JP2020074177A - Machine learning device for learning display of operation menu, numerical control device, machine tool system, manufacturing system, and machine learning method - Google Patents

Abstract

To provide a machine learning device capable of displaying an optimum operation menu to each operator, a numerical control device, a machine tool system, a manufacturing system, and a machine learning method.SOLUTION: A machine learning device 2 for detecting an operator, communicating with a database in which the information of the operator is registered, and learning the display of the operation menu based on the information of the operator includes a state observation part 21 for observing the operation history of the operation menu and a learning part 22 for learning the display of the operation menu on the basis of the operation history of the operation menu observed by the state observation part.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a machine learning device, a numerical control device, a machine tool system, a manufacturing system and a machine learning method for learning the display of operation menus.

  Conventionally, for example, in a numerical control device (NC (Numerical Control) device) for controlling a machine tool, various people (operators) use its operation menu. Here, the operators include, for example, persons of various positions and authority levels such as a machine tool maker (MTB (Machine Tool Builder) developer and person in charge, an operator (user) and a service person (service engineer)). ..

  In this specification, the NC device also includes a computer numerical control device (CNC device) and the like. The operation menu of the NC device is displayed on a display device such as a liquid crystal panel having a touch (position input) function, and changes according to the content displayed on the display device. The display device for displaying the operation menu can be provided integrally with the NC device, for example, but it is provided at a position separated from the NC device by wire or wirelessly, or the display device is operated near the machine tool or the like. You can also operate the menu.

  By the way, conventionally, for example, there has been proposed a method of arranging function icons as cells in a matrix and rearranging the function icons in an arrangement desired by the user according to the number of times the icons are used (see, for example, Patent Document 1). Further, there is also proposed a method of rearranging menu items in an order suitable for the current user situation when displaying a menu of a navigation device (for example, see Patent Document 2) without bothering the user. Further, for example, it is possible to set an authority level for an operator who operates the machine tool and change the operation menu according to the authority level (see, for example, Patent Document 3).

JP, 2009-181501, A JP, 2010-127814, A JP, 2009-086964, A

  As described above, for example, arranging the menus according to the number of times the menus are used may result in poor access to the menus that the user wants to use, for example, when a service person other than the operator who normally operates the machine operates the menus. Conceivable. In addition, after detecting the operator, for example, a method of rearranging the menus according to the number of times of use for each operator may be considered. In this case, in order to obtain a menu screen that is easy for the user to use, It is necessary to operate by learning to some extent.

  In addition, by preparing in advance a table that determines the menu display item order from parameters, when displaying a menu suitable for the situation, it is not possible to dynamically respond to changes from the assumed situation. , It is necessary to manually recreate the table every time the operating person's position or authority level increases.

  An object of the present invention is to provide a machine learning device, a numerical control device, a machine tool system, a manufacturing system and a machine learning method capable of displaying an optimum operation menu for each operator in view of the above-mentioned problems of the prior art. It is in.

  According to the first embodiment of the present invention, machine learning that detects an operator, communicates with a database in which information of the operator is registered, and learns to display an operation menu based on the information of the operator. A state observing unit that observes the operation history of the operation menu, and a learning unit that learns the display of the operation menu based on the operation history of the operation menu observed by the state observing unit. A machine learning device is provided.

  The operation history of the operation menu preferably includes the number of times the operation menu is accessed and transition information of the operation menu. The status observing unit further includes information on a currently selected operation menu, information indicating whether or not the machine tool is in a machining operation, alarm information of the numerical control device and the machine tool, and a program being edited. It is preferable to include at least one of the information indicating whether or not The machine learning device may further include a decision making unit that refers to the display of the operation menu learned by the learning unit and determines the position and the order of the operation menu displayed on the display device.

  The learning unit calculates the reward based on the output of the state observation unit, and the position and order of the operation menu displayed on the display device based on the outputs of the state observation unit and the reward calculation unit. A value function updating unit that updates the value function that determines the value of the value according to the reward. The reward calculation unit gives a positive reward when operated from the position of the operation menu which is easy to access and the menu arranged in order, and when operated from the position of the operation menu which is difficult to access and the menu arranged in order. It is preferable to give a negative reward.

