Cell control system

Abstract

Cell control system (1), comprising: at least one manufacturing cell (4) comprising at least one manufacturing machine (41); and a cell controller (3) for transmitting an operating instruction to the production machine (41) based on a production plan received from a production planning device (2), wherein the cell control (3) comprises a machine operating instruction unit (31) for transmitting the operating instruction to the production machine (41) based on the production plan, a fault value collecting unit (33) for collecting detected fault information, an operating information collecting unit (34) for collecting operating information of the production machine (41), a learning unit (35) for creating a learning model by performing machine learning using the operation information collected by the operation information collecting unit (34) as an input signal and the disturbance information collected by the disturbance value collecting unit (33) as an instruction signal, an estimation unit (36) for analyzing the learning model created by the learning unit (35) to estimate operational information corresponding to a fault cause detected by the manufacturing machine (41), wherein the learning model is analyzed to calculate a respective correlation coefficient of a respective operational information item and a respective disturbance value for randomly extracted operational information items, and wherein an average value of the correlation coefficients is calculated, and wherein the disturbance value is considered as a function of the operational information, and an operation instruction changing unit (37) for instructing the machine operation instruction unit (31) to change instruction content based on the operation information corresponding to a disturbance factor estimated by the estimation unit (36).

Description

GENERAL STATE OF THE ART1. Field of the InventionThe present invention relates to a cell control system, and more particularly to a cell control system that estimates an influence of disturbances in multiple machines.2. Description of the Prior ArtFIG. 7 is a diagram illustrating a cell control system (manufacturing management system) that manages a manufacturing cell including a plurality of manufacturing machines. The cell control system improves an operation rate of the entire manufacturing cell by instructing each manufacturing cell to operate based on a manufacturing plan indicated by a manufacturing planning device during management of use states of the manufacturing machines or determination of a manufacturing machine to be used.When an alarm occurs in a manufacturing machine included in the operated manufacturing cell at the time of operating a manufacturing cell according to the manufacturing plan so that an operation is interrupted, productivity in the manufacturing cell decreases. As a result, an operation rate of the entire manufacturing cell decreases.As a conventional technology for responding to the occurrence of the alarm in the manufacturing machine, for example, Japanese Patent Laid-Open Publication No. 2011-243,118 discloses a monitoring diagnostic apparatus that uses a plurality of parts in a monitoring target apparatus as an object to be monitored, collects physical quantities of an attached sensor over time, detects a fault, diagnoses a cause thereof, and determines the presence / absence of a causal relationship from a correlation coefficient between sensor data elements. In addition, Japanese Patent Laid-Open No. 08-320,726 A discloses a diagnostic / analysis apparatus which interprets a correlated pair of signals picked up from a plurality of objects to be controlled, analyzes the presence / absence of a fault by comparison with correlation information at the normal time, and outputs a conclusion of fault location of presence / absence information on the fault.Generally, measures against troubles (electrical trouble or physical vibration) are taken in the manufacturing machine to prevent the alarm from occurring. These measures are carried out in the following procedure.Procedure a1) A failure state of each machine is measured.Procedure a 2) A probability that an alarm occurs is analyzed from a measurement result in Procedure a 1.Procedure a3) A cause of the alarm is specified from a result of the analysis, and measures against the disturbance are taken to prevent the cause.However, in an environment such as a factory where a plurality of manufacturing machines are continuously operating, an influence of disturbance changes in connection with a combination of a plurality of states such as arrangement or wiring of the manufacturing machines, an operation pattern, etc. Therefore, it is difficult to specify the cause of the alarm in many cases even when the above procedures are executed. In such a state, both the fault and an operating state must be analyzed simultaneously. However, in practice, since the plurality of manufacturing machines operate simultaneously, the operation pattern as a whole is complex and the analysis of the cause is difficult.On the other hand, in the technology disclosed in Japanese Patent Laid-Open No. 2011-243, 118, it is not effective to apply the technology to a manufacturing cell in which a plurality of machines operate simultaneously because diagnosis is performed using only one sensor in the device. In addition, in the technology disclosed in Japanese Patent Laid-Open No. H08-320,726A, although a fault can be detected in a machine system including a plurality of control devices, a cause of the fault can be estimated only in a state in which the fault has been recorded in a database in advance.The document "Sum: Predictive maintenance, its implementation and latest trends. Proceedings of the Institution of Mechanical Engineers" by SELCUK Sum in Part B: of the Journal of Engineering Manufacture, 2016, relates to a predictive maintenance concept (PdM) in the industrial environment. EP 1 390 822 B1 relates to a field device having a diagnostic