / output device 100f, and an output device 100g.

[0229] The CPU 100a is an arithmetic processing unit that reads and executes the program code of the software that realizes the functions performed by the device control device 100 from the ROM 100b.

[0230] Temporarily write variables, parameters, etc. generated during the arithmetic processing into the RAM 100c.

[0231] In the non-volatile memory 100d, a large-capacity information storage medium such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive) is used, for example. A program (device control program) for executing the processing functions performed by the device control device 100 is recorded in the non-volatile memory 100d. In addition, data required for machine learning is recorded in the non-volatile memory 100d.

[0232] The network interface 100e transmits and receives various information to and from the outside via a LAN (Local Area Network), a dedicated line, or the like.

[0233] The input / output device 100f inputs various information from the controlled device 190 (rolling mill 301) and outputs information for instructing each operation terminal 103, 104 (203, 204).

[0234] The display device 100g displays the control state of the controlled device 190 (rolling mill 301).

[0235] In addition, information on the program for implementing each processing function performed by the device control device 100 can be placed not only in non-volatile memories such as HDDs and SSDs but also in recording media such as semiconductor memories, IC cards, SD cards, and optical discs.

[0236] In addition, in the case where a part or all of the processing units of the device control device are configured by hardware, an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit) can also be used.

Claims

1. An apparatus control device performs a first operation process on a controlled object device with a response speed to an operation being a predetermined response speed and a second operation process with a response speed to the operation slower than the first operation process, and is characterized in that the apparatus control device includes: an operation terminal control unit that obtains a state quantity that is a target of the controlled object device and gives an instruction for the first operation process; an operation terminal setting unit that gives an instruction for the second operation process; a first operation terminal that executes the first operation process of the controlled object device according to the instruction of the operation terminal control unit; a second operation terminal that executes the second operation process of the controlled object device according to the instruction of the operation terminal setting unit; a safe operation range determination unit that determines a safe operation range of the first operation process performed by the first operation terminal based on the performance of the first operation terminal; a learning unit that learns the position of the first operation terminal of the first operation process of the controlled object device in units of production materials and the state quantity that is a target according to the operation performance in the controlled object device; and an operation terminal position prediction unit that uses the learning result in the learning unit to predict the position of the first operation terminal during the production of the next production material, and the operation terminal setting unit determines the position of the second operation terminal based on the operation range prediction value of the position of the first operation terminal during the next production material by the operation terminal position prediction unit and the determination of the safe operation range determination unit, so that the position of the first operation terminal does not deviate from the safe operation range during the operation of the next production material.

2. The apparatus control device according to claim 1, characterized in that the control time response of the first operation process performed by the first operation terminal is faster than that of the second operation process, and the influence on the controlled object state quantity of the controlled object device is limited, the control time response of the second operation process performed by the second operation terminal is slower than that of the first operation process, and the entire area of the controlled object state quantity of the controlled object device is affected.

3. The apparatus control device according to claim 1, characterized in that the safe operation range determination unit uses the performance data of the controlled object device to identify the occurrence of an operation abnormality, sets the performance data at the time of the occurrence of the operation abnormality as supervised data, thereby learns the relationship between the performance data and the operation abnormality, and determines the safe operation range, the operation terminal position prediction unit sets the position performance of the first operation terminal and the second operation terminal of the controlled object device in units of production materials and the product performance data including product information as supervised data, thereby predicting the position change of the first operation terminal and the second operation terminal in units of production materials.

4. The apparatus control device according to claim 3, characterized in that machine learning is performed based on the collected performance data to obtain the relationship between the performance data and the operation abnormality.

5. The apparatus control device according to any one of claims 1 to 4, characterized in that the controlled object device is a rolling mill, The first operation process is a mechanical shape operation process that changes the shape through a mechanical structure. The second operation process is a coolant shape operation process that changes the shape by changing the injection amount of the coolant in the plate width direction. The safe operation range determination unit determines the safe operation range of the first operation process based on the actual value of the mechanical position where no operation abnormality occurs at the first operation end. The equipment control device includes: an operation end position prediction unit that learns the position of the first operation end and the position of the second operation end of the rolling mill, which is the equipment to be controlled, in units of production materials based on the operation performance of the rolling mill, and uses the learning result to predict the position of the first operation end during the production of the next production material. The operation end setting unit determines the position of the second operation end based on the predicted value of the operation range of the position of the first operation end during the production of the next production material by the operation end position prediction unit and the judgment of the safe operation range determination unit, so that the position of the first operation end does not deviate from the safe operation range during the operation of the next production material.

6. An equipment control method, in which an arithmetic processing unit executes a first operation process with a predetermined response speed for an operation performed by a first operation end and a second operation process for an operation performed by a second operation end with a response speed slower than that of the first operation process on the equipment to be controlled. It is characterized in that The equipment control method includes: A control step in which the arithmetic processing unit obtains a state quantity that is the target of the equipment to be controlled and gives an instruction for the first operation process. An operation end setting step for giving an instruction for the second operation process. A first operation execution step in which, according to the instruction of the control step, the arithmetic processing unit executes the first operation process of the equipment to be controlled. A second operation execution step in which, according to the instruction given in the operation end setting step, the arithmetic processing unit executes the second operation process of the equipment to be controlled. A safe operation range determination step in which the arithmetic processing unit determines the safe operation range of the first operation process performed in the first operation execution step based on the actual performance of the first operation process. A learning step for learning the position of the first operation end of the equipment to be controlled in units of production materials and the state quantity that is the target based on the operation performance in the equipment to be controlled. And An operation end position prediction step for predicting the position of the first operation end during the production of the next production material using the learning result in the learning step. In the operation end setting step, based on the predicted value of the operation range of the position of the first operation end during the production of the next production material in the operation end position prediction step and the judgment in the safe operation range determination step, the position of the second operation end is determined so that the position of the first operation end does not deviate from the safe operation range during the operation of the next production material.

7. A computer-readable recording medium records a computer program that causes a computer to perform a first operation process with a predetermined response speed for an operation performed by a first operation terminal on a controlled object device, and a second operation process for an operation performed by a second operation terminal with a response speed slower than that of the first operation process. The computer program is characterized in that: The computer program causes the computer to execute: A control step of obtaining a state quantity that is a target of the controlled object device and giving an instruction for the first operation process; An operation terminal setting step of giving an instruction for the second operation process; A first operation execution step of executing the first operation process of the controlled object device according to the instruction of the control step; A second operation execution step of executing the second operation process of the controlled object device according to the instruction of the operation terminal setting step; A safe operation range determination step of determining a safe operation range of the first operation process performed by the first operation execution step based on the result of the first operation process; A learning step of learning the position of the first operation terminal of the first operation process of the controlled object device in units of production materials and the state quantity that is a target according to the operation result in the controlled object device; And An operation terminal position prediction step of predicting the position of the first operation terminal during the production of the next production material using the learning result in the learning step, The operation terminal setting step determines the position of the second operation terminal according to the predicted value of the operation range of the position of the first operation terminal during the next production material in the operation terminal position prediction step and the determination in the safe operation range determination step, so that the position of the first operation terminal does not deviate from the safe operation range during the operation of the next production material.

Images