  The learning unit is an output of the state observation unit, and an error calculation unit that calculates an error based on the input teacher data, and the display device based on the outputs of the state observation unit and the error calculation unit. A learning model updating unit that updates a learning model that determines the position and order error of the displayed operation menu. The machine learning device may include a neural network. The operator information includes information on the operator's position or authority level, and the operation menu based on the operator information is changed based on the operator's position or authority level information. Good.

  According to the second embodiment of the present invention, a detection unit that detects the operator, a communication unit that communicates with a database in which information of the operator is registered, and the machine learning device according to the first embodiment described above. And a display device for displaying the operation menu learned by the machine learning device.

  According to the third embodiment of the present invention, there is provided a machine tool system including a numerical control device, a machine tool controlled by the numerical control device, and the machine learning device according to the first embodiment described above. ..

  According to a fourth embodiment of the present invention, there is provided a manufacturing system comprising a plurality of machine tool systems according to the above-mentioned third embodiment, wherein the machine learning device is provided in each of the machine tool systems, and a plurality of the machine learning systems are provided. A manufacturing system is provided in which a plurality of machine learning devices provided in a machine tool system share or exchange data with each other via a communication medium. The machine learning device may exist on a cloud server.

  According to the fifth embodiment of the present invention, the machine learning for detecting the operator, communicating with the database in which the information of the operator is registered, and learning the display of the operation menu based on the information of the operator. A machine learning method for observing an operation history of the operation menu and learning the display of the operation menu based on the observed operation history of the operation menu. The operation history of the operation menu preferably includes the number of times the operation menu is accessed and transition information of the operation menu.

  According to the machine learning device, the numerical controller, the machine tool system, the manufacturing system and the machine learning method according to the present invention, it is possible to display the optimum operation menu for each operator.

FIG. 1 is a block diagram schematically showing an embodiment of a machine tool system according to the present invention. FIG. 2 is a block diagram showing an embodiment of the numerical controller according to the present invention. FIG. 3 is a diagram schematically showing a neuron model. FIG. 4 is a diagram schematically showing a three-layer neural network configured by combining the neurons shown in FIG. FIG. 5 is a flowchart showing an example of the operation of the machine learning device shown in FIG. FIG. 6 is a view (No. 1) for explaining the display of the operation menu learned by the machine learning device shown in FIG. FIG. 7 is a diagram (part 2) for explaining the display of the operation menu learned by the machine learning device shown in FIG. FIG. 8 is a view (No. 3) for explaining the display of the operation menu learned by the machine learning device shown in FIG. FIG. 9 is a block diagram schematically showing another embodiment of the machine tool system according to the present invention.

  Hereinafter, embodiments of a machine learning device, a numerical control device, a machine tool system, a manufacturing system, and a machine learning method according to the present invention will be described in detail with reference to the accompanying drawings. FIG. 1 is a block diagram schematically showing an embodiment of a machine tool system according to the present invention.

  As shown in FIG. 1, the machine tool system includes a machine tool 1, a machine learning device 2, and a numerical controller (NC device, CNC device) 3. Here, the machine tool 1 is, for example, a lathe, a drilling machine, a boring machine, a milling machine, a grinding machine, a gear cutting machine / gear finishing machine, a machining center, an electric discharge machine, a punch press, a laser machine, a carrier machine and a plastic injection molding machine. It is controlled by the numerical controller 3 including a machine. Although the machine learning device 2 is provided separately from the numerical control device 3 in FIG. 1, it may be provided as a part of the numerical control device 3. The display device 30 is, for example, a liquid crystal panel having a touch function, and displays the operation menu of the numerical control device 3. The display device 30 may be provided integrally with the numerical control device 3, but may be provided at a position separated from the numerical control device 3 by wire or wirelessly.

