Control system

Abstract

To provide a control system capable of quickly grasping a cause of a failure or abnormality.SOLUTION: A control system according to an embodiment includes: a control device that outputs an open / close command to control an electromagnetic valve that controls opening / closing of a spray nozzle for cooling a steel plate to the electromagnetic valve; a camera that images a state of cooling water sprayed from the spray nozzle and generates image data; an image processing device that performs image processing of the image data, determines whether or not there is an abnormality in the sprayed cooling water, and outputs a determination result; a performance data collection device that collects the open / close command, the determination result, a drive current past record of the electromagnetic valve, and flow rate past record data of the cooling water flowing through piping of the electromagnetic valve being opened and closed, synchronizes them with the time, and outputs time-series data; and a machine learning device that generates a database by classifying a content of the failure or abnormality on the basis of the time-series data and generates a defect determination model to be updated.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] Embodiments of the present invention relate to a control system for detecting a failure or abnormality of a cooling system that cools a steel plate in a run-out table of a hot rolling line.

Background Art

[0002] Conventionally, failures and abnormalities in the cooling system have been detected by visually monitoring the states of pressure sensors and flow sensors commonly installed in a plurality of electromagnetic valves, or the ejection state of cooling water from spray nozzles. In recent years, a technique has been developed in which the ejection state of cooling water ejected from each spray nozzle is photographed by a camera and judged as good or bad by an image processing device (for example, Patent Document 1). In this technique, when the ejection state of the cooling water is determined to be abnormal, the control device disconnects the electromagnetic valve that has caused the failure or abnormality from the control system and opens other electromagnetic valves to adjust the flow rate of the cooling water.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Since the cooling process of the steel sheet in the run-out table is carried out at the end of the hot rolling line, if a failure or abnormality occurs in the cooling system during the cooling process, the impact on the entire line is significant. In order to suppress the stoppage of the hot rolling line and the occurrence of defective products, it is required to grasp in advance the deterioration state of electromagnetic valves and the like in the cooling system and take countermeasures before a failure or abnormality occurs. Also, even if a failure or abnormality occurs in the cooling system, it is important to repair the failure or abnormality in a short period of time without causing a decrease in the operating rate of the cooling system. Since there can be multiple types of failures and abnormalities, not only the pass / fail judgment but also recording the content of the failure or abnormality and referring to the record can enable quick recovery from the failure or abnormality.

[0005] An embodiment of the present invention aims to provide a control system that enables quick grasping of the cause of a failure or abnormality.

Means for Solving the Problem

[0006] The control system according to an embodiment of the present invention includes a control device that outputs a plurality of opening / closing commands for controlling a plurality of electromagnetic valves that respectively control the ejection of cooling water from a plurality of spray nozzles arranged along a run-out table on which a steel plate is conveyed, a camera that sequentially captures the state of the ejection of cooling water from the plurality of spray nozzles to generate image data, an image processing device that performs image processing on the image data to determine the ejection state of the cooling water of each of the plurality of spray nozzles and outputs one or more determination results, a data collection device that respectively detects the plurality of opening / closing commands, the flow rates of the cooling water supplied to the plurality of spray nozzles, a plurality of flow rate achievements output by a plurality of flow sensors, a plurality of drive current achievements flowing through the plurality of electromagnetic valves respectively, and collects a plurality of time-series data in which the determination results are time-synchronized, and a machine learning device that generates a defect determination model that classifies the failure or abnormality of the one or more determination results based on the plurality of time-series data to generate a database. The machine learning device classifies the failure or abnormality of the one or more determination results based on the variations over time of the cluster data including each of the plurality of opening / closing commands, each of the plurality of flow rate achievements, each of the plurality of drive current achievements, and the determination results.

Advantages of the Invention

[0007] According to an embodiment of the present invention, it is possible to provide a control system that can quickly identify the cause of a failure or abnormality.

Brief Description of the Drawings

[0008]

Figure 1

Figure 2

Embodiments for Carrying Out the Invention

[0009] Hereinafter, embodiments will be described with reference to the drawings. Note that the drawings are schematic or conceptual, and the relationships between the thickness and width of each part, the size ratios between parts, etc. are not necessarily the same as the actual ones. Also, even when representing the same part, there may be cases where the dimensions and ratios are shown differently in the drawings. In the specification and each figure of the present application, the same reference numerals are given to the same elements as those described above with respect to the previously shown figures, and the detailed description thereof will be omitted as appropriate.