circuit that measures a property associated with a process control and measurement system (10) to provide a diagnostic output indicative of a state of the process control and measurement system. The document "A neural network integrated decision support system for condition-based optimal predictive maintenance policy" by WU, Sze-jung [et al.] in IEEE Transactions on Systems, Man, and Cybernetics-Part A: Systems and Humans, 2007, Vol. 37, No. 2, pp. 226-236 relates to predictive maintenance with artificial neural networks.The article "Supervised Learning" as retrievable on Wikipedia on 26.01.2015 under "https: / / de.wikipedia.org / w / index.php?title=%C3%9Cberwachtes_Lernen&oldid=1381806 11" describes the term supervised learning as a sub-region of machine learning and refers to training and validating artificial neural networks.SUMMARY OF THE INVENTIONIn this regard, an object of the invention is to provide a cell control system capable of estimating a cause of an alarm by estimating an influence of disturbance in a plurality of machines. This object is achieved by a cell control system according to claim 1.In the invention, a control device of a machine included in a manufacturing cell is connected to a cell controller. Each cell controller monitors and collects the following conditions for each manufacturing machine. 1. a disturbance value at each measurement position (a main body of the control device, an amplifier, a power supply, a signal line, etc.) 2. operation information of each machine of the manufacturing cell (a speed, acceleration, and load of each axis, a serial number of the block)Then, the cell controller analyzes the collected information in the following procedure.Procedure b 1) A correlation between the disturbance value and the operation information is calculated by machine learning.Procedure b2) A state in which noise is likely to occur (=combination of the operation information) is estimated from the correlation.Procedure b3) A machine or a part thereof in which an interference level is likely to become high is detected from the correlation.The cell controller notifies an estimated result to an operator of the manufacturing machine or a high-order server.The operator can take measures against troubles by concentrating on a specific part of the manufacturing machine based on the estimated result. Therefore, it is possible to take measures to efficiently improve an operation rate of the manufacturing cell in a small number of processes.Further, a cell control system according to the invention includes at least one manufacturing cell including at least one manufacturing machine and a cell controller for transmitting an operation instruction to the manufacturing machine based on a manufacturing plan received from a manufacturing planning device, wherein the cell controller includes a machine operation instruction unit for transmitting the operation instruction to the manufacturing machine based on the manufacturing plan, a disturbance value collection unit for collecting detected disturbance information, an operation information collection unit for collecting operation information of the manufacturing machine, a learning unit for constructing a learning model by executing machine learning using the operation information collected by the operation information collection unit as an input signal, and the disturbance information collected by the disturbance value collection unit as an instruction signal, an estimation unit, to analyze the learning model created by the learning unit and estimate operation information corresponding to a cause of disturbance detected by the manufacturing machine, and an operation instruction changing unit to instruct the machine operation instruction unit to change instruction content based on the operation information corresponding to a disturbance estimated by the estimating unit.In the control system according to the invention, the estimating unit performs prediction using the learning model created by the learning unit to estimate a manufacturing machine having low fault resistance in the manufacturing machine, and the operation instruction changing unit instructs the machine operation instruction unit to change a communication content based on information on fault resistance estimated by the estimating unit.In the cell control system according to the invention, the operation instruction changing unit instructs the machine operation instruction unit to change an operation instruction estimated to affect the manufacturing machine with the low fault resistance estimated by the estimating unit.According to the invention, it is possible to estimate operation information having a large correlation with noise with respect to each manufacturing machine. It is possible to prevent an alarm from occurring by taking measures against troubles based on an estimated result. In this way, it is possible to improve an operation rate of a manufacturing machine. In addition, it is possible to detect correlation of breakdowns between manufacturing machines and detect deterioration of breakdown resistance of a manufacturing machine by continuous monitoring.BRIEF DESCRIPTION OF THE DRAWINGSThe above-described object and characteristic of the invention and other objects and characteristics will be understood from the description of embodiments with reference to the accompanying drawings. In the drawings: FIG. 1 is a schematic block diagram of a cell control system according to an embodiment of the invention; FIG. 2 is a diagram illustrating an example of a learning unit using a multilayer neural network; FIG. 3 is a diagram for describing a procedure for estimating operation information corresponding to a cause of failure by an estimating unit; FIG. 4 is a diagram for describing a procedure for estimating operation information corresponding to a cause of failure by the estimating unit; FIG. 5 is a diagram for describing a procedure for estimating operation information corresponding to a cause of failure by