  The machine learning device 2 learns, for example, the display of the operation menu on the display device 30, and includes a state observation unit 21, a learning unit 22, and a decision making unit 25. The state observation unit 21 observes the operation history of the operation menu, for example, the state quantities (state variables) such as the number of times the operation menu is accessed and the transition information of the operation menu. Further, the state observing unit 21 is in the process of editing the information of the currently selected operation menu, the information indicating whether the machine tool is in the machining operation, the alarm information of the numerical control device and the machine tool, and the program being edited. At least one piece of information indicating whether or not there is a state quantity can be observed.

  Here, the state observing unit 21, for example, the number of times the operation menu is accessed, the transition information of the operation menu, the information of the operation menu currently selected, the information indicating whether the machine tool 1 is in the machining operation, and the numerical value. It is possible to receive from the numerical control device 3 alarm information of the control device 3 and information indicating whether or not the program is being edited. Furthermore, the state observing unit 21 can receive, for example, information indicating whether or not the machine tool 1 is in a machining operation, alarm information of the machine tool, and the like from the machine tool 1.

  The learning unit 22 learns the display of the operation menu based on the state quantity observed by the state observing unit 21, and includes a reward calculating unit 23 and a value function updating unit 24. The reward calculation unit 23 calculates the reward based on the output of the state observation unit 21, and the value function update unit 24 displays the operation menu displayed on the display device 30 based on the outputs of the state observation unit 21 and the reward calculation unit 23. The value function that determines the value of the position and order of is updated according to the reward. Here, the reward calculation unit 23 gives a positive reward when operated from the position of the operation menu which is easy to access and the menu arranged in order, and operates from the position of the operation menu which is difficult to access and the menu arranged in order. Give a negative reward when given. This will be described later with reference to FIG.

  FIG. 2 is a block diagram showing an embodiment of the numerical control device according to the present invention, and shows an example in which the machine learning device 2 shown in FIG. 1 is built in the numerical control device 3. 2 also shows how a plurality of machine learning devices 2 (21, 22, ..., 2n) share or exchange data with each other via a communication medium.

  As shown in FIG. 2, the numerical control device 3 includes a display device 30 for displaying the above-described operation menu, a detection unit 31, a communication unit 32, and a machine learning device 2 (21). The detection unit 31 detects the operator 6 based on a predetermined operation by the operator 6, for example, based on the input of a predetermined code by the operator 6 or the reading of an IC card held by the operator 6. To do. The communication unit 32 communicates with the database 5 provided outside the numerical control device 3, for example, based on the output of the detection unit 31.

  Here, the information of the operator 6 is registered in the database 5, for example, the position and authority level of the operator 6 are registered in advance. That is, in the output data D1 from the communication unit 32 to the machine learning device 2, information on the operator 6 corresponding to the operator 6 detected by the detection unit 31, for example, the operator 6 is a developer of a machine tool maker or It contains information on various positions and authority levels, such as person in charge, operator (user), or service engineer.

  Here, the detection unit 31 is not limited to the keyboard operated by the operator 6 and the reader of the IC card, and various known input devices and sensors may be used as long as the operator 6 can be detected. It is also possible to do so. In the present embodiment, it is preferable that the operation menu is changed based on the information of the operator 6 such as the detected position and authority level of the operator 6, for example. That is, it is preferable that the machine learning device 2 learns the optimum operation menu for the operator's position or authority level based on the operator's position or authority level and displays it on the display device 30.

  Further, as shown in FIG. 2, the plurality of machine learning devices 2 (21 to 2n) may be configured to share or exchange data with each other via a communication medium. For example, in a machine tool factory having a plurality of machine tools 1 each controlled by a numerical controller 3, that is, in a manufacturing system having a plurality of machine tool systems (1, 2), a machine learning device for each machine tool system. 2 (21 to 2n) can share or exchange data with each other via a communication medium. The machine learning device 2 (21 to 2n) may be provided on the cloud server instead of being provided on each numerical control device 3, for example.

