Factory equipment control system, control method, and computer-readable recording medium

Abstract

The present invention provides a plant equipment control system, a control method, and a computer-readable recording medium. In the plant equipment control system, control rules are efficiently corrected while minimizing the risk of disrupting the control of the plant equipment. The system comprises: a control method learning unit that learns a combination of performance data and control operations of the target plant equipment; a control execution unit that executes control of the target plant equipment based on the combination of performance data and control operations learned by the control method learning unit; and a quality judgment rule learning unit that learns a combination of performance data and control operations of the target plant equipment and the quality of control results. The quality of the control output is judged based on a determined combination of the performance data of the target plant equipment, the control operations, and the quality of the control results, and the control rules are learned using the quality judgment results, performance data, and supervisory data as learning data.

Description

Technical Field

[0001] The present invention relates to a factory equipment control system, a factory equipment control method, and a computer-readable recording medium storing a program. Background Art

[0002] Conventionally, in various types of plant equipment, plant equipment control based on various control theories is performed in order to obtain appropriate control results by controlling the plant equipment.

[0003] An example of factory equipment will be described. For example, in rolling mill control, fuzzy control and neuro-fuzzy control are applied as control theories for shape control of the fluctuating state of the control plate. Fuzzy control is applied to shape control using coolant, while neuro-fuzzy control is applied to shape control of Sendzimir mills. As shown in Patent Document 1, shape control using neuro-fuzzy control is performed in the following manner: the difference between the actual shape pattern detected by a shape detector and the target shape pattern, as well as the similarity ratio with a predetermined reference shape pattern, is calculated. Then, based on the calculated similarity ratio, the control output for the control terminal is calculated according to the control rule represented by the control terminal operation amount for the predetermined reference shape pattern.

[0004] Hereinafter, a conventional technique for shape control of a Sendzimir mill using neuro-fuzzy control will be described.

[0005] In the shape control of Sendzimir mill, neuro-fuzzy control is used. Figure 25 As shown, the Sendzimir mill 50 uses a pattern recognition unit 51 to perform shape pattern recognition based on the actual shape detected by a shape detector 52, calculating which of the two predetermined reference shape patterns is closest to the actual shape. The data on the actual shape detected by the shape detector 52 undergoes pre-processing for pattern recognition in a shape detection pre-processing unit 54.

[0006] Then, the control calculation unit 53 executes control using a control rule composed of the control operation terminal operation amount for a preset shape pattern.

[0007] Here, if Figure 26 As shown, the pattern recognition unit 51 calculates the difference between the actual shape pattern (ε) detected by the shape detector 52 and the target shape (ε ref ) is closest to which of the shapes of Patterns 1 to 8. Then, the control operation unit 53 selects and executes any one of the control methods of Patterns 1 to 8 based on the operation result.

[0008] Prior art literature

[0009] Patent Literature

[0010] Patent Document 1: Japanese Patent No. 2804161

[0011] Patent Document 2: Japanese Patent Application Laid-Open No. 2018-180799 Summary of the Invention

[0012] Problems to be solved by the invention

[0013] In the prior art described in Patent Document 1, a representative shape is pre-set as a reference shape pattern, and control is performed based on a control rule that expresses the relationship between the amount of control terminal operation relative to the reference shape pattern. Learning of the control rule also involves learning the relationship between the amount of control terminal operation relative to the reference waveform pattern, directly using the pre-determined representative reference shape pattern. Consequently, there is a problem with shape control that responds only to a specific shape pattern.

[0014] Reference shape patterns are pre-determined by humans based on knowledge of the target rolling mill, accumulated shape performance, and experience gained through manual intervention. However, they are difficult to capture all shapes that can occur in the target rolling mill and rolled material. Consequently, when a shape differs from the reference shape pattern, shape control may not be executed, leaving shape deviations unchecked. Alternatively, the shape may be mistakenly recognized as a similar reference shape pattern, leading to incorrect control operations and ultimately worsening the shape.

[0015] Therefore, in conventional shape control, a control rule is learned and control is executed using a preset reference shape pattern and a control rule for the reference shape pattern. This leads to a problem that there is a limit to improving control accuracy.

[0016] To address this issue, for example, the technology described in Patent Document 2 has been proposed. Patent Document 2 describes a process that generates interference during control and gradually improves the intelligence of a neural network through learning. However, the process described in Patent Document 2 for generating control interference occurs during actual operation of the controlled plant equipment. The generation of control interference during operation of the controlled plant equipment disrupts the actual operation of the controlled plant equipment, making this method unsuitable for practical purposes. Furthermore, unless the controlled plant equipment is operated to a certain degree, the neural network will not function properly, and there is a high probability that proper control will not be achieved during the brief period from the initial operation.

[0017] An object of the present invention is to provide a plant control system, a plant control method, and a program that can reduce the risk of disrupting control of plant equipment and efficiently correct control rules.

[0018] Means for solving problems

[0019] In order to solve the above-mentioned problems, for example, the structure described in the claims of the patent is adopted.

[0020] The present application includes a plurality of means for solving the above-mentioned problems. As one example, a plant control system is applied to a system that recognizes a combination pattern of performance data of controlled plant equipment and executes control thereof.

[0021] In addition, the factory equipment control system includes: a control method learning unit, which learns the combination of performance data and control operations of the controlled factory equipment; a control execution unit, which executes control of the controlled factory equipment based on the combination of performance data and control operations learned by the control method learning unit; and a quality judgment rule learning unit, which learns the combination of performance data and control operations of the controlled factory equipment and the quality of control results.

[0022] Here, the control execution unit includes:

[0023] a control rule execution unit that provides a control output according to a combination of performance data of the controlled plant equipment and a determination of the control operation;

[0024] a control output quality judgment rule execution unit that judges the quality of the control output based on a combination of performance data of the controlled plant equipment, control operations, and control result quality determination;

[0025] a new search operation amount calculation unit that calculates an operation amount for a new operation search based on the quality determination in the control output quality determination rule execution unit; and

[0026] A control output suppression unit uses the quality judgment of the control output quality judgment rule execution unit and the simulation data of the control simulator to prevent the control output from being output to the controlled object factory equipment when it is determined that the actual performance data of the controlled object factory equipment has deteriorated when the control output is output to the controlled object factory equipment.

[0027] In addition, the learning unit for judging the pros and cons also includes:

[0028] a control result quality determination unit for determining the quality of the control result after a time delay until the control effect is reflected in the actual performance data when the control execution unit outputs the control output to the controlled plant equipment; and

[0029] The quality judgment rule learning unit performs learning using the quality of the control result, the actual performance data, and the control output in the control result quality judgment unit as learning data.

[0030] In addition, the control method learning unit also includes:

[0031] a learning data production unit that obtains supervisory data using the control output quality judgment and the control output in the control output quality judgment rule execution unit; and

[0032] The control rule learning unit performs learning using the actual performance data and the supervisory data as learning data.

[0033] Effects of the Invention

[0034] According to the present invention, the control rules for shape patterns and operating methods used in shape control can reduce risks to factory equipment and automatically and efficiently adjust them to optimal control rules that adapt to environmental changes in factory equipment over time. Consequently, the present invention can improve control accuracy, shorten the startup period of the control unit, and respond to changes over time.

[0035] Furthermore, according to the present invention, by preliminarily evaluating the performance of the control law, there are effects of reducing the risk to plant equipment by applying a new control law and improving the control performance by selecting an optimal control law.

[0036] Other problems, structures, and effects than those described above will become clear from the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a schematic configuration diagram showing a plant equipment control system according to an example of an embodiment of the present invention.

[0038] Figure 2 This is a diagram showing a specific configuration example of a control rule execution unit according to an example of an embodiment of the present invention.

[0039] Figure 3 This is a diagram showing a configuration of an example of a control output quality determination rule execution unit according to one embodiment of the present invention.

[0040] Figure 4 This is a diagram showing a specific configuration example of a control rule learning unit according to an example of an embodiment of the present invention.

[0041] Figure 5 This is a diagram showing a configuration of an example of a quality determination rule learning unit according to one embodiment of the present invention.

[0042] Figure 6 This is a diagram showing an example of judging the quality of control results of a control method in shape control of a Sendzimir mill.

[0043] Figure 7 This is a diagram showing a neural network configuration when used for shape control of a Sendzimir mill according to an example of an embodiment of the present invention.

[0044] Figure 8 It is a diagram illustrating a shape deviation and a control method according to an example of an embodiment of the present invention.

[0045] Figure 9 This is a structural diagram showing an example of a control input data creation unit according to an example of an embodiment of the present invention.

[0046] Figure 10 This is a diagram showing a configuration example of a control output calculation unit according to an example of an embodiment of the present invention.

[0047] Figure 11 This is a diagram showing a neural network configuration when used for quality determination of a Sendzimir mill according to an example of an embodiment of the present invention.

[0048] Figure 12 This is a diagram showing an operation amount calculation method in a new search operation amount calculation unit according to an example of an embodiment of the present invention.

[0049] Figure 13 This is a diagram showing a configuration example of a control output determination unit according to an example of an embodiment of the present invention.

[0050] Figure 14 This is a diagram showing a configuration example of a control output calculation unit according to an example of an embodiment of the present invention.

[0051] Figure 15 This is a diagram showing the processing stages and processing contents in a learning data creation unit according to an example of an embodiment of the present invention.

[0052] Figure 16 This is a diagram showing an example of data stored in a learning data database according to an example of an embodiment of the present invention.

[0053] Figure 17 This is a diagram showing an example of a neural network management table according to an example of an embodiment of the present invention.

[0054] Figure 18 This is a structural diagram showing an example of a learning data database according to an embodiment of the present invention.

[0055] Figure 19 This is a diagram showing a configuration of an example of a control quality determination unit according to an example of an embodiment of the present invention.

[0056] Figure 20 This is a diagram showing an example of data stored in a learning data database according to an example of an embodiment of the present invention.

[0057] Figure 21This is a diagram showing an example of a neural network management table according to an example of an embodiment of the present invention.

[0058] Figure 22 This is a diagram showing an example of a learning data database according to an example of an embodiment of the present invention.

[0059] Figure 23 This is a configuration diagram showing an example in which a plant equipment control system according to an example of an embodiment of the present invention includes a control rule evaluation unit.

[0060] Figure 24 This is a block diagram showing a hardware configuration example of a plant equipment control system according to an example of an embodiment of the present invention.