the estimating unit; FIG. 6 is a diagram for describing a procedure for estimating a manufacturing machine having low fault resistance by the estimating unit; and FIG. 7 is a diagram illustrating a cell control system (manufacturing management system) that manages a manufacturing cell including a plurality of manufacturing machines.DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTSHereinafter, an embodiment of the invention will be described with reference to the drawings.FIG. 1 is a schematic block diagram of a cell control system according to an embodiment of the invention. In FIG. 1, a dotted arrow indicates an information flow in a conventional technology, and a continuous arrow indicates an information flow introduced in the invention. A cell control system 1 of the present embodiment is configured by connecting a manufacturing planning device 2, a cell controller 3, and at least one manufacturing cell 4 through a network, etc.The manufacturing planning device 2 plans all the manufacturing work performed in at least one manufacturing cell 4, and sends the planned manufacturing work as a manufacturing plan to the cell controller 3 that manages each manufacturing cell 4.The cell controller 3 instructs each manufacturing cell to operate while managing a use state of a manufacturing machine 41 included in a managed manufacturing cell 4 or determines a manufacturing machine 41 to be used based on the manufacturing plan indicated by the manufacturing planning device 2. The cell controller 3 illustrated in FIG. 1 manages the at least one manufacturing cell 4. the cell controller 3 includes a manufacturing plan receiver 30, a machine operation instruction unit 31, a communication unit 32, a disturbance value collection unit 33, an operation information collection unit 34, a learning unit 35, an estimation unit 36, and an operation instruction change unit 37.The manufacturing plan receiver 30 receives the manufacturing plan displayed by the manufacturing planning device 2, and outputs the received manufacturing plan to the machine operation instruction unit 31.The machine operation instruction unit 31 transmits an operation instruction to the at least one manufacturing machine 41 included in the manufacturing cell 4 managed by the cell controller 3 through the communication unit 32 based on the manufacturing plan input from the manufacturing plan receiver 30. The machine operation instruction unit 31 has a function of generating an operation time plan based on the manufacturing plan for each manufacturing machine 41 included in the managed manufacturing cell 4.The disturbance value collection unit 33 collects a value associated with disturbances detected by the manufacturing machine 41 included in the managed manufacturing cell 4 from the manufacturing machine 41 and stores the collected value together with time information, etc. in an operation information database 38. With reference to the disturbance value collected by the disturbance value collection unit 33, in addition to a disturbance value detected by a disturbance detector 42 included in the manufacturing machine 41, it is possible to collect a disturbance value detected by a sensor, etc. (not illustrated) provided in a factory or outside, or it is possible to collect all disturbance values that can be obtained by the cell control system 1.The operation information collection unit 34 collects operation information indicating an operation state of the manufacturing machine 41 included in the managed manufacturing cell 4 from the manufacturing machine 41, and stores the collected operation information together with time information, etc. in the operation information database 38 for each manufacturing machine 41. in addition to the information indicating the operation state of the manufacturing machine 41, the operation information collection unit 34 may collect, as operation information, for example, all kinds of information such as time, a power state of the plant, etc. that can be acquired by the cell control system 1.The learning unit 35 performs machine learning on a relationship between the operation information collected from each manufacturing machine and the disturbance value detected by each manufacturing machine 41 based on the disturbance value and the operation information stored in the operation information database 38. Further, a model learned by the learning unit 35 is used for estimation by the estimation unit 36 described below. Each model may be used as a model used for learning by the learning unit 35 when the model can estimate a trend in the change of a disturbance value due to a change in certain operation information.For example, it is possible to use a multilayer neural network, a Bayesian network, etc. described below.When a multilayer neural network illustrated in FIG. 2 is used as the learning unit 35, for example, operation information of each manufacturing machine 41 may be given as an input signal and a disturbance value of each manufacturing machine 41 may be given as an instruction signal. In an intermediate layer 1, operation information of the same manufacturing machine may be learned together in operation information of each manufacturing machine 41. In an overall composite layer, a correlation between operation information of each manufacturing machine 41 and disturbances of each manufacturing machine 41 can be learned.In the estimating unit 36, a disturbance factor that largely affects disturbances is estimated from the operation information using the model learned by the learning unit 35. For example, the operation information corresponding to the disturbance may be estimated by the following procedure.Procedure c1) Addressed Disturbance information is set to Disturbance A.Procedure c 2) Operation information of plural times is randomly selected from the collected operation information.Procedure c 3) With respect to each selected operation information element, a slope of an input value with respect to a disturbance value A is calculated