  As described with reference to FIG. 1, the machine learning device 2 (state observing unit 21), for example, outputs the output data D2 of the display device 30 as the operation history of the operation menu (state amount: for example, the access count of the operation menu). Change data D3 (operation amount: for example, the order of menu display and the menu display) that controls the operation menu displayed on the display device 30 by performing machine learning (for example, reinforcement learning) by receiving the information and transition information of the operation menu). Position etc.). It is needless to say that FIG. 2 is merely an example, and various modifications and changes can be made.

  According to the present embodiment, for example, when an operator operates a machine (machine tool system) for the first time, if he / she learns the operation history of a person who is in a position close to the operator or has an authority level, it is appropriate to some extent from the beginning. The menu can be displayed. Also, every time the operating person's position or authority level increases, it becomes possible to gradually display an appropriate menu without creating a dedicated table.

  By the way, the machine learning device 2 extracts useful rules, knowledge expressions, judgment criteria, etc., contained in the data input to the device by analysis, outputs the judgment results, and learns knowledge ( It has the function of performing machine learning). Although there are various machine learning methods, they can be roughly classified into, for example, “supervised learning”, “unsupervised learning”, and “reinforcement learning”. Furthermore, in order to realize these methods, there is a method called "deep learning" that learns extraction of the feature amount itself.

  The machine learning device 2 shown in FIG. 1 applies “reinforcement learning (Q learning)”, and the machine learning device 4 described later with reference to FIG. 9 applies “supervised learning”. It was done. For these machine learning (machine learning devices 2 and 4), a general-purpose computer or processor can be used, but if GPGPU (General-Purpose computing on Graphics Processing Units) or a large-scale PC cluster is applied, for example, High-speed processing becomes possible.

  First, with supervised learning, a large amount of teacher data, that is, a set of input and result (label) data, is given to a machine learning device to learn features in those data sets, and to obtain results from the input. The model (error model) to be estimated, that is, the relationship is recursively acquired. For example, it can be realized using an algorithm such as a neural network described later.

  In addition, unsupervised learning learns what kind of distribution the input data has by giving a large amount of input data only to the machine learning device. On the other hand, this is a method of learning with a device that performs compression, classification, shaping, and the like. For example, features in those datasets can be clustered among similar individuals. Output prediction can be realized by using the result and assigning the output so as to set some standard and optimize it.

  As an intermediate problem setting between unsupervised learning and supervised learning, there is also what is called semi-supervised learning. This is because, for example, only a part of input and output data sets exists, and other than that, This corresponds to the case where the data is input only. In the present embodiment, data (image data, simulation data, etc.) that can be acquired without actually moving the machine tool system (machine tool 1 and numerical controller 3) is used by unsupervised learning. , It becomes possible to perform learning efficiently.

Next, reinforcement learning will be described. First, consider the following as a problem setting for reinforcement learning.
A machine tool system (that is, a machine tool and a numerical control device: in the following description, also referred to as a numerical control device for simplification of description) observes a state of an environment and determines an action.
-The environment changes according to some rules, and in addition, one's behavior may change the environment.
・ Every time you act, a reward signal comes back.
・ What we want to maximize is the total of (discount) rewards in the future.
・ Learning starts from a state of completely or incompletely knowing the consequences of actions. That is, the numerical control device can obtain the result as data only after actually acting. In other words, it is necessary to search for the optimum behavior by trial and error.
-It is possible to start learning from a good starting point by setting the state of pre-learning (methods such as supervised learning and inverse reinforcement learning described above) as the initial state so as to imitate human movements.

  Here, reinforcement learning means not only judgment and classification but also learning behaviors to learn appropriate behaviors based on the interaction of behaviors with the environment, that is, maximizing rewards obtained in the future. Is for learning how to learn. Hereinafter, as an example, the description will be continued in the case of Q learning, but the present invention is not limited to Q learning.