[0061] Figure 25 This is a structural diagram showing an example of a Sendzimir rolling mill.

[0062] Figure 26 This is a diagram showing an example of a list of control rules in shape control of a Sendzimir mill.

[0063] Description of Reference Numerals

[0064] 1…Control target plant equipment, 2…Control input data creation unit, 3…Control output calculation unit, 4…Control output suppression unit, 5…Control output determination unit, 6…Control result quality determination unit, 7…Learning data creation unit, 10…Control rule execution unit, 16…Control operation interference generation unit, 17…Control output quality determination rule execution unit, 18…Control output operation method selection unit, 20…Control execution unit, 21…Control method learning unit, 22…Quality determination rule learning unit, 23…Control rule evaluation unit, 31…Quality determination rule learning unit, 32… 3…New search operation amount calculation unit, 34…Quality judgment rule accuracy verification unit, 35…Control rule quality judgment data collection unit, 36…Control rule evaluation data calculation unit, 37…Control rule database update unit, 50…Sendzimir mill, 51…Pattern recognition unit, 52…Shape detector, 53…Control calculation unit, 54…Shape detection preprocessing unit, 101…Neural network, 102…Neural network selection unit, 110…Neural network processing unit, 111…Neural network, 112…Neural network learning control unit, 113…Neural network selection unit, 114 ...input data creation unit, 115...supervisory data creation unit, 171...neural network, 172...neural network selection unit, 201...normalized shape deviation, 202...shape deviation stage, 210...shape deviation PP value calculation unit, 211...shape deviation stage calculation unit, 310...neural network processing unit, 311...neural network, 312...neural network learning control unit, 313...neural network selection unit, 314...input data creation unit, 315...supervisory data creation unit, 501...rolling phenomenon model, 502...shape correction quality judgment unit, 503... …shape deviation correction amount prediction data, 504…shape deviation performance data, 505…shape deviation prediction data, 602…shape change quality judgment unit, 801…learning data production unit, 802…control rule learning unit, DB1…control rule database, DB2…learning data database, DB3…output judgment database, DB4…quality judgment database, DB5…quality judgment rule database, DB6…learning data database, DB7…verification data database, DB8…control rule evaluation data database, DB9…control rule evaluation value database. DETAILED DESCRIPTION

[0065] Hereinafter, a plant equipment control system according to an example of an embodiment of the present invention (hereinafter referred to as “this example”) will be described with reference to the drawings.

[0066] First, before explaining this example, the process of completing the present invention and its outline will be explained by taking the case where a plant equipment control system is applied to a shape control device of a rolling mill as an example.

[0067] First, in order to obtain a plant equipment control system that can efficiently correct control rules while reducing risks to plant equipment, which is one of the objectives of the present invention, the following requirements (1), (2), (3), and (4) are necessary.

[0068] Requirement (1): In order to improve the control rules, if a control operation with good control results cannot be learned, the control operation is significantly changed. If the control results are good, the control operation is adopted as a new control operation method. If a control operation with good control results can be learned, the control operation is not changed or only slightly changed. If the control results for this situation are good, the control operation is adopted as a new control operation method.

[0069] Requirement (2): Based on actual machine data, the combination of shape patterns, control operations, and the quality of control results is learned to construct a model that can estimate the quality of control results with higher accuracy than a simulator using a machine model. Through regular automatic learning, a model that is always suitable for the latest factory equipment status is constructed.

[0070] Requirement (3): Using a model for determining the quality of estimated control results, the reliability of the control output suppression function for factory equipment, which was previously performed using only a simple mechanical model, is improved.

[0071] · Requirement (4): In the prior art, in the function of generating control rule learning data in the judgment of the quality of a control result, by using a model for judging the quality of the estimated control result, the influence of noise contained in the plant equipment data can be suppressed, and fine adjustments with small effects can also be included as objects of learning data. At the same time, by preventing erroneous judgments of the control effect, changes in the learning data can be suppressed, and the control performance can be stabilized.

[0072] To achieve these requirements (1) to (4), it is preferable to construct a neural network within the control device that can learn the combination of the shape pattern, control operation, and the quality of the control result used in shape control. The control device then uses the value obtained by inputting the shape pattern generated by the rolling mill and the output of the control operation into the neural network to estimate the quality of the control result generated by the output of the control operation with respect to the shape pattern generated by the rolling mill. Furthermore, the control device uses the estimated value of the quality of the control result to select a method for calculating the control operation amount used to search for a new control operation.

[0073] For outputs that indicate significant shape degradation, as determined through verification using a simple model of the rolling mill, the control device prevents shape degradation by suppressing outputs at the mill's control terminals. In this case, the control device uses an estimated value of the control result to determine whether output suppression is effective. This improves protection reliability and optimizes the suppression range, thereby expanding the control function's applicable range.

[0074] In the early stages of application when the accuracy of estimating the quality of control results is insufficient, even control operation outputs estimated to be poor need to be output to factory equipment, thereby expanding the scope of learning for combinations of unlearned shape patterns, control operations, and quality of control results.

[0075] When the estimation accuracy of the control result is sufficiently high, the control result can be estimated without outputting the manipulated variable to the plant equipment. Therefore, the control device can generate learning data for the control rule.

[0076] By using a neural network capable of estimating the quality of control results, the control device can reduce the impact of noise on plant equipment data and determine the quality of even small adjustments. This allows the control device to generate learning data. Furthermore, by preventing erroneous quality determinations caused by noise, the control device can improve the accuracy of the learning data.

[0077] In addition, when the accuracy of the estimated quality judgment of the control results is reduced due to environmental changes of the factory equipment over time, the control device can estimate the quality judgment of the control results suitable for the latest factory equipment status by relearning using the performance data of the most recent factory equipment.

[0078] To verify the estimated accuracy of the control result quality judgment, test data for accuracy verification is prepared separately from the data used for neural network training. The control device can then verify the estimated accuracy of the quality judgment based on the difference between the value output by the neural network when the shape pattern and control operation output included in this accuracy verification test data are input to the neural network and the control result quality judgment included in the test data.

[0079] Figure 1 The structure of the factory equipment control system of this example is shown.

[0080] Figure 1 The plant equipment control system includes a control execution unit 20, a control method learning unit 21, a quality determination rule learning unit 22, a plurality of databases DB (DB1 to DB8), and a management table TB of each database DB.

[0081] The control execution unit 20 receives the performance data Si from the controlled plant 1 and provides the control operation variable output SO determined according to the control rule (neural network) to the controlled plant 1 to control the controlled plant 1. The controlled plant 1 is the already described Figure 25 A Sendzimir rolling mill 50 is shown.

[0082] Here, the control rules are as follows: Figure 26 As described in, for example, the detected shape pattern A(ε) and the target shape (ε ref The control execution unit 20 selects and executes a control method for any of the patterns based on the calculation result of the control rule.

[0083] The control method learning unit 21 inputs the control input data S1 and the like generated by the control execution unit 20 to perform learning, and reflects the learned control rules in the control rules in the control execution unit 20 .

[0084] The performance determination rule learning unit 22 inputs the performance data Si before and after control of the control target plant 1 and performs learning, and reflects the learned performance determination rule in the performance determination rule in the control execution unit 20 .

[0085] The control execution unit 20 includes a control input data production unit 2, a control rule execution unit 10, a control output calculation unit 3, a control output suppression unit 4, a control output judgment unit 5, a control output quality judgment rule execution unit 17, a new search operation amount calculation unit 33 and a control output operation method selection unit 18.

[0086] The control execution unit 20 creates input data S1 for the control rule execution unit 10 using the control input data creation unit 2 based on the performance data Si of the rolling mill as the control target plant 1 .

[0087] The control rule execution unit 10 uses a neural network (control rule) that represents the relationship between the performance data Si of the controlled object and the control terminal operation command S2 to execute the control rule execution process, generating the control terminal operation command S2 based on the performance data Si of the controlled object. The control output calculation unit 3 calculates the control operation variable S3 for the control terminal based on the control terminal operation command S2. Thus, the control execution unit 20 generates the control operation variable S3 using the neural network based on the performance data Si of the controlled object plant 1.

[0088] Furthermore, the control output quality judgment rule execution unit 17 uses a neural network (quality judgment rule) that represents the relationship between the actual performance data Si of the controlled object, the control operation variable S3, and the control result quality data S6 of the control result to perform control output quality judgment rule execution processing to generate a control output quality judgment estimate value S9 based on the actual performance data Si of the controlled object and the control operation variable S3. Furthermore, the control output quality judgment rule execution unit 17 generates a control result quality judgment estimate value S11 based on the actual performance data Si of the controlled object and the selected control operation variable S8 described later.

[0089] The new search operation amount calculation unit 33 performs a new search operation amount calculation process of calculating a new search control operation amount S12 based on the control output quality determination estimated value S9.

[0090] The control output operation method selection unit 18 creates a selected control operation amount S8 and a control method selection flag S14 based on the control operation amount S3 or the newly searched control operation amount S12.

[0091] Furthermore, the control output determination unit 5 within the control execution unit 20 uses the performance data Si from the controlled plant 1 and the control manipulated variable S3 from the control output calculation unit 3 to determine control manipulated variable output permission data S4 for the controlled manipulated terminal. Based on the control manipulated variable output permission data S4 and the control result quality evaluation estimate S11, the control output suppression unit 4 determines whether or not the selected control manipulated variable S8 can be output to the controlled manipulated terminal. The selected control manipulated variable S8, which has been determined to be permitted, is output as the control manipulated variable output SO to the controlled plant 1. Consequently, the selected control manipulated variable S8, if determined to be abnormal, is not output from the control execution unit 20 to the controlled plant 1.

[0092] The control execution unit 20 configured as described above refers to the control rule database DB1 , the output determination database DB3 , and the quality determination rule database DB5 in order to execute the processing.

[0093] The control rule database DB1 is connected to both the control rule execution unit 10 in the control execution unit 20 and the control rule learning unit 802 in the control method learning unit 21 to be described later in an accessible manner.

[0094] The control rule database DB1 stores control rules (neural networks) that are learning results of the control rule learning unit 802. The control rule execution unit 10 refers to the control rules stored in the control rule database DB1.

[0095] The learning data database DB2 stores the learning data obtained by the control rule learning unit 802 .

[0096] The output determination database DB3 is connected to the control output determination unit 5 in the control execution unit 20 in an accessible manner, and the output determination results are stored in the output determination database DB3.