using the learned model.Procedure c4) An average value of slopes is calculated, and an input signal having a particularly large slope is estimated as operation information corresponding to a noise factor. In a case where a difference in slope is small compared to another input signal, the input signal is not estimated as a noise factor.FIGS. 3A-1 and 3A-2 and FIGS. 4B-1 and 4B-2 are diagrams illustrating an image of a procedure for estimating the above-described noise factor. In FIGS. 3A-1 and 3A-2 and FIGS. 4B-1 and 4B-2, for example, an interference value corresponding to an estimation object of a factor is set to A, w1 denotes an operation information feed rate value, w2 denotes an operation information spindle speed value, X1 denotes a correlation coefficient of a feed rate and an interference value A(w), X2 denotes a correlation coefficient of a spindle speed and an interference value A(w), and the prime sign (n) denotes a consecutive number n assigned to a plurality of operation information items selected in the above-described procedure c2. The disturbance value A modeled by machine learning is considered as a function of the operation information w and is therefore denoted by A(w) in the figure.As illustrated in FIGS. 3A-1, the model learned by the learning unit 35 is analyzed to calculate a correlation coefficient X 1( 1) of a feed rate and a disturbance value A(w) for specific operation information w( 1). Similarly, as illustrated in FIG. 3A-2, the model learned by the learning unit 35 is analyzed to calculate a correlation coefficient X 1( 2) of a feed rate and a disturbance value A(w) for specific operation information w( 2)). In this way, as illustrated in FIG. 4B- 1, with respect to the feed speed w 1 in each of the randomly extracted operation information items w( 1) to w(n), each of the correlation coefficients X 1( 1) to X 1( n) is obtained by analyzing the model learned by the learning unit 35, and a value obtained by averaging these correlation coefficients is calculated. When an average value (0.5 in FIG. 4B- 1 ) of the correlation coefficients calculated in this manner is larger than a predetermined threshold value (for example, a threshold value 0.1) determined in advance, a change in the operation information (the feed rate w 1 in FIG. 4B- 1 ) largely affects the disturbance value A(w), and therefore the feed rate w 1 can be estimated as a disturbance factor of the disturbance value A. In an example illustrated in FIG. 4B-2, since an average value (0.025) of the respective correlation coefficients X 1( 1) to X 1( n) of the spindle rotation speed w 2 is a small value, a change in the spindle rotation speed w 2 does not largely affect the disturbance value (w), and therefore it can be estimated that the spindle rotation speed w 2 is not a disturbance factor of the disturbance value A.In FIGS. 3A-1 and 3A-2 and FIGS. 4B-1 and 4B-2 described above, for convenience of description, each operation information element is illustrated as a two-dimensional (2D) graph with respect to a disturbance value, and then a slope value is obtained as a correlation coefficient. However, in practice, as illustrated in FIG. 5, a slope value for each direction at each point in a multi-dimensional function corresponding to the number of inputs of operation information is calculated as a correlation coefficient.In the estimating unit 36, a machine having low disturbance resistance may be estimated using the model learned by the learning unit 35. The machine having the low disturbance resistance can be estimated, for example, in the following procedure.Procedure d 1) A part of the data corresponding to an appropriate time is selected from collected operation information.Procedure d 2) The selected operation information is input as an input value to the model learned by the learning unit 35, and a slope of an input value at which each output value increases is calculated.Procedure d 3) Each input value is increased and decreased in a direction of the calculated slope.Procedure d 4) Procedures d 2 to d 3 are repeated until one of the outputs exceeds or converges a predetermined threshold.Procedure d 5) A manufacturing machine corresponding to an output signal exceeding the threshold in Procedure d 4 is estimated to have low fault resistance.FIGS. 6C-1 and 6C-2 are diagrams illustrating an image of a procedure for estimating the low fault resistance machine described above. In a case where the disturbance value A(w) exceeds the predetermined threshold value, when the input value is increased in the procedure d 3 as illustrated in FIG. 6C- 1, it is possible to estimate that disturbance may occur in the manufacturing machine during operation, and the manufacturing machine may be estimated to have low disturbance resistance. In addition, in a case where the disturbance A(w) does not exceed the predetermined threshold and converges (the disturbance value (w) falls off), when the input value is increased in the procedure d 3 as illustrated in FIG. 6C-2, it is possible to estimate that a disturbance value A corresponding to a level that causes a problem during operation is not generated in the manufacturing machine and estimate that there is no problem of disturbance resistance of the manufacturing machine.The operation instruction changing unit 37 instructs the machine operation instruction unit 31 to change instruction content sent to each manufacturing machine 41 based on a result of estimating operation information corresponding to a disturbance factor by the estimating unit 36 or a result of estimating a manufacturing machine having low disturbance resistance by the estimating unit. As an example of changing an instruction sent to the machine operation instruction unit 31 by the operation instruction changing unit 37 at the time of operating a manufacturing machine, for example, corresponding to a cause of operation information actually estimated as a disturbance or a manufacturing machine estimated to have low disturbance resistance, a message