  Q-learning is a method of learning the value Q (s, a) of selecting an action a under a certain environmental condition s. That is, in a certain state s, the action a having the highest value Q (s, a) may be selected as the optimum action. However, at first, the correct value of the value Q (s, a) is not known for the combination of the state s and the action a. Therefore, the agent (action subject) selects various actions a under a certain state s, and a reward is given to the action a at that time. Thereby, the agent learns a better action selection, that is, the correct value Q (s, a).

The results of behavioral, we want to maximize the sum of the rewards future, finally Q (s, a) = E aims to _{^{[Σ (γ t) r t}} ] become so. Here, the expected value is taken when the state changes in accordance with the optimum action, and since it is not known, learning is performed while searching. Such an updating formula of the value Q (s, a) can be expressed by the following formula (1), for example.

上記の式(１)において、ｓ_tは、時刻ｔにおける環境の状態を表し、ａ_tは、時刻ｔにおける行動を表す。行動ａ_tにより、状態はｓ_t+1に変化する。r_t+1は、その状態の変化により得られる報酬を表している。また、ｍａｘの付いた項は、状態ｓ_t+1の下で、その時に分かっている最もＱ値の高い行動ａを選択した場合のＱ値にγを乗じたものになる。ここで、γは、０＜γ≦１のパラメータで、割引率と呼ばれる。また、αは、学習係数で、０＜α≦１の範囲とする。

In the above formula (1), s _t represents the state of the environment at time t, a _t represents the action at time t. By the action a _t, the state changes to s _{t + 1.} r _{t + 1} represents the reward obtained by changing the state. In addition, the term with max is the Q value when the action a with the highest Q value known at that time is selected under the state _{st + 1} , and is multiplied by γ. Here, γ is a parameter of 0 <γ ≦ 1 and is called a discount rate. Further, α is a learning coefficient, and is set in a range of 0 <α ≦ 1.

The above-mentioned formula (1) as a result of the trial a _t, based on the reward r _{t + 1} came back, represents a method for updating the evaluation value Q of the action a _t in state _{s t (s t, a t} ) ing. In other words, the evaluation value Q (s _t, a _t) of the action a in the state s than, reward r _{t +} 1 ₊ evaluation value Q of the best action max a in the next state by the action a (s _t + 1, max If a _{t + 1} ) is larger, then Q (s _t , a _t ) is increased, and if smaller, Q (s _t , a _t ) is decreased. That is, the value of a certain action in a certain state is brought closer to the reward that immediately returns as a result and the value of the best action in the next state due to the action.

  Here, the method of expressing Q (s, a) on the computer is to hold the values of all state-action pairs (s, a) as a table, and to use Q (s, a) There is a method to prepare a function that approximates. In the latter method, the above equation (1) can be realized by adjusting the parameters of the approximation function by a method such as the stochastic gradient descent method. A neural network described later can be used as the approximation function.

  Also, a neural network can be used as an approximation algorithm of the value function in the reinforcement learning. FIG. 3 is a diagram schematically showing a neuron model, and FIG. 4 is a diagram schematically showing a three-layer neural network configured by combining the neurons shown in FIG. That is, the neural network is composed of, for example, an arithmetic unit and a memory that imitate a neuron model as shown in FIG.

As shown in FIG. 3, the neuron outputs an output (result) y for a plurality of inputs x (in FIG. 3, inputs x1 to x3 as an example). Each input x (x1, x2, x3) is multiplied by the weight w (w1, w2, w3) corresponding to this input x. As a result, the neuron outputs the result y expressed by the following equation (2). The input x, the result y, and the weight w are all vectors. Further, in the following formula (2), θ is a bias, and f _k is an activation function.

  A three-layer neural network configured by combining the neurons shown in FIG. 3 will be described with reference to FIG. As shown in FIG. 4, a plurality of inputs x (here, input x1 to input x3 as an example) are input from the left side of the neural network, and a result y (here, result y1 to input y3 as an example here) is input from the right side. ) Is output. Specifically, the inputs x1, x2, and x3 are input with corresponding weights applied to each of the three neurons N11 to N13. The weights applied to these inputs are collectively labeled as W1.