[0097] The quality judgment database DB4 stores data used for quality judgment.

[0098] The quality judgment rule database DB5 stores the quality judgment rules (neural networks) learned by the quality judgment rule learning unit 31. This quality judgment rule database DB5 is connected to both the control output quality judgment rule execution unit 17 within the control execution unit 20 and the quality judgment rule learning unit 31 within the quality judgment rule learning unit 22, described later, in an accessible manner. The control output quality judgment rule execution unit 17 refers to the quality judgment rules stored in the quality judgment rule database DB5.

[0099] The learning data database DB6 stores the learning data learned by the control method learning unit 21 .

[0100] The verification data database DB7 stores verification data required for quality determination.

[0101] Figure 2 A specific configuration example of the control rule execution unit 10 of this example is shown.

[0102] The control rule execution unit 10 receives the input data S1 generated by the control input data generation unit 2. The control rule execution unit 10 processes the input data S1 and provides the control operation terminal operation instruction S2 to the control output calculation unit 3. The control rule execution unit 10 includes a neural network 101. The neural network 101 outputs a control signal that follows the control signal. Figure 26 The control operation terminal operation instruction S2 of the shape control rule shown.

[0103] The control rule execution unit 10 further includes a neural network selection unit 102 , which refers to the control rules stored in the control rule database DB1 , selects the optimal control rule as the control rule in the neural network 101 , and causes the neural network 101 to execute the optimal control rule.

[0104] In this way, the control rule execution unit 10 selects and uses a desired neural network from among multiple neural networks classified by operator group and control purpose. The control rule database DB1 may also include data from the controlled plant 1, including performance data (such as operator group data) Si that enables selection of a neural network and a quality assessment criterion.

[0105] Furthermore, since executing a neural network becomes a control law, in this specification, the terms neural network and control law are used synonymously.

[0106] Figure 3 The specific structure of the control output quality determination rule execution unit 17 is shown.

[0107] The control output quality judgment rule execution unit 17 receives input data S1 generated by the control input data generation unit 2 and control operation variable S3 generated by the control output calculation unit 3. Based on these input data, the control output quality judgment rule execution unit 17 generates a control output quality judgment estimated value S9 and supplies it to the new search operation variable calculation unit 33.

[0108] Furthermore, the control output quality judgment rule execution unit 17 receives input data S1 generated by the control input data generation unit 2 and the selected control operation variable S8 generated by the control output operation method selection unit 18. Based on these input data, the control output quality judgment rule execution unit 17 generates a control result quality judgment estimate S11 and provides it to the control output suppression unit 4.

[0109] The control output quality determination rule execution unit 17 includes a neural network 171 and a neural network selection unit 172 .

[0110] The neural network 171 estimates a quality judgment value of a control result when a control operation variable S3 (control pattern) is outputted for input data S1 (shape pattern) based on past control performance.

[0111] The neural network selection unit 172 refers to the grade determination rules stored in the grade determination rule database DB5 and selects the best grade determination rule as the grade determination rule in the neural network 171 .

[0112] In this way, the control output quality judgment rule execution unit 17 selects a necessary neural network from a plurality of neural networks divided according to differences in the properties of the material to be controlled and differences in the quality judgment criteria.

[0113] The quality judgment rule database DB5 may include data from the controlled plant equipment 1, such as performance data Si (such as data on operation groups) that allows selection of material properties to be controlled and quality judgment criteria. Furthermore, since the execution of a neural network generates quality judgment rules, the terms "neural network" and "quality judgment rules" are used synonymously in this specification.

[0114] Return to Figure 1 As described above, the control method learning unit 21 performs learning of the neural network 101 used in the control execution unit 20.

[0115] The control method learning unit 21 includes a learning data creating unit 801 and a control rule learning unit 802 .

[0116] The learning data generation unit 801 within the control method learning unit 21 uses the selected control operation variable S8 generated by the control execution unit 20, the control method selection flag S14, and the control result quality judgment estimate S11 generated by the control output quality judgment rule execution unit 17 to generate new supervisory data S7a for use in learning the neural network. The supervisory data S7a generated by the learning data generation unit 801 is provided to the control rule learning unit 802.

[0117] The supervisory data S7 a corresponds to the control operation terminal operation instruction S2 output by the control rule execution unit 10 .

[0118] The learning data production unit 801 uses the control result quality judgment estimated value S11 produced by the control output quality judgment rule execution unit 17 to obtain data obtained by estimating the control operation terminal operation instruction S2 output by the control rule execution unit 10 as new supervision data S7a.

[0119] Figure 4 A specific configuration example of the control rule learning unit 802 is shown.

[0120] The control rule learning unit 802 includes an input data creating unit 114 , a supervisory data creating unit 115 , a neural network processing unit 110 , and a neural network selecting unit 113 .

[0121] The control rule learning unit 802 receives external inputs such as input data S1 from the control input data generating unit 2 and new supervisory data S7a from the learning data generating unit 801. The control rule learning unit 802 also refers to data stored in the control rule database DB1 and the learning data database DB2.

[0122] In the control rule learning unit 802 , the input data S1 is taken into the neural network processing unit 110 via the input data generating unit 114 .

[0123] Furthermore, in the control rule learning unit 802, the new supervisory data S7a from the learning data generation unit 801 is provided to the neural network processing unit 110 in the form of supervisory data S7c, which is a combination of the supervisory data S7b stored in the learning data database DB2 by the supervisory data generation unit 115 and the past supervisory data S7b. These supervisory data S7a and S7b are appropriately stored in the learning data database DB2 for use.

[0124] Similarly, input data S8a from the control input data generator 2 is provided to the neural network processing unit 110 in the form of input data S8c, which is a sum of input data S8b stored in the learning data database DB2 and the past input data S8b stored in the learning data database DB2 by the input data generator 114. These input data S8a and S8b are also appropriately stored in the learning data database DB2 for use.

[0125] The neural network processing unit 110 is composed of a neural network 111 and a neural network learning control unit 112 .

[0126] The neural network 111 takes in the input data S8c from the input data generating unit 114, the supervisory data S7c from the supervisory data generating unit 115, and the control rule (neural network) selected by the neural network selecting unit 113, and stores the finally determined neural network in the control rule database DB1.

[0127] The neural network learning control unit 112 controls the input data creation unit 114, the supervisory data creation unit 115, and the neural network selection unit 113 at appropriate times to obtain input to the neural network 111 and store the processing results in the control rule database DB1.

[0128] Here, Figure 2 The neural network 101 in the control rule execution unit 10 and Figure 4 The neural networks 111 in the control method learning unit 21 are all neural networks with the same concept, but there are differences as follows.

[0129] That is, the neural network 101 in the control rule execution unit 10 is a neural network with predetermined contents, and is a neural network that obtains the control operation terminal operation instruction S2 as a corresponding output when the input data S1 is provided.

[0130] On the other hand, the neural network 111 in the control method learning unit 21 is a neural network that finds the input-output relationship through learning when the input data S1 and the control operation terminal operation instruction S2 are set as learning data.

[0131] The basic processing in the control method learning unit 21 is considered as follows.

[0132] First, if the control operation variable output permission data S4 indicates "yes," the control execution unit 20 outputs the control operation variable output SO to the controlled plant 1. If the control result quality evaluation estimate S11 indicates "good" (changing toward a better performance data Si), the learning data creation unit 801 determines that the selected control operation variable S8 output by the control output operation method selection unit 18 is correct, and creates learning data so that the output of the neural network corresponds to the selected control operation variable S8.

[0133] On the other hand, when the content of the control operation quantity output possibility data S4 is "No", or the content of the control result quality judgment estimated value S11 of the control operation quantity output SO output to the controlled object factory equipment 1 is "No" (changing in the direction of deterioration of the performance data Si), the learning data production unit 801 determines that the selected control operation quantity S8 output by the control output operation method selection unit 18 is wrong.

[0134] In this case, the learning data generation unit 801 checks whether the control operation variable S3 is selected in the control output operation method selection unit 18 based on the control method selection flag S14. If the control operation variable S3 is selected during this confirmation, the learning data generation unit 801 determines that the control operation terminal operation instruction S2 output by the control rule execution unit 10 is incorrect and generates learning data so as not to output the neural network output. In this case, the neural network output is configured such that two outputs, one in the positive direction and one in the negative direction, are output to the same control operation terminal as the control output, and the learning data is generated so as not to output the output-side control operation terminal operation instruction S2.

[0135] In addition, about Figure 4 The control rule learning unit 802 shown performs the following processing as a result of the data processing by the neural network learning control unit 112 .

[0136] First, the control rule learning unit 802 performs learning of the neural network 101 used by the control rule execution unit 10 using learning data that is a combination of data S8c obtained from the input data S1 to the control execution unit 20 and supervisory data S7c generated by the supervisory data generation unit 115 .

[0137] In fact, the control rule learning unit 802 has a neural network 111 that is the same as the neural network 101 of the control rule execution unit 10. It performs application tests under various conditions and learns the responses at that time. As a result of the learning, it is confirmed that the control rule produces better results.

[0138] Since learning requires multiple learning data sets, multiple sets of past learning data sets are retrieved from the learning data database DB2, which stores previously generated learning data, for learning and processing. The current learning data is then stored in the learning data database DB2. Furthermore, the learned neural network is stored in the control rule database DB1 for use in the control rule execution unit 10.

[0139] Neural network learning can be performed by using past learning data each time new learning data is generated, or by using past learning data after a certain amount of learning data (e.g., 100) has been accumulated.

[0140] With such a configuration, the control output operation method selection unit 18 selects a new search operation amount, thereby outputting the new search operation amount to the target plant equipment, creating learning data based on the control result, and learning a new control method.

[0141] Return to Figure 1 As described above, the quality judgment rule learning unit 22 executes the neural network 171 ( Figure 3 ) learning. When the control execution unit 20 outputs the control operation variable output SO to the controlled plant equipment 1, it takes time for the actual control effect to be expressed as a change in the performance data Si. Therefore, learning is performed using data obtained by delaying the time. In addition, Figure 1 etc., recorded as "Z -1 The processing unit DL of " indicates that there is an appropriate time delay when transmitting each data.

[0142] The quality judgment rule learning unit 22 includes a control result quality judgment unit 6 , a quality judgment rule learning unit 31 , and a quality judgment database DB4 .