prompting a user to take measures against the disturbance associated with operation information estimated as a disturbance is displayed on a screen, etc., of an operation panel of the manufacturing machine, or alerts the user to change the operation information estimated as a disturbance to a value having a level at which disturbance does not occur (recommends, for example, lowering a feed speed of the manufacturing machine). Further, the operation instruction changing unit 37 may report information associated with the operation information estimated as having the disturbance factor or the manufacturing machine estimated as having the low disturbance resistance to a higher server as the manufacturing planning device 2.The operation instruction changing unit 37 may instruct the machine operation instruction unit 31 to change operation instruction content sent to each manufacturing machine 41 so that a failure of the machine due to disturbance does not occur, based on a result of estimating operation information corresponding to a disturbance factor by the estimating unit 36 or a result of estimating a manufacturing machine having low disturbance resistance by the estimating unit. As an example of changing the operation instruction content of the manufacturing machine, a program used is changed such that an operation is executed using a method that does not cause such an operation state with respect to a manufacturing machine corresponding to a cause of the operation information estimated as the disturbance factor, or an order is changed such that, while the manufacturing machine estimated to have low disturbance resistance is being operated, a manufacturing machine corresponding to a disturbance factor with respect to the manufacturing machine is not instructed to operate. In addition, it is possible to consider a change such that the operation information estimated as the disturbance is reduced to a value corresponding to a level at which disturbance does not occur (for example, the operation instruction is automatically changed to lower a feed speed of the manufacturing machine).The manufacturing cell 4 includes the at least one manufacturing machine 41. in addition, each manufacturing machine 41 includes the interference detector 42 and an operation information transmitter 43.The interference detector 42 detects interference occurring using a sensor (not illustrated), etc. installed in each part of the manufacturing machine 41, and transmits a value of the detected interference to the cell controller 3. In addition, the operation information transmitter 43 transmits information collected by a controller of the manufacturing machine 41 (not illustrated) from each part of the manufacturing machine 41 to the cell controller 3 as operation information.Although the embodiment of the invention has been described above, the invention is not limited to the example of the above-described embodiment and can be implemented in various ways by making appropriate modifications.

Claims

A cell control system (1) comprising: at least one manufacturing cell (4) including at least one manufacturing machine (41); and a cell controller (3) for transmitting an operation instruction to the manufacturing machine (41) based on a manufacturing plan received from a manufacturing planning device (2), wherein the cell controller (3) includes a machine operation instruction unit (31) for transmitting the operation instruction to the manufacturing machine (41) based on the manufacturing plan, a disturbance value collection unit (33) for collecting detected disturbance information, an operation information collection unit (34) for collecting operation information of the manufacturing machine (41), a learning unit (35) for creating a learning model by executing machine learning using the operation information collected by the operation information collection unit (34) as an input signal and the disturbance information collected by the disturbance value collection unit (33), as an instruction signal, an estimating unit (36) for analyzing the learning model created by the learning unit (35) to estimate operation information corresponding to a cause of disturbance detected by the manufacturing machine (41), the learning model being analyzed to calculate a respective correlation coefficient of a respective operation information element and a respective disturbance value for randomly extracted operation information elements, and an average value of the correlation coefficients being calculated, and the disturbance value being regarded as a function of the operation information, and an operation instruction changing unit (37) for instructing the machine operation instruction unit (31) to change instruction content based on the operation information corresponding to a disturbance factor estimated by the estimating unit (36).The cell control system (1) according to claim 1, wherein the estimating unit (36) performs prediction using the learning model created by the learning unit (35) to estimate a manufacturing machine (41) having low fault resistance in the manufacturing machine (41), and the operation instruction changing unit (37) instructs the machine operation instruction unit (31) to change communication content based on information on fault resistance estimated by the estimating unit (36).The cell control system (1) according to claim 2, wherein the operation instruction changing unit (37) instructs the machine operation instruction unit (31) to change an operation instruction estimated to affect the manufacturing machine (41) with the low fault resistance estimated by the estimating unit (36).The cell control system (1) according to claim 1, wherein operation information is estimated as a disturbance factor when the calculated average value of the correlation coefficients is larger than a predetermined threshold value.The cell control system (1) according to claim 1, wherein an operation information item is a feed speed or a spindle rotational speed.

Images