  The neurons N11 to N13 output z11 to z13, respectively. In FIG. 4, z11 to z13 are collectively labeled as a feature vector Z1 and can be regarded as a vector obtained by extracting the feature amount of the input vector. The feature vector Z1 is a feature vector between the weight W1 and the weight W2. z11 to z13 are input with corresponding weights applied to each of the two neurons N21 and N22. The weights applied to these feature vectors are collectively labeled as W2.

  The neurons N21 and N22 output z21 and z22, respectively. In FIG. 4, these z21 and z22 are collectively referred to as a feature vector Z2. The feature vector Z2 is a feature vector between the weight W2 and the weight W3. z21 and z22 are inputted with corresponding weights applied to each of the three neurons N31 to N33. The weights applied to these feature vectors are collectively labeled as W3.

  Finally, the neurons N31 to N33 output the result y1 to the result y3, respectively. The operation of the neural network has a learning mode and a value prediction mode. For example, in the learning mode, the weight W is learned using the learning data set, and the behavior of the numerical controller is determined in the prediction mode using the parameter. It should be noted that although it is written as prediction for convenience, it goes without saying that various tasks such as detection, classification, and inference can be performed.

  Here, the data obtained by actually operating the numerical control device in the prediction mode can be immediately learned and reflected in the next action (online learning), or the learning collected using the data group collected in advance. It is also possible to perform the detection mode with that parameter after that (batch learning). Alternatively, it is also possible to sandwich the learning mode every time some intermediate data is accumulated.

  The weights W1 to W3 can be learned by an error back propagation method (error back propagation method: back propagation). Note that the error information enters from the right side and flows to the left side. The error back-propagation method is a method of adjusting (learning) the respective weights for each neuron so as to reduce the difference between the output y when the input x is input and the true output y (teacher). Such a neural network can increase the number of layers to three or more (called deep learning). Further, it is also possible to automatically acquire the arithmetic unit that performs the feature extraction of the input stepwise and regresses the result from only the teacher data.

  Therefore, as described above, the machine learning device 2 of the present embodiment includes, for example, the state observing unit 21, the learning unit 22, and the decision making unit 25 in order to carry out the Q learning. However, the machine learning method applied to the present invention is not limited to Q learning. As described above, the machine learning (machine learning device 2) can be realized by applying, for example, GPGPU or a large-scale PC cluster.

  FIG. 5 is a flowchart showing an example of the operation of the machine learning device shown in FIG. As shown in FIG. 5, when machine learning is started (learning start), the position or authority level of the operator is acquired in step ST1, and the process proceeds to step ST2. In step ST2, the operation history is acquired, and the process proceeds to step ST3. Here, the operation history includes, for example, the menu access count and transition information.

  In step ST3, the menu is displayed, in step ST4, the operator selects the menu, and in step ST5, it is determined whether or not the menu is operated from the menu at a position where it is easy to operate. If it is determined that the menu is operated at a position where it is difficult to operate, the procedure proceeds to step ST8 to set a negative reward, and if it is determined that it is operated from a menu at a position that is difficult to operate, the procedure proceeds to step ST6 to set a positive reward.

  In this way, the rewards set in steps ST6 and ST8 are summarized by the reward calculation in step ST7, and the process proceeds to step ST9 to update the action value table based on the reward calculated in step ST7. ..

  That is, the value function updating unit 24 updates the value function (action value table). Then, for example, the decision making unit 25 decides the position and the order of the operation menu displayed on the display device 30, based on the value function updated by the value function updating unit 24. When the process of step ST9 ends, the process returns to step ST1 and the same process is repeated. As described above, according to this embodiment, it is possible to display the optimum operation menu for each operator.