[0143] The control result evaluation unit 6 uses the actual performance data Si and the previous performance data value Si0 from the controlled plant 1, as well as the evaluation data S5 stored in the evaluation database DB4, to perform a control result evaluation process to determine whether the actual performance data Si is changing in an improving or deteriorating direction. The control result evaluation unit 6 then outputs control result evaluation data S6 indicating the evaluation result.

[0144] Figure 5 The specific structure of the quality determination rule learning unit 31 is shown.

[0145] The quality determination rule learning unit 31 includes an input data creating unit 314 , a supervisory data creating unit 315 , a neural network processing unit 310 , and a neural network selecting unit 313 .

[0146] The quality judgment rule learning unit 31 receives, as external inputs, data S12a1 obtained by time-delaying the input data S1 from the control input data generating unit 2 and data S12a2 obtained by time-delaying the control output amount S0 from the control output suppressing unit 4. Furthermore, the quality judgment rule learning unit 31 receives control result quality data S6 from the control result quality judgment unit 6 (S13a).

[0147] Then, the grade determination rule learning unit 31 refers to the data accumulated in the grade determination rule database DB5 and the learning data database DB6 .

[0148] The input data S1 and the control output amount S0 are taken into the neural network processing unit 310 via the input data creation unit 314 after appropriate time delay compensation.

[0149] Furthermore, the control result quality data S6 (S13a) from the control result quality determination unit 6 is provided to the neural network processing unit 310 in the form of supervisory data S13c, which is a total of supervisory data S13c including the past supervisory data S13b stored in the learning data database DB6 in the supervisory data creation unit 315. These supervisory data S13a and S13b are appropriately stored in the learning data database DB6 for use.

[0150] Similarly, the input data S12a1 and S12a2 from the control input data generator 2 and the control output suppressor 4 are provided to the neural network processing unit 310 in the form of total input data S12c, including the past input data S12b stored in the learning data database DB6 in the input data generator 314. These input data S12a1, S12a2, and S12b are appropriately stored in the learning data database DB6 for use.

[0151] The neural network processing unit 310 is composed of a neural network 311 and a neural network learning control unit 312 .

[0152] The neural network 311 receives input data S12c from the input data generator 314, supervisory data S13c from the supervisory data generator 315, and the control rule (neural network) selected by the neural network selector 313. The neural network 311 then stores the final neural network in the quality judgment rule database DB5.

[0153] The neural network learning control unit 312 controls the input data generation unit 314, the supervisory data generation unit 315, and the neural network selection unit 313 at appropriate times to obtain input to the neural network 311. Furthermore, the neural network learning control unit 312 stores the processing results in the quality judgment rule database DB5 via the neural network selection unit 313.

[0154] Figure 6 This is a diagram showing a specific example of judging the quality of control results of a control method in shape control of a Sendzimir mill. Figure 6 express Figure 26 The control results of each shape control rule are shown.

[0155] Here, Figure 3 The neural network 171 and the control execution unit 20 shown Figure 5 The neural networks 311 in the quality judgment rule learning unit 22 shown are all neural networks of the same concept, but differ in the following aspects.

[0156] The neural network 171 in the control execution unit 20 is a neural network with predetermined contents. Specifically, the neural network 171 calculates the control output quality judgment estimated value S9 or S11 as the corresponding output when input data S1 and the selected control operation variable S8 or control operation variable S3 are provided. This is a neural network utilized for so-called one-way processing.

[0157] In contrast, the neural network 311 in the quality judgment rule learning unit 22 satisfies the input-output relationship between the input data S1 and the input data S12c and the supervisory data S13c of the control output S0 when these are set as learning data.

[0158] Next, a specific example of a plant equipment control method will be described with respect to shape control in a Sendzimir mill. The shape control will be described assuming that the following specifications A and B are employed.

[0159] Specification A is a priority specification and contains information on the priority across the plate width. For example, in shape control, it is often difficult to maintain target mechanical properties across the entire plate width. Therefore, two priority specifications, A1 and A2, are set for the plate width. Priority specification A1 prioritizes the plate ends. Priority specification A2 prioritizes the center.

[0160] Control is performed according to the two priority levels of specifications A1 and A2. That is, when the plant equipment control system performs control, either specification A1 or A2 regarding the priority level is considered.

[0161] Specification B is a specification for responding to pre-determined conditions. For example, the relationship between the shape pattern and the control method changes depending on the conditions. Therefore, for example, specification B1 is defined as the plate width, and specification B2 is defined as the steel type. The degree of influence on the shape of the shape manipulation end varies with each specification change.

[0162] In this example, the controlled plant 1 is a Sendzimir mill, and the performance data is shape performance. A Sendzimir mill is a rolling mill with multiple cluster rolls used for cold rolling hard materials such as stainless steel. To apply high pressure to hard materials, Sendzimir mills use small-diameter work rolls. Therefore, control to achieve flat steel plates is difficult with Sendzimir mills. To address this, Sendzimir mills employ a cluster roll structure and various shape control units.

[0163] Sendzimir mills typically have shiftable, single-tapered upper and lower intermediate rolls. In addition, they are equipped with six upper and lower split rolls and two AS-U rolls. In the example described below, the actual shape data Si uses detection data from a shape detector, and the input data S1 uses the shape deviation, the difference from the target shape. Furthermore, the control operation variable S3 represents the roll shift amount for AS-Us #1 to #n and the upper and lower intermediate rolls.

[0164] Figure 7 The following shows the structure of a neural network used for shape control in a Sendzimir mill. Here, neural network 101 is shown as the neural network used in the control rule execution unit 10. Furthermore, neural network 111 is shown as the control rule learning unit 802. Neural networks 101 and 111 have the same structure.

[0165] In the example of shape control of a Sendzimir mill, the performance data Si from the controlled plant 1 is the performance data of the Sendzimir mill, including data from a shape detector (here, the output is the difference between the actual shape and the target shape, i.e., the shape deviation). The control input data generator 2 receives the normalized shape deviation 201 and the shape deviation level 202 as input data S1. Thus, the input layer of the neural network 101, 111 is composed of the normalized shape deviation 201 and the shape deviation level 202. In addition, Figure 7 In the example, the shape deviation stage 202 is set as the input to the neural network input layer, but the neural network can also be switched according to the stage.

[0166] The output layers of neural networks 101 and 111 correspond to the AS-U and first intermediate roll, which are the shape control operating terminals of the Sendzimir mill, and are composed of an AS-U operation level 301 and a first intermediate operation level 302. Each operation level has an AS-U opening direction (the direction in which the roll gap (the gap between the upper and lower working rolls of the rolling mill) is opened) and an AS-U closing direction (the direction in which the roll gap is closed).

[0167] Regarding the first intermediate rolls, the upper and lower first intermediate rolls have a first intermediate roll opening direction (direction in which the first intermediate rolls move outward from the mill center) and a first intermediate roll closing direction (direction in which the first intermediate rolls move toward the mill center side).

[0168] For example, if the shape detector has 20 regions and the shape deviation stage 202 is set to three stages (large, medium, and small), the input layer has 23 inputs. Furthermore, if the AS-U has seven saddles and the upper and lower first intermediate rollers are displaceable in the sheet width direction, the output layer has 14 AS-U operation levels 301 and 4 first intermediate operation levels, for a total of 18. The number of intermediate layers and the number of neurons in each layer are appropriately set.

[0169] Furthermore, the shape control operation terminals of the Sendzimir mill as the output layer constitute a neural network output so that two types of outputs, a + direction and a - direction, are output to each control operation terminal.

[0170] Figure 8 Represents the shape deviation and control method in this example.

[0171] Figure 8 (a) shows the control method when the shape deviation is large. Figure 8 (b) shows a control method when the shape deviation is small. Figure 8 The height direction (vertical axis direction) of (a) and (b) represents the size of the shape deviation, and the horizontal axis direction represents the plate width direction. The two ends of the plate width represent the plate ends, and the center represents the plate center.

[0172] like Figure 8 As shown in (a), when the shape deviation is large, priority is given to correcting the overall shape rather than local shape deviation in the plate width direction.

[0173] On the other hand, Figure 8 As shown in (b), when the shape deviation is small, priority is given to reducing the local shape deviation.

[0174] In this way, the control method needs to be changed according to the size of the shape deviation, so Figure 7As shown, a shape deviation stage 202 is provided and provided to neural networks 101 and 111 to determine the magnitude of the shape deviation. Regardless of the magnitude of the shape deviation, a normalized shape deviation between 0 and 1 can be used, for example. This is an example; the shape deviation can be directly input to the input layer of the neural network without normalization, or the neural network itself can be modified according to the magnitude of the shape deviation. For example, two neural networks can be prepared: one for use when the shape deviation is large, and one for use when the shape deviation is small.

[0175] In this example of factory equipment control, the Figure 7 Neural networks 101 and 111 with such structures learn how to operate with shape patterns and use the learned neural networks to perform shape control. Even neural networks with the same structure can have different characteristics depending on the learning conditions, and can output different control outputs for the same shape pattern.

[0176] Therefore, by using multiple neural networks separately according to other conditions of shape performance, it is possible to configure optimal control for various conditions. This is a response to specification B. Figure 2 The structure of represents a specific example of the case where such specifications are implemented.

[0177] That is, in Figure 2 In the configuration example, different neural networks are prepared for the neural network 101 used in the control rule execution unit 10 based on rolling performance, mill operator name, steel type of rolled material, plate width, and the like, and are registered in the control rule database DB1. The neural network selection unit 102 selects a neural network that matches the conditions at that point in time and sets it as the neural network 101 of the control rule execution unit 10.

[0178] Furthermore, as the condition at this time point in the neural network selection unit 102, the data of the plate width can be obtained from the actual performance data Si in the control target plant 1, and a neural network can be selected according to the data. Figure 7 For input layers and output layers as shown, the number of intermediate layers and the number of units in each layer can also be different.

[0179] Figure 9 The structure of the control input data generating unit 2 for generating data S1 (normalized shape deviation 201 , shape deviation stage 202 ) to be input to the input layer of the neural network 101 , 111 is shown.

[0180] The control input data generator 2 receives, as input, shape detector data from a shape detector that detects the shape of a sheet during rolling in a Sendzimir mill, the controlled plant 1, as performance data Si. The control input data generator 2 then uses a shape deviation PP value calculator 210 to calculate a shape deviation PP value (Peak to Peak value) SPP, which is the difference between the maximum and minimum values ​​of the detection results for each shape detector region.