[0010] FIG. 1 is a schematic block diagram illustrating a control system according to an embodiment. As shown in FIG. 1, a control system 100 according to an embodiment includes cameras 14 and 15, a control device 11, an image processing device 16, a data collection device 17, and a machine learning device 18. In FIG. 1, in addition to the control system 100, the configuration of the cooling system in the run-out table 2 is also shown. Although not shown in FIG. 1, for example, a finishing rolling mill is provided upstream of the run-out table 2. The run-out table 2 conveys the steel plate 1 finished-rolled by the finishing rolling mill, and the steel plate 1 is cooled by the cooling system.

[0011] First, with reference to FIG. 1, the configuration of the cooling system of the run-out table 2 will be described. The run-out table 2 is composed of a plurality of table rolls. The plurality of table rolls are arranged along the conveyance direction of the steel plate 1. In FIG. 1, the conveyance direction of the steel plate 1 is indicated by a white arrow. In the plurality of table rolls, the direction of the rotation axis is set so as to be orthogonal to the conveyance direction.

[0012] The cooling system has a plurality of spray nozzles 5, 6, 7 arranged along the conveyance direction of the steel plate 1 and a plurality of electromagnetic valves 3, 4. The spray nozzles 5, 6, 7 and the electromagnetic valves 3, 4 are provided in the piping and control the ejection of the cooling water 30 flowing through the piping.

[0013] The spray nozzles 5 and 6 are arranged along the conveyance direction above the pass line along which the steel plate 1 is conveyed. The spray nozzles 5 and 6 eject the cooling water 30 from above the steel plate 1 and cool the conveyed steel plate 1 from the upper surface. The spray nozzle 7 is arranged along the conveyance direction below the pass line. The spray nozzle 7 ejects the cooling water 30 from below the steel plate 1 and cools the conveyed steel plate 1 from the lower surface.

[0014] The electromagnetic valve 3 is provided in the middle of the pipe through which the cooling water 30 can flow. In this example, it is provided for every two spray nozzles 5 and 6. The electromagnetic valve 4 is provided in the middle of the pipe through which the cooling water 30 can flow. In this example, it is provided for each spray nozzle 7. The electromagnetic valves 3 and 4 are electrically connected to the control device 11, open the pipe path according to the open command output from the control device 11, and close the pipe path according to the close command.

[0015] In this example, on the upper surface side of the steel plate 1, the opening and closing of the electromagnetic valve 3 are selected and set for every two spray nozzles 5 and 6, and on the lower surface side of the steel plate 1, the opening and closing of the electromagnetic valve 4 are selected and set for each spray nozzle 7.

[0016] The flow rate sensor 12 is provided for each pipe provided with the spray nozzles 5 and 6 and the electromagnetic valve 3. The flow rate sensor 13 is provided for each pipe provided with the spray nozzle 7 and the electromagnetic valve 4. The flow rate sensors 12 and 13 are electrically connected to the control device 11, detect the flow rate of the cooling water 30 flowing through the pipe, and output the detected data to the control device 11. Hereinafter, it is assumed that the flow rate sensor 12 is provided for each electromagnetic valve 3 and the flow rate sensor 13 is provided for each electromagnetic valve 4.

[0017] The control device 11 outputs an open command to the selected electromagnetic valves 3 and 4 from among a plurality of electromagnetic valves 3 and 4 so as to cool the steel plate 1 at a flow rate set according to the conveyance speed of the steel plate 1 conveyed by the run-out table 2, the width, thickness, and material of the steel plate 1, etc., and cools the steel plate 1 with the cooling water 30. In FIG. 1, the open command is indicated by a solid arrow and the close command is indicated by a dashed arrow. In the specific example of FIG. 1, the control device 11 selects and sets to open the electromagnetic valves 3 and 4 every other one both vertically and horizontally.

[0018] Subsequently, the configuration of the control system 100 will be described. The cameras 14 and 15 are communicably connected to the image processing device 16. The image processing device 16 is communicably connected to the data collection device 17. The data collection device 17 is also communicably connected to the control device 11 and is electrically connected to the outputs of the electromagnetic valves 3 and 4 and the flow rate sensors 12 and 13. The data collection device 17 is communicably connected to the machine learning device 18. The machine learning device 18 is communicably connected to the control device 11.