[0181] The shape deviation stage calculation unit 211 classifies shape deviation into three stages: large, medium, and small, based on the shape deviation PP value SPP. Shape represents the distribution of elongation in the width direction of the rolled material, using units ranging from I to UNIT, where elongation is expressed in 10-5 units. For example, classification is performed as shown in the following formula.

[0182] Here, the shape deviation stages are classified as (large = 1, medium = 0, small = 0) by satisfying [Formula 1], (large = 0, medium = 1, small = 0) by satisfying [Formula 2], and (large = 0, medium = 1, small = 0) by satisfying [Formula 3]. The shape deviations of each region are normalized using SPM, where SPM = SPP.

[0183] [Formula 1]

[0184] SPP≥50I-UNIT

[0185] [Formula 2]

[0186] 50I-UNIT>Spp≥10I-UNIT

[0187] [Formula 3]

[0188] 10I-UNIT>Spp As described above, the control input data generator 2 generates the normalized shape deviation 201 and the shape deviation level 202 as input data to the neural network 101 . The normalized shape deviation 201 and the shape deviation level 202 are input data S1 of the control rule execution unit 10 .

[0189] Figure 10 The configuration of the control output calculation unit 3 is shown.

[0190] The control output calculation unit 3 generates an operation instruction, or control operation amount S3, for each shape control operation terminal based on the control operation terminal operation instruction S2, which is the output from the neural network 101 within the control rule execution unit 10. In the case of shape control of a Sendzimir mill, the control operation terminal operation instruction S2 corresponds to the AS-U operation level 301 and the first intermediate operation level 302.

[0191] exist Figure 10 , one data example is shown for each of the AS-U operation levels 301 and the first intermediate operation level 302, which exist in plurality. Each data example is composed of a pair of data of an opening direction level and a closing direction level.

[0192] In the control output calculation unit 3, the input AS-U operation degree 301 has outputs of the opening direction and the closing direction of each AS-U, so the difference between them is calculated by the subtractor 303. Then, the output of the subtractor 303 is multiplied by the conversion gain G in the multiplier 304. ASU , generate and output the operation instructions to each AS-U. Since the control output to each AS-U is the AS-U position change (unit is length), the conversion gain G ASU It becomes the conversion gain from the operation degree to the position change amount.

[0193] Similarly, the first intermediate operation level 302 inputted has outputs outside and inside the first intermediate level, and thus the difference between them is calculated by the subtractor 305. Then, the output of the subtractor 305 is multiplied by the conversion gain G by the multiplier 306. 1ST , thereby generating and outputting an operation command for the displacement of each first intermediate roller. Since the control output to each first intermediate roller becomes the displacement position change amount of the first intermediate roller (unit is length), the conversion gain G 1ST It becomes the conversion gain from the operation degree to the position change amount.

[0194] Using the above method, the control output calculation unit 3 can calculate the control operation variable S3. The control operation variable S3 is composed of the #1 to #n AS-U position change amounts (n depends on the number of saddles of the AS-U roller), the upper first intermediate shift position change amount, and the lower first intermediate shift position change amount.

[0195] Figure 11 The following shows the structure of a neural network used for determining the quality of the shape control results of a Sendzimir mill used in the control output quality determination rule execution unit 17 and the quality determination rule learning unit 31. The neural network shown here is neural network 171 when used in the control output quality determination rule execution unit 17, and is neural network 311 when used in the quality determination rule learning unit 31, but the structures are the same in both cases.

[0196] As input data S1, the normalized shape deviation 201 and the shape deviation stage 202 are used. Figure 7 The same signal as that described in the input to the neural network input layer is input to the input layer. Furthermore, the control operation variable S3 or the selected control operation variable S8 described later is input to the input layer. The control operation variable S3 or the selected control operation variable S8 is composed of the position change amount of each control operation device.

[0197] The output layer also outputs a value that determines the quality of the control result when the control operation variable S3 or the selected control operation variable S8 is output for the input data S1. The number of layers in the intermediate layer and the number of neurons in each layer are appropriately set.

[0198] Figure 12 The operation amount calculation method in the new search operation amount calculation unit 33 is shown.

[0199] The new search operation amount calculation unit 33 uses the control output quality judgment estimated value S9 output by the control output quality judgment rule execution unit 17 to calculate a new search control operation amount S12 according to the following policy.

[0200] That is, when the control output quality evaluation value S9 is large, the quality evaluation of the control operation is estimated to be good, so the new search operation amount calculation unit 33 performs a fine adjustment as the new search operation amount.

[0201] When the control output quality evaluation value S9 is small, the quality evaluation of the control operation is estimated to be poor. Therefore, the new search operation amount calculation unit 33 searches for a new appropriate operation method by significantly changing the control operation.

[0202] Based on the above principles, the formula for calculating the new search operation amount Crand is set as follows.

[0203] IF(S9>th) THEN Crand=Cref * (1+β * th1)

[0204] IF(th≥S9)THEN Crand=Cref+γ * th2 * G

[0205] Here, β and γ represent random values ​​generated between -1 and 1. th1 represents the degree of fine-tuning.

[0206] For example, when a range of ±10% of the original instruction is used as a fine adjustment, th1 is set to 0.1.

[0207] th2 is a setting that significantly changes the operation method. For example, if th2 is set to 0.1, a 10% offset is added to the original instruction, which may cause the operation polarity to change or output an instruction to a device that was not originally operated.

[0208] The values ​​of β and γ are different for each operating device, and the operation amount of each device is changed independently. G represents the maximum operating position control command of each control operating device. By multiplying it with the command % above, the value of % is converted into the operating position control command.

[0209] The control output operation method selection unit 18 selects either the control operation amount S3 or the new search control operation amount S12 and outputs it as the selected control operation amount S8. The selection of either the control operation amount S3 or the new search control operation amount S12 is probabilistically determined, and the user can set the probability Prand of using the new search control operation amount S12 between 0 and 1. The value δ, which randomly takes a value between 0 and 1, is determined by the following equation.

[0210] IF(δ>Prand)THEN C”ref=Cref, α=1

[0211] ELSE C"ref=Crand、α=0

[0212] Here, C"ref represents the selected control operation amount S8 output by the control output operation method selection unit 18 to the subsequent operation unit. δ uses a shared value in the calculation of the operation amounts of all devices, and all devices use the operation amount on the same side. α is the control method selection flag S14, which takes 1 when the control operation amount S3 is selected and takes 0 when the new search control operation amount S12 is selected. The control method selection flag S14 is output to the subsequent operation unit together with the selected control operation amount S8. As a setting method of Prand, in the control of the actual machine, it is set to 0 when you do not want to impose risks on factory equipment through random operations, and it is set to a ratio other than 0 when you want to output a new search operation amount in order to improve the control rules.

[0213] Figure 13 The configuration of the control output determination unit 5 is shown.

[0214] The control output determination unit 5 is composed of a rolling phenomenon model 501 and a shape correction quality determination unit 502. Furthermore, the control output determination unit 5 receives performance data Si from the controlled plant equipment 1, control operation variables S3 from the control output calculation unit 3, and information from the output determination database DB3, and provides control operation variable output permission data S4 to the control operation terminal.

[0215] The control output determination unit 5 with this configuration predicts the change in shape when the selected control operation variable S8 calculated by the control output operation method selection unit 18 is output to the rolling mill, which is the control target plant 1, by inputting it into a known model of the control target plant 1. Here, the model of the known control target plant 1 is the rolling phenomenon model 501. In this prediction, if shape deterioration is predicted, the control output determination unit 5 suppresses the control operation variable output SO to prevent significant shape deterioration.

[0216] More specifically, the control output determination unit 5 inputs the selected control operation amount S8 into the rolling phenomenon model 501 , predicts the shape change due to the selected control operation amount S8 , and calculates shape deviation correction amount prediction data 503 .

[0217] Meanwhile, the control output determination unit 5 adds the shape deviation correction amount prediction data 503 to the shape detector data Si from the controlled plant 1 to obtain shape deviation prediction data 505, and then evaluates the shape deviation prediction data 505. This allows the control output determination unit 5 to predict how the shape will change when the control operation variable S3 is output to the controlled plant 1. The shape detector data Si here refers to the current point in time's actual shape deviation performance data 504.

[0218] The control output determination unit 5 determines whether the shape correction quality determination unit 502 is changing toward improvement or deterioration based on the current shape deviation performance data 504 and the shape deviation prediction data 505, and obtains the control operation amount output permission data S4.

[0219] Specifically, the shape correction quality determination unit 502 performs shape correction quality determination as follows. First, as shown in the specifications A and B regarding shape control priority, a weight coefficient w(i) in the width direction is pre-set in the output determination database DB3 for each of specifications A1 and A2 to take into account control priority in the sheet width direction. Using this weight coefficient, the quality of the shape change is determined using, for example, an evaluation function J such as the following [Formula 4]. In [Formula 4], w(i) is the weight coefficient, εfb(i) is the actual shape deviation data 504, εest(i) is the predicted shape deviation data 505, i is the shape detector area, and rand is the random number term.

[0220] [Formula 4]

[0221]

[0222] When using the evaluation function J in [Equation 4], the evaluation function J is positive when the shape improves and negative when it deteriorates. Furthermore, rand is a random number term that randomly changes the evaluation result of the evaluation function J. This allows the evaluation function J to be positive even when the shape deteriorates, making it possible to learn the relationship between the shape pattern and the control method even when the rolling phenomenon model 501 is incorrect.

[0223] Here, the random number term rand is changed as appropriate: when the model of the controlled plant 1 is unreliable, such as at the beginning of a test run, the maximum value is increased, and when it is desired to learn the control method to some extent and perform stable control, the value is set to 0.

[0224] The shape correction quality determination unit 502 calculates the evaluation function J, and outputs permission data S4=1 (yes) as the control operation amount when J≥0, and outputs permission data S4=0 (no) as the control operation amount when J<0, and outputs the control operation amount permission data S4.

[0225] As already explained, the control output quality judgment rule execution unit 17 receives the normalized shape deviation 201, the shape deviation level 202, and the selected control operation variable S8 as inputs, and outputs a control result quality judgment estimate S11. The control result quality judgment estimate S11 takes a value of 1 if the control result is estimated to be good, and takes a value of 0 otherwise.

[0226] The control output suppression unit 4 determines whether to output the control manipulated variable output SO to the controlled plant 1 based on the control manipulated variable output permission data S4, which is the determination result of the control output determination unit 5, and the control result quality determination estimated value S11. The control manipulated variable output permission data S4 includes the position change outputs #1 to #nAS-U, the upper first intermediate shift position change output, and the lower first intermediate shift position change output, and is determined by the following conditions.

[0227] IF (control method selection flag = 1) THEN

[0228] IF (Control operation quantity output permission data S4 = 0 OR Control result quality judgment estimated value S11 ≤ thprot) THEN

[0229] #1~#nAS-U position change output=0

[0230] Upper first intermediate shift position change output = 0

[0231] Lower first intermediate shift position change output = 0

[0232] ELSE

[0233] #1~#nAS-U position change output = #1~#nAS-U position change

[0234] Upper first intermediate shift position change output = Upper first intermediate shift position change

[0235] Lower first intermediate shift position change output = Lower first intermediate shift position change

[0236] ENDIF

[0237] ELSE

[0238] IF((control operation variable output permission data S4=0 OR control result quality judgment estimated value S11≤thprot) AND (PTRIAL<η)) THEN

[0239] #1~#nAS-U position change output=0

[0240] Upper first intermediate shift position change output = 0

[0241] Lower first intermediate shift position change output = 0

[0242] ELSE

[0243] #1~#nAS-U position change output = #1~#nAS-U position change

[0244] Upper first intermediate shift position change output = Upper first intermediate shift position change

[0245] Lower first intermediate shift position change output = Lower first intermediate shift position change

[0246] ENDIF

[0247] ENDIF

[0248] Here, thprot sets a baseline value for applying output suppression based on the estimated value of the control result quality judgment. Specifically, it is believed that the accuracy of the quality judgment estimation is also low during the initial startup period when the operating data of the factory equipment is insufficient. Therefore, the baseline value is lowered in advance, and output suppression based on the estimated quality judgment is not applied much.

[0249] On the other hand, after sufficient operational performance data has been accumulated and the accuracy of the quality judgment has increased, the reference value is increased to enhance the effectiveness of output suppression for the quality judgment estimation based on the control results. The quality judgment accuracy is determined based on the verification results of the estimated accuracy of the currently used quality judgment rule, received from the quality judgment rule accuracy verification unit 34 in the quality judgment rule learning unit, S15.

[0250] Furthermore, η is a variable that takes a random value between 0 and 1, and PTRIAL represents the probability of disabling output suppression and outputting a new search operation to the plant. When the control method selection flag S14 is 0, including when verifying the effectiveness of the control method in an unknown region, output suppression to the plant is ignored with a certain probability, and output is performed to the plant.

[0251] The control execution unit 20 performs the aforementioned calculations based on performance data Si from the controlled plant 1 (rolling mill), and executes shape control by outputting the control operation variable output SO to the controlled plant 1 (rolling mill). Furthermore, the control method learning unit 21 utilizes the data used by the control execution unit 20.

[0252] Next, the operation performed by the learning data creation unit 801 will be described.

[0253] like Figure 1 As shown, the learning data production unit 801 produces supervision data S7a for the neural network 111 used in the control rule learning unit 802 based on the control result quality judgment estimated value S11 from the control output quality judgment rule execution unit 17, the control operation terminal operation instruction S2, the selected control operation amount S8, the control method selection flag S14, and the judgment result of the control output inhibition unit (control operation amount output possibility data S4).

[0254] In this case, the supervisory data S7a becomes Figure 7 The output from the output layer of the neural network 111 is shown as the AS-U operation level 301 and the first intermediate operation level 302. The learning data generation unit 7 uses the control operation terminal operation instruction S2 (AS-U operation level 301, first intermediate operation level 302) as the output of the neural network 101 and the #1 to #n AS-U position change amount outputs, the upper first intermediate shift position change amount output, and the lower first intermediate shift position change amount output as the selected control operation amount S8 to generate supervision data S7a for the neural network 111 used in the control rule learning unit 802.

[0255] When explaining the operation of the learning data generating unit 801, Figure 14 Shown in Figure 10 The relationship between the data and reference numerals of each component in the control output calculation unit 3 is shown. Here, the AS-U operation level 301 is representatively shown for the control operation terminal operation instruction S2 as the output of the neural network 101, the data on the positive side of the operation level is denoted as OPref, the data on the negative side of the operation level is denoted as OMref, the conversion gain is denoted as G, and the control operation amount S3 is denoted as Cref.

[0256] The difference between the positive and negative operation degree data OPref and OMref is obtained by a subtractor 701 and multiplied by a conversion gain G by a multiplier 702 to obtain a control operation variable output Cref. This control operation variable output Cref is supplied to the control output operation method selection unit 18, which obtains the selected operation command value C"ref.

[0257] For simplicity, the randomly generated operation degree from the positive and negative sides of the operation degree, and from the control operation disturbance generating unit 16, is used as the output from the output layer of the neural network 101 of the control rule execution unit 10 as the operation degree random number. Furthermore, the control operation variable output SO to the control operation terminal is used as the operation command value.

[0258] Figure 15 The processing stages and processing contents in the learning data creation unit 7 are shown.

[0259] In the first processing stage 71 , the operation command value C″ ref refers to the selected control operation amount S8 which is the output value of the control output operation method selection unit 18 .

[0260] In the next processing stage 72, the operation command value Cref is corrected based on the control result quality judgment estimate S11 and the control operation variable output permission data S4, and is set to C'ref. Specifically, when the control result quality judgment estimate S11 = 0 or the control operation variable output permission data S4 = 0, the correction value C'ref of the operation command value C"ref is obtained using the following [Formula 5]. When the control result quality judgment estimate S11 = 1 and the control operation variable output permission data S4 = 1, the correction value C'ref is obtained using the following [Formula 6].

[0261] [Formula 5]

[0262] IFC”ref>OTHEN C’ref=C”ref-Δcref

[0263] IFC”ref<OTHEN C’ref=C”ref+Δcref

[0264] [Formula 6]

[0265] IFC”ref>OTHEN C’ref=C”ref+Δcref

[0266] IFC”ref<OTHEN C’ref=C”ref-Δcref

[0267] In the processing stage 73 , the operation degree correction amount ΔOref is calculated using [Formula 7] and [Formula 8] based on the corrected operation command value C′ref.

[0268] [Formula 7]

[0269] C'ref=G-((OPref+ΔOref)-(OMref-ΔOref))

[0270] [Formula 8]

[0271]

[0272] In the processing stage 74, the supervisory data OP′ref and OM′ref for the neural network 111 are obtained by [Formula 9].

[0273] [Formula 9]

[0274]

[0275] Thus, in the learning data production unit 7, as shown in FIG. Figure 14 As shown, based on the control result quality judgment estimated value S11 of the control output quality judgment rule execution unit 17 and the control operation variable output permission data S4 of the control output suppression unit 4, the operation command value correction value C'ref is calculated for the operation command value C"ref actually output to the control target plant 1.

[0276] Specifically, when the control result quality determination estimated value S11 = 1 and the control operation variable output permission data S4 = 1, if it is determined that the operation is good, the operation command value is increased in the same direction by Δcref.

[0277] Conversely, if the control result quality judgment estimate S11 = 0 or the control manipulation variable output permission data S4 = 0, the operation is judged to be unacceptable, and the operation command value is reduced in the opposite direction by Δcref. Since the conversion gain G is predetermined and known, the correction amount ΔOref can be calculated simply by knowing the values ​​of the positive and negative manipulation degrees. Here, ΔCref is set to an appropriate value in advance through simulations, etc. Through the above steps, the supervisory data OP'ref and OM'ref used in the control rule learning unit 802 can be calculated using [Equation 9].

[0278] In addition, Figure 14 Although a simple example is used to illustrate this, in reality, all of these steps are performed with respect to the AS-U operation degree 301 for #1 to #nAS-U and the first intermediate operation degree 302 for the upper first intermediate roller shift and the lower first intermediate roller shift, and are set as the supervision data (AS-U operation degree supervision data, first intermediate operation degree supervision data) of the neural network 111 used in the control rule learning unit 802.

[0279] Figure 16 An example of data stored in the learning data database DB2 is shown.

[0280] Learning the neural network 111 requires multiple combinations of input data S8a and supervisory data S7a. Therefore, the supervisory data S7a generated by the learning data generation unit 7 is combined with the input data S1 (S8a) input to the control rule execution unit 10 by the control execution unit 20 and stored as a set of learning data in the learning data database DB2. The supervisory data S7a here is the AS-U operation level supervisory data and the first intermediate operation level. Furthermore, the input data S1 (S8a) is the normalized shape deviation 201 and the shape deviation stage.

[0281] also, Figure 1 The factory equipment control system uses various databases DB1, DB2, DB3, and DB4, but each database DB1, DB2, DB3, and DB4 is managed and operated in a coordinated manner through a neural network management table TB.

[0282] Figure 17 It shows the structure of the neural network management table TB.

[0283] The neural network management table TB is divided according to the specifications (B1) plate width, (B2) steel type, and the specifications A1 and A2 for control priority. As the (B1) plate width, for example, four categories are used: 3-foot width, meter width, 4-foot width, and 5-foot width. As the steel type, ten categories are used: steel type (1) to steel type (10). Furthermore, regarding the specification A for control priority, two categories are used: A1 and A2. In this case, there are 80 categories, and 80 neural networks are used according to the rolling conditions.

[0284] The neural network learning control unit 112 follows Figure 17 The neural network management table TB shown in FIG. Figure 16 The combination of input data and supervision data, i.e. learning data, is associated with the corresponding neural network No. and stored in Figure 18 The learning data database DB2 as shown.

[0285] Each time the control execution unit 20 executes shape control on the controlled plant 1, it generates two sets of learning data. This is because, for the same input data and control output, two evaluation criteria—specifications A1 and A2—are used to determine the quality of the control results. Therefore, two types of supervisory data are generated. Once a certain amount of supervisory data has been accumulated (e.g., 200 sets) or new data has been added to the learning data database DB2, the neural network learning control unit 112 instructs the neural network 111 to begin learning.

[0286] In the control rule database DB1, according to Figure 17The management table TB shown in the figure stores a plurality of neural networks. The neural network learning control unit 112 specifies the neural network number to be learned, and the neural network selection unit 113 extracts the neural network from the control rule database DB1 and sets it as the neural network 111.