[0019] The camera 14 is arranged to photograph the state of the cooling water ejected from each of the plurality of spray nozzles 5 and 6 arranged above the steel plate 1, i.e., from which spray nozzle the cooling water is ejected and how, or whether it is ejected or not. The camera 14 photographs the ejection state of each of the plurality of spray nozzles 5 and 6, generates image data, and outputs it to the image processing device 16. The camera 14 may photograph each spray nozzle 5 and 6 to generate image data, or photograph several spray nozzles 5 and 6 in groups in multiple times, or photograph all the spray nozzles 5 and 6 at once to generate image data for each photograph. Alternatively, a plurality of cameras 14 may be prepared according to the number of spray nozzles 5 and 6.

[0020] The camera 15 is arranged to photograph the state of the cooling water ejected from each of the plurality of spray nozzles 7 arranged below the steel plate 1, including from which spray nozzle the cooling water is ejected, how it is ejected, or whether it is ejected or not. The camera 15 photographs the ejection state of each of the plurality of spray nozzles 7, generates image data, and outputs it to the image processing device 16. The camera 15 may photograph each spray nozzle 7 to generate image data, or photograph several spray nozzles 7 in groups multiple times, or photograph all the spray nozzles 7 at once to generate image data for each photographing. Alternatively, according to the number of spray nozzles 7, a plurality of cameras 15 may be prepared.

[0021] The image processing device 16 performs image processing on the image data output from the cameras 14 and 15 respectively, and determines whether the cooling water is ejected from the spray nozzles 5 to 7, and the ejection state of the ejected cooling water. The image processing device 16 determines the ejection state of the cooling water of the spray nozzles 5 to 7 arranged above and below the steel plate 1 for each spray nozzle. Since the determination for the spray nozzles above and below the steel plate 1 is performed independently and in the same manner, in order to avoid complicating the explanation, hereinafter, the case of the ejection of the spray nozzles 7 arranged below the steel plate 1 generated by the camera 15 will be described.

[0022] The data collection device 17 obtains, as time-series data, the result of the image processing device 16 determining the ejection state of the cooling water for each spray nozzle 7 based on the image data. The data collection device 17 obtains the opening command and closing command (hereinafter, the opening command and closing command are collectively referred to as the opening / closing command) output by the control device 11 as time-series data. The data collection device 17 obtains the actual data of the drive current flowing through the electromagnetic valve 4 as time-series data, and obtains the actual data of the flow rate of the cooling water output by the flow rate sensor 13 as time-series data.

[0023] The data collection device 17 synchronizes a plurality of determination results of the ejection state of the cooling water, a plurality of opening / closing commands, a plurality of valve current actual data, and a plurality of flow rate actual data in terms of time, and outputs them to the machine learning device 18.

[0024] The machine learning device 18 updates a failure determination model by machine learning using the plurality of time-series data synchronized in terms of time acquired from the data collection device 17. Based on the updated failure determination model, the machine learning device 18 classifies the content of a failure or abnormality (hereinafter, a failure or abnormality is simply referred to as a failure, etc.) of the cooling system, and stores it as a database.

[0025] In updating the failure determination model, the machine learning device 18 may classify the state at a stage prior to a failure, etc. of the cooling system and store it as a database. The stage prior to a failure, etc. is, for example, a state indicating a tendency to reach a failure, etc., and more specifically, for example, represents a state of deterioration of components of the cooling system.

[0026] The operation of the control system 100 according to the embodiment will be described. FIG. 2 is a schematic block diagram for explaining the operation of the control system according to the embodiment. FIG. 2 shows an example in which the machine learning device 18 updates the failure determination model 19 by machine learning the time-series data synchronized in terms of time, and classifies the content of a failure, etc. of the cooling system according to the failure determination model 19.

[0027] As shown in FIG. 2, the machine learning device 18 classifies the content of a failure, etc. according to the updated failure determination model 19, and updates the maintenance information list 23 and the failure information list 24 as a database.

[0028] The maintenance information list 23 is classification information on the content of the state at the pre-failure stage, etc. For example, when it is determined that the state is deterioration as the pre-failure stage, etc., it is a database associated with the image data representing the water spray state of the cooling water, the opening / closing command output from the control device 11, the valve current actual data, and the time when the flow rate actual data was acquired. The classification categories are set in advance and are, for example, categories such as "valve mechanical part deterioration", "valve electrical part deterioration", "pipe leakage", "pipe blockage", etc.

[0029] For example, "valve mechanical part deterioration" is a case where an open command is output from the control device 11 to the target electromagnetic valve 4, and the water spray state of the cooling water is between the failure level and the non-defective level even though the drive current of the electromagnetic valve 4 is within the normal range. For example, "valve electrical part deterioration" is a case where, even though an open command is output from the control device 11 to the target electromagnetic valve 4, the drive current of the electromagnetic valve 4 is outside the normal range, and the water spray state of the cooling water is between the failure level and the non-defective level.

[0030] The failure information list 24 is classification information on the content of failures, etc. For example, when it is determined that there is a failure, etc., it is a database associated with the image data representing the water spray state of the cooling water, the opening / closing command output from the control device 11, the valve current actual data, and the time when the flow rate actual data was acquired. The classification categories are set in advance and are, for example, categories such as "valve failure", "electrical circuit failure", etc.

[0031] For example, "valve failure" is a case where an open command is output from the control device 11 to the target electromagnetic valve 4, and the water spray state of the cooling water is inappropriate even though the drive current of the electromagnetic valve 4 is within the normal range. For example, "electrical circuit failure" is a case where an open command is not output from the control device 11 to the target electromagnetic valve 4 and the drive current of the electromagnetic valve 4 is outside the normal range.

[0032] For example, in the machine learning device 18, a clustering method can be used to set and update the defect determination model 19. The cluster data is formed by a plurality of time-series data synchronized in time, and the level and direction of variation that vary with the passage of time are evaluated to perform classification determination of the content in the pre-failure stage such as a failure and classification determination of the content of a failure or the like. The level of variation and the direction of variation are set in advance in the defect determination model 19.

[0033] The direction of variation of the cluster data is the direction in a four-dimensional space of the determination result by image data, the opening / closing command, the valve current actual data, and the flow rate actual data. This four-dimensional space is set for each electromagnetic valve 4. The level of variation of the cluster data is the magnitude of the variation of the data in a specific direction set in this four-dimensional space.

[0034] The machine learning device 18 determines that there is a failure or the like, for example, when the level of variation according to the time change of the cluster data is a predetermined value, and determines that a level lower than the variation level determined to be a failure or the like is a tendency of a failure or the like. The variation level is not limited to a fixed value, and may be the magnitude of variation or the variation rate within a predetermined period, or the magnitude of variation or the variation rate with respect to the initial value.

[0035] Note that the cluster data is not limited to the case where it is set in a four-dimensional data space of the determination result of the ejection state of the cooling water, the opening / closing command, the valve current actual data, and the flow rate actual data. For example, there may be cases where the ejection state of the cooling water is determined for a plurality of types of states. The third spray nozzle 7 from the left in FIG. 2 has an insufficient ejection amount of the cooling water, which is, for example, a state of failure of the electromagnetic valve 21 or the spray nozzle. Even if the ejection amount of the cooling water is sufficient, if the shape of the ejected cooling water is different from the shape of the cooling water ejected from other spray nozzles, there is a possibility of a precursor to the failure of the spray nozzle. In such a case, the cluster data becomes a more multi-dimensional data space according to the type of determination.

[0036] In the maintenance information list 23 and the failure information list 24, for the pre-set content at the stage prior to a failure or the like and the classification categories of the content of the failure or the like, do not give names having attributes such as "valve failure", set initial classifications such as "1", "2",... After updating the defect determination model 19, the classified data may be associated with the time-series data before classification so that a human assigns an attribute name such as "valve failure" to the initial classification. In advance, if the failure mode and the degradation mode are known, an attribute name may be assigned to the initial classification.

[0037] In the initial state where the control system 100 has no operation record of the cooling system, the defect determination model 19 is in the initial state. Therefore, sufficient time-series data is collected and an effective defect determination model 19 is formed.

[0038] In the initial state of the control system 100, the generated defect determination model 19 may not be based on sufficient data. By operating the cooling system, the machine learning device 18 collects sufficient time-series data and generates a defect determination model 19. Whether sufficient time-series data has been collected may be determined by setting a collection period in advance, or a human may view and judge the content of the maintenance information list 23 and the failure information list 24 generated by the defect determination model 19. At that time, an attribute name representing a failure state such as "electromagnetic valve failure" may be given to the initial classification category.

[0039] The machine learning device 18 sequentially updates the maintenance information list 23 and the failure information list 24 using the defect determination model 19.

[0040] For example, when the failure information list is updated during the test injection of the electromagnetic valve 4, the machine learning device 18 closes the electromagnetic valve 4 of the failed spray nozzle 7 and commands the control device 11 to issue an alarm and recalculate the cooling amount to the cooling model preset in the control device 11.