[0287] The neural network learning control unit 112 instructs the input data generation unit 114 and the supervisory data generation unit 115 to retrieve the input data and supervisory data corresponding to the corresponding neural network from the learning data database DB2 and use these data to perform learning of the neural network 111. Various methods have been proposed for learning neural networks, and any method can be used.

[0288] When the learning of the neural network 111 is completed, the neural network learning control unit 112 writes the neural network 111 as the learning result back to the position of the corresponding neural network No. in the control rule database DB1, thereby completing the learning.

[0289] Learning can be targeted at Figure 17 All neural networks defined as shown may be executed simultaneously at fixed time intervals (e.g., every day), or only the neural network No. for which a certain amount (e.g., 100 sets) of new learning data has been accumulated may be trained at that time point.

[0290] Next, the operation of the quality determination rule learning unit 22 will be described.

[0291] The quality judgment rule learning unit 22 uses the time delay data of the data used in the control execution unit 20. Here, the time delay Z -1 e-TS stands for delay by a preset time T.

[0292] Since the controlled plant 1 has a time response, there is a time delay until the actual data changes according to the control operation variable output SO. Therefore, learning is performed using the actual data at the time when the delay time T has passed after the control operation is performed.

[0293] In shape control, it takes several seconds for the shape meter to detect a shape change after an operation command is issued to the AS-U or the first intermediate roll. Therefore, it is preferable to set T to approximately 2 to 3 seconds. Furthermore, the delay time also varies depending on the type of shape detector and the rolling speed. Therefore, it is preferable to set T to the optimal time until a change in the control operation terminal results in a shape change.

[0294] Figure 19 1 and 2 show the operation of the control result quality judging unit 6. The shape change quality judging unit 602 uses the quality judging evaluation function Jc shown in [Formula 10].

[0295] [Formula 10]

[0296]

[0297] In [Equation 10], εfb(i) represents the actual shape deviation performance data included in the performance data Si, εlast(i) represents the previous value of the shape deviation performance data, and wC(i) represents the weight coefficient in the width direction used for quality judgment. The weight coefficient wC(i) for quality judgment is set from the quality judgment database DB4 based on the control priority specifications A1 and A2. The quality of the control result is determined using the quality judgment evaluation function Jc.

[0298] Based on the threshold condition (LCU ≥ 0 ≥ LCL), an upper threshold limit LCU and a lower threshold limit LCL are pre-set. At this time, if the comparison result with the quality evaluation function Jc is Jc > LCU, the control result quality data S6 = 0 (shape deteriorates). If Jc < LCL, the control result quality data S6 = 1 (shape improves).

[0299] Thus, the weight coefficient wC(i) in the width direction changes depending on the control priority specifications A1 and A2, resulting in different evaluation functions Jc. Consequently, it is expected that the control result evaluation data S6 will also be different. Therefore, the evaluation rule learning unit 22 performs evaluations on the control result evaluation data S6 for both control priority specifications A1 and A2.

[0300] This control result quality data S6 is used as it is as supervisory data S13 a for the neural network 311 used in the quality determination rule learning unit 31 .

[0301] Figure 20 An example of data stored in the learning data database DB6 is shown.

[0302] To learn the neural network 311, multiple combinations of input data S12a and supervisory data S13a are required. Therefore, the supervisory data S13a (control result quality data) generated by the control result quality determination unit 6 is combined with the time delay data S12a of the input data S1 (normalized shape deviation 201 and shape deviation stage) input from the control execution unit 20 to the control rule execution unit 10, and the combined data is stored as a set of learning data in the learning data database DB6.

[0303] At this time, the learning data is stored in the verification data database DB7 at a certain ratio, and can be used for the verification of the quality judgment rule in the quality judgment rule accuracy verification unit 34 .

[0304] The quality assessment rule accuracy verification unit 34 includes a neural network that performs calculations in only one direction, similar to the control output quality assessment rule execution unit 17. The quality assessment rule accuracy verification unit 34 then retrieves test data from the verification data database DB7 and calculates the error between the output data obtained by inputting the neural network with the input data of the test data and the output data of the test data. For example, the quality assessment rule accuracy verification unit 34 calculates the average of the errors across all test data as the quality assessment rule accuracy S15 for the quality assessment rule.

[0305] also, Figure 1 The factory equipment control system uses various databases DB5, DB6, but in Figure 21 1 shows the structure of the neural network management table TB for managing and operating the databases DB5 and DB6 in a coordinated manner. That is, the management table TB includes a standardized management table.

[0306] Specifically, if Figure 21 As shown, the management table TB is divided according to the specifications (B1) plate width, (B2) steel type, and the specifications A1 and A2 for control priority. As the (B1) plate width, for example, four categories are used: 3-foot width, meter width, 4-foot width, and 5-foot width. As the steel type, ten categories are used: steel type (1) to steel type (10). In addition, the specification A for control priority is set to two categories: A1 and A2. In this case, there are 80 categories, and 80 neural networks are used according to the rolling conditions.

[0307] The neural network learning control unit 312 follows Figure 21 The neural network management table TB will Figure 20 The combination of input data and supervision data, i.e. learning data, is associated with the corresponding neural network No. and stored in Figure 22 The learning data database DB6 as shown.

[0308] Each time the control execution unit 20 executes shape control on the controlled plant 1, it generates two sets of learning data. This is because, for the same input data and control output, two evaluation criteria—specifications A1 and A2—are used to determine the quality of the control results. Therefore, two types of supervisory data are generated. Once a certain amount of supervisory data has accumulated (e.g., 200 sets) or new supervisory data has been added to the learning data database DB6, the neural network learning control unit 312 instructs the neural network 311 to begin learning.

[0309] The quality judgment rule database DB5 is based on Figure 21The management table TB shown in the figure stores multiple neural networks. Then, the neural network learning control unit 312 specifies the neural network number to be learned, and the neural network selection unit 313 retrieves the corresponding neural network from the quality judgment rule database DB5 and sets it as the neural network 311. The neural network learning control unit 312 retrieves the input data and supervisory data corresponding to the corresponding neural network from the learning data database DB6, and instructs the input data creation unit 314 and supervisory data creation unit 315 to use this data to execute the training of the neural network 311. Various methods have been proposed for neural network learning, and any method can be used.

[0310] When the learning of the neural network 311 is completed, the neural network learning control unit 312 writes the neural network 311 as the learning result back to the position of the neural network No. in the control rule database DB6, thereby completing the learning.

[0311] For Figure 21 All neural networks defined in the management table TB shown above are trained simultaneously at fixed time intervals (e.g., every day). Alternatively, only neural networks with neural network numbers that have accumulated a certain amount (e.g., 100 sets) of new training data may be trained at that time.

[0312] Furthermore, by including rolling performance, steel grade, and plate width in the input data for the quality judgment rules, it is possible to learn the differences in the quality judgment criteria using a single neural network. In this case, there is no need to switch the quality judgment rules according to the rolling conditions when executing them.

[0313] As described above, to improve the control rules for the controlled plant 1, if a control operation with good control results cannot be learned, the control operation is significantly modified. Alternatively, if the control results are good, a new control operation method is adopted. Furthermore, if a control operation with good control results can be learned, the control operation is not modified or is modified only slightly. Then, if the control results for these actions are good, it is effective to adopt them as new control operation methods.

[0314] In addition, by learning the combination of shape patterns, control operations, and the quality of control results based on actual machine data, it is possible to build a model that can infer the quality of control results with high precision compared to simulators that use machine models. Through regular automatic learning, it is possible to build a model that is always suitable for the latest factory equipment status.

[0315] Furthermore, the use of a model for determining the quality of estimated control results can improve the reliability of the control output suppression function for plant equipment, which was conventionally performed using only a simple mechanical model.

[0316] Furthermore, while control rule learning data, which was previously generated during a single control result quality assessment, is generated using a model that estimates the quality of control results, this example mitigates the effects of noise contained in plant equipment data and allows even minor adjustments with minimal effects to be included in the learning data. Furthermore, this example prevents erroneous control effect assessments, suppresses fluctuations in the learning data, and stabilizes control performance.

[0317] Furthermore, the control rule database DB1 stores the neural network used by the control execution unit 20. If the stored neural network only performs initial processing using random numbers, it takes time to learn the neural network until appropriate control is possible. Therefore, when the control unit is constructed for the controlled plant 1, control rule learning is performed in advance through simulation based on the control model of the controlled plant 1 determined at that time. By then storing the learned neural network in the simulator in the database, a certain level of control performance can be achieved from the initial startup of the controlled plant.

[0318] Alternatively, by enabling the quality judgment rule learning unit 22 to learn the quality judgment rule based on the actual performance data of the operation data in the actual machine, the control rule can be learned even without controlling the actual machine, and a certain degree of performance control can be performed before being applied to the control object factory equipment.

[0319] Figure 23 The plant equipment control system of this example is configured to include a control rule evaluation unit 23 that performs control rule evaluation processing.

[0320] The control rule evaluation unit 23 includes a control rule quality determination data collection unit 35 , a control rule evaluation data calculation unit 36 ​​, a control rule database update unit 37 , a control rule evaluation data database DB8 , and a control rule evaluation value database DB9 .

[0321] The control rule quality assessment data collection unit 35 receives the control output quality assessment estimate S9 from the control output quality assessment rule execution unit 17 and the quality assessment rule accuracy S15 from the quality assessment rule accuracy verification unit 34. The control rule quality assessment data collection unit 35 then stores the control rule quality assessment data S16 in the control rule evaluation data database DB8, along with the control rule number used by the control execution unit 20. The control rule quality assessment data S16 is the control output quality assessment estimate S9. However, if the quality assessment rule accuracy S15 is below a certain value, it is not stored in the database DB8.

[0322] Each time the control execution unit 20 performs a control output calculation using a control rule, new control rule quality determination data S16 is generated. The generated control rule quality determination data S16 is stored in the control rule evaluation data database DB8. In this case, since a large amount of data is stored for each control rule, the control rule evaluation data database DB8 predetermines an upper limit on the data stored for each control rule. When the upper limit exceeds a certain value, old data is deleted and new data is stored.

[0323] The control rule evaluation data calculation unit 36 ​​retrieves the control rule quality judgment data S17 accumulated for each control rule from the control rule evaluation data database DB8 and calculates the average value thereof as control rule evaluation data S18. The average value thus obtained corresponds to the evaluation value.