[0041] For example, when the failure information list is updated while the steel plate 1 is being rolled, the machine learning device 18 instructs the control device 11 to close the solenoid valve of the target spray nozzle 7 and issue an alarm. Further, the machine learning device 18 instructs the control device 11 to automatically switch the control to the nozzle with the next highest priority among the failed nozzles so as to compensate for the injection amount of the insufficient cooling water 30 by the failed spray nozzle 7.

[0042] In this way, the control system 100 according to the embodiment can operate.

[0043] The effects of the control system 100 according to the embodiment will be described. The control system 100 according to the embodiment includes cameras 14 and 15 and an image processing device 16. The image processing device 16 can analyze, by image processing, image data including the state of the cooling water ejected from each of the spray nozzles 5 to 7, and determine how the cooling water is ejected or not ejected.

[0044] The control system 100 includes a data collection device 17 and a control device 11. The data collection device 17 can collect the determination result of the state of the cooling water ejection, the opening / closing commands of the solenoid valves 3 and 4 that control the cooling water ejection, the actual opening / closing states, and the performance data of the flow rate of the cooling water controlled by the solenoid valves 3 and 4. The data collection device 17 can output a plurality of time-series data by synchronizing the collected performance data and the like for each spray nozzle with time.

[0045] The control system 100 includes a machine learning device 18 having a defect determination model 19, and can classify a plurality of collected time-series data from the data collection device 17 by machine learning for each content of a failure or the like and the content of the stage before the failure or the like, and store them in a database. Therefore, in the database, the determination results of the state of the cooling water jet, the states of the opening / closing commands of the electromagnetic valves 3 and 4 that control the cooling water jet, the actual opening / closing states, and the performance data of the flow rate of the cooling water controlled by the electromagnetic valves 3 and 4, which are classified for each content of the stage before the failure or the like, are classified and stored for each mode of the failure or the like. By referring to this database (maintenance information list 23), it is possible to determine before a problem occurs whether there is a problem with elements such as parts related to which spray nozzle, and take appropriate measures. As a result, it is possible to make it difficult for problems such as the stop of the cooling water jet to occur due to an unexpected accident or the like, and improve the operating rate of the facility.

[0046] In the database of the control system, the determination results of the state of the cooling water jet, the states of the opening / closing commands of the electromagnetic valves 3 and 4 that control the cooling water jet, the actual opening / closing states, and the performance data of the flow rate of the cooling water controlled by the electromagnetic valves 3 and 4, which are classified for each content of a failure or the like, are classified and stored for each mode of the failure or the like. Therefore, it becomes easy to investigate the cause of a failure or the like, and even when the operation of the facility is started or stopped, the facility can be quickly restored.

[0047] Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the gist of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are included in the invention described in the claims and its equivalent scope.

Explanation of Reference Numerals

[0048] 1...Steel plate, 2...Run-out table, 3, 4...Electromagnetic valves, 5, 6, 7...Spray nozzles, 8...Control device, 11...Control device, 12, 13...Flow sensors, 14, 15...Cameras, 16...Image processing device, 17...Data collection device, 18...Machine learning device, 19...Defect determination model, 23...Maintenance information list, 24...Fault information list, 30...Cooling water, 100...Control system

Claims

1. A control device that outputs a plurality of opening / closing commands for controlling a plurality of electromagnetic valves that respectively control the ejection of cooling water from a plurality of spray nozzles arranged along a run-out table on which a steel plate is conveyed; A camera that sequentially captures the state of the ejection of cooling water from the plurality of spray nozzles to generate image data; An image processing device that performs image processing on the image data, determines the ejection state of the cooling water of each of the plurality of spray nozzles, and outputs one or more determination results; A data collection device that respectively detects the plurality of opening / closing commands, the flow rates of the cooling water supplied to the plurality of spray nozzles, the plurality of flow rate achievements output by the plurality of flow sensors, the plurality of drive current achievements flowing through the plurality of electromagnetic valves, and collects a plurality of time-series data in which the determination results are time-synchronized; A machine learning device that generates a defect determination model that classifies the failure or abnormality of the one or more determination results based on the plurality of time-series data to generate a database; Comprising; The machine learning device is a control system that classifies the failure or abnormality of the one or more determination results based on the fluctuations over time of the cluster data including each of the plurality of opening / closing commands, each of the plurality of flow rate achievements, each of the plurality of drive current achievements, and the determination results.

2. The control system according to claim 1, wherein the database includes the classification of the failure or abnormality and the classification of the state at the stage preceding the failure or abnormality.

3. The control system according to claim 2, wherein the machine learning device issues a failure notification when a failure or abnormality of the one or more determination results is detected in one of the plurality of spray nozzles.

Images