[0324] The control rule evaluation data S18 calculated by the control rule evaluation data calculation unit 36 ​​is stored in the control rule evaluation value database DB 9. However, if the number of control rule quality determination data is less than a certain number, the reliability of the evaluation value is low, so the evaluation result is not stored.

[0325] In the database management table TB, neural network numbers (control rules) used according to the conditions are registered one by one. In contrast, the control rule evaluation value database DB9 manages the evaluation values ​​of multiple control rules. The control rule database update unit 37 refers to the control rule evaluation value database DB9, compares the control rule evaluation value of the neural network number (control rule) registered in the database management table TB with the control rule evaluation values ​​of other control rules applicable to the conditions, and updates the control rule with the highest evaluation value as the neural network number (control rule) in the database management table TB.

[0326] Figure 23 Other parts of the plant equipment control system shown are Figure 1 The plant equipment control system shown is similarly constructed. Figure 23 In the illustrated plant control system, control rule evaluation unit 23 performs evaluation based on past performance data of the controlled plant 1. Therefore, control execution unit 20 does not need to actually control the controlled plant 1. Specifically, control output suppression unit 4 does not need to supply control output S0 to the controlled plant 1.

[0327] According to the Figure 23The plant control system shown here sets the control rule to be evaluated in the control rule execution unit 10 and assigns past performance data as Si. This allows the control rule evaluation value database DB9 to be updated even if no control output is actually performed on the controlled plant 1.

[0328] Modifications

[0329] The present invention is not limited to the above-described embodiments, but includes various modifications. For example, the above-described embodiments are described in detail to facilitate understanding of the present invention, and are not necessarily limited to having all the described configurations.

[0330] For example, Figure 1 、 Figure 23 The plant equipment control system shown is configured to include a processing unit that performs data creation, learning, control, and other processing. Figure 1 、 Figure 23 The control execution unit 20, control method learning unit 21, quality determination rule learning unit 22, and control rule evaluation unit 23 shown may be composed of programs (software) that enable a processor to realize their respective functions, and the computer may execute the programs. Figure 24 An example of the configuration of a computer in this case is shown.

[0331] That is, Figure 24 As shown, the computer constituting each unit 20 to 23 includes a CPU (Central Processing Unit) a, a ROM (Read Only Memory) b, and a RAM (Random Access Memory) c connected to a bus, a nonvolatile memory d, and a network interface e.

[0332] The CPUa is a processing unit that reads the program code for the software that executes the processing in each unit 20 to 23 from the ROMb and executes it. Variables and parameters generated during the processing are temporarily written to the RAMc. The nonvolatile memory d, using a high-capacity information storage unit such as an HDD (Hard Disk Drive) or SSD (Solid State Drive), stores programs executed by each unit 20 to 23 and data from various databases.

[0333] Furthermore, the units 20 to 23 may be constituted by separate computers, but the respective programs may be installed in a small number of computers such as one and executed simultaneously.

[0334] The network interface e uses, for example, a NIC (Network Interface Card) and performs data transmission and reception with other units and the controlled plant equipment 1 .

[0335] Information such as programs that realize each processing function in this case can be stored in a nonvolatile memory d such as HDD or SSD, or in a recording medium such as a memory, an IC card, an SD card, or an optical disk.

[0336] Furthermore, part or all of the functions performed by the units 20 to 23 may be realized by hardware such as an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit).

[0337] In addition, Figure 1 、 Figure 23 In the block diagrams shown in FIG. 1 , only the control lines and information lines are shown as necessary for explanation, and not all control lines and information lines on the product are shown. In fact, it can be assumed that almost all the components are connected to each other.

[0338] In addition, in the above-described embodiment, the control target plant 1 is applied to a Sendzimir rolling mill. However, the present invention can be applied to the control of various other plant equipment. The control rules for the case of application to a Sendzimir rolling mill are also shown as an example, and the present invention is not limited to the above-described embodiment.

Claims

1. A plant control system for performing control on a plant to be controlled by recognizing a pattern of combinations of performance data of the plant to be controlled, characterized in that: The factory equipment control system has: a control method learning unit that learns a combination of performance data and control operations of the control target plant equipment; a control execution unit that executes control of the control target plant equipment based on a combination of the performance data and the control operation learned by the control method learning unit; as well as A quality judgment rule learning unit learns the combination of the performance data of the control target plant equipment and the control operation and the quality of the control result. The control execution unit comprises: a control rule execution unit that provides a control output according to a combination of performance data of the controlled plant and a determination of a control operation; a control output quality determination rule execution unit for determining the quality of the control output based on a combination of performance data of the control target plant equipment, control operations, and control result quality determination; a new search operation amount calculation unit that calculates a new search operation amount based on the quality judgment of the control output quality judgment rule execution unit; as well as a control output suppressing unit that uses the performance determination made by the control output performance determination rule executing unit and simulation data using a control simulator to suppress output of the control output to the control target plant if it is determined that the performance data of the control target plant has deteriorated when the control output is output to the control target plant; The superiority and inferiority judgment rule learning unit has: a control result quality determination unit configured to determine the quality of the control result after a time delay until the control effect is reflected in actual performance data when the control execution unit outputs the control output to the controlled plant equipment; as well as a quality judgment rule learning unit that performs learning using the quality of the control result in the control result quality judgment unit, the actual performance data, and the control output as learning data; The control method learning unit comprises: a learning data generating unit that obtains supervisory data using the control output and the control output as determined by the control output quality judgment rule executing unit; and The control rule learning unit performs learning using the actual performance data and the supervisory data as learning data.

2. The factory equipment control system according to claim 1, characterized in that: The control method learning unit performs learning to obtain different combinations of performance data and control operations for a plurality of control targets according to the state of the control target plant equipment. The obtained combination of the performance data and the control operation is used as the combination of the performance data and the control operation determined by the control rule execution unit for the control target plant.

3. The factory equipment control system according to claim 1, characterized in that: The control output quality judgment rule execution unit stores a combination of the performance data of the control target plant equipment, the control operation, and the control result quality determination as a first neural network. The quality judgment rule learning unit stores the combination of performance data, control operation and quality of control results as a second neural network. The second neural network obtained as a result of learning in the grade determination rule learning unit is used as the first neural network in the grade determination rule execution unit.

4. The factory equipment control system according to claim 1, characterized in that: The quality judgment rule learning unit is provided with a quality judgment rule accuracy verification unit. The criterion of output suppression using the quality of the control result in the control output suppression unit is changed using the quality of the judgment rule generated by the quality judgment rule accuracy verification unit.

5. The plant equipment control system according to any one of claims 1 to 4, characterized in that: The plant equipment control system further comprises a control rule evaluation unit, The control rule evaluation unit has: a control rule quality judgment data collection unit for storing the quality judgment data of the quality judgment rule execution unit of the control execution unit and the accuracy verification results of the quality judgment rules obtained by the quality judgment rule learning unit in a database; as well as a control rule evaluation data calculation unit that calculates control rule evaluation data based on the quality judgment data accumulated in the database and the accuracy verification result of the quality judgment rule; The plant control system executes the evaluation of the control rule used by the control execution unit without outputting it to the controlled plant.

6. A method for controlling plant equipment, comprising: identifying a pattern of combinations of performance data of a control target plant equipment, and controlling the control target plant equipment by a computer, wherein: The processing executed by the computer includes: a control method learning process for learning a combination of performance data and control operations of the control target plant equipment; a control execution process for executing control of the control target plant equipment based on a combination of the performance data and the control operation learned by the control method learning process; and The quality judgment rule learning process learns the combination of the performance data of the control target plant equipment and the control operation and the quality of the control result. The control execution process includes: A control rule execution process for providing a control output based on a combination of performance data of the control target plant and a determination of a control operation; a control output quality determination rule execution process for determining the quality of the control output based on a combination of performance data of the control target plant equipment, control operations, and control result quality determination; A new search operation amount calculation process is performed to calculate a new search operation amount based on the quality judgment of the control output quality judgment rule execution process; and a control output suppression process that uses the quality judgment of the control output quality judgment process executed based on the control output quality judgment rule and simulation data using a control simulator to prevent the control output from being output to the control target plant if it is determined that the performance data of the control target plant has deteriorated when the control output is output to the control target plant; The quality judgment rule learning process includes: a control result quality determination process for determining the quality of the control result after a time delay until the control effect is reflected in actual performance data when the control output is output to the controlled plant equipment by the control execution process; and a quality judgment rule learning process, wherein the quality of the control result in the control result quality judgment process, the actual performance data, and the control output are used as learning data for learning; The control method learning process includes: a learning data production process, wherein the quality of the control output is judged by executing the control output quality judgment rule, and supervisory data is obtained using the control output; and The control rule learning process performs learning using the actual data and the supervisory data as learning data.

7. A computer-readable recording medium storing a program for identifying a pattern of combinations of performance data of a control target plant equipment and causing a computer to execute plant equipment control, wherein: The program causes the computer to execute: a control method learning step of learning a combination of performance data and control operations of the control target plant equipment; a control execution step of executing control of the control target plant equipment based on a combination of the performance data and the control operation learned in the control method learning step; as well as The quality judgment rule learning step is to learn the combination of the performance data of the control target plant equipment and the control operation and the quality of the control result. The control execution step includes: a control rule execution step of providing a control output based on a combination of performance data of the control target plant equipment and a determination of the control operation; a control output quality determination rule execution step, performing quality determination of the control output based on a combination of actual performance data of the controlled plant equipment, control operations, and determination of quality of control results; A new search operation amount calculation step, calculating a new search operation amount according to the quality judgment of the control output quality judgment rule execution step; and a control output suppression step, using the quality judgment of the control output quality judgment rule execution step and simulation data using a control simulator, and, when outputting the control output to the control target plant, determining that the actual performance data of the control target plant has deteriorated, preventing the control output from being output to the control target plant; The step of learning the quality judgment rules includes: a control result quality determination step of determining the quality of the control result after a time delay until the control effect is reflected in the actual performance data when the control output is output to the controlled plant equipment in the control execution step; and a quality judgment rule learning step, wherein the quality of the control result in the control result quality judgment step, the actual performance data, and the control output are used as learning data for learning; The control method learning step includes: a learning data preparation step of obtaining supervisory data using the control output quality judgment of the control output quality judgment rule execution step and the control output; and The control rule learning step performs learning using the actual performance data and the supervisory data as learning data.

Images