  6 to 8 are views for explaining the display of the operation menu learned by the machine learning device shown in FIG. Here, FIG. 6A and FIG. 6B are views for explaining the manner in which the operation menu is updated (learned) based on the access count (state amount) of the operation menu, and FIG. 8A to 8B are diagrams for explaining how to learn the operation menu based on the transition information (state amount) of the operation menu. Further, in order to simplify the description, in each figure (the display screen 300 of the display device 30 on which the operation menu is displayed), for example, the positions that are easy to operate as the operation menu are the upper side and the left side, and the positions that are difficult to operate are the lower side. On the side and on the right.

  FIG. 6A shows an operation menu (home screen) in an initial state, and FIG. 6B shows an example accessed by an operator (for example, a person in charge of a machine tool maker (MTB)). Note that reference symbols P1 to P6 indicate the positions of the respective icons of the operation menu displayed on the display screen 300 of the display device 30. As shown in FIG. 6A, in the operation menu in the initial state, a "browser" icon is displayed at position P1 on the display screen 300, a "memo" icon is displayed at position P2, and A "manual" icon is displayed at position P3. Further, a "NC operation" icon is displayed at position P4 on the display screen 300, a "maintenance" icon is displayed at position P5, and a "setting" icon is displayed at position P6.

  Here, the "browser" icon is used to browse the Internet, the "memo" icon is used to display the memo screen, and the "manual" icon is used to browse the manual. It These "browser", "memo" and "manual" icons are arranged in the row direction to form [category A]. The "NC operation" icon is used to create a machining program and check the machining status, the "Maintenance" icon is used to check machine alarm information, and "Setting". The icon is used when setting parameters and the like. These "NC operation", "maintenance" and "setting" icons are arranged in the row direction to form [Category B].

  As shown in FIG. 6 (b), for example, when the person in charge of MTB accesses the operation menu (display screen 300) as the operator 6, the number of times of operation of each icon is 2 times for the “browser” icon and “ If the "Memo" icon is once, the "Manual" icon is three times, the "NC operation" icon is five times, the "Maintenance" icon is once, and the "Settings" icon is ten times, then [Category A] is selected. The total number of operations of the constituent icons is 6 and the total number of operations of the icons forming [Category B] is 16. Therefore, [Category B] is easier to operate than [Category A]. Is changed to the line.

  Furthermore, in each [Category A] and [Category B], the icons are arranged in the descending order of the number of times of operation from the left side where the operation is easy. That is, a "setting" icon is displayed at position P1, a "NC operation" icon is displayed at position P2, and a "maintenance" icon is displayed at position P3. A "manual" icon is displayed at position P4, a "browser" icon is displayed at position P5, and a "memo" icon is displayed at position P6.

  7 (a) and 7 (b), for example, the person in charge of the MTB as the operator 6 operates the “setting” icon in FIG. 6 (a) (for example, touch operation), and then the display screen 300 is displayed. An example of the operation menu to be displayed is shown, and two display screens (operation menu) of <first page> and <second page> can be selected by the "page forward" and "page return" icons. Reference numerals P11 to P15 and P21 to P25 indicate the positions of the respective icons of the <first page> and <second page> operation menus displayed on the display screen 300 of the display device 30. Here, the "page forward" icon on the <first page> display screen is fixed at position P15, and the "page return" icon on the <second page> display screen is fixed at position P25. To do.

  Then, the person in charge of the MTB as the operator 6 performs machine learning by the machine learning device 2 based on the number of times the icons “A” to “H” have been operated in the past. For example, as shown in FIG. The operation menu is displayed as shown in FIG. 8 (b). For example, the number of operations of each icon “A” to “H” is 10 times for the “A” icon, 8 times for the “B” icon, 7 times for the “C” icon, 5 times for the “D” icon, When the “E” icon is 6 times, the “F” icon is 9 times, the “G” icon is 4 times, and the “H” icon is 3 times, the operation is easy as shown in FIG. 8 (a). From left to right on the <first page>, in order of the number of operations, the position P11 is "A", P12 is "F", P13 is "B", P14 is "C", and P15 is " The "Send page" icon is arranged, and further, as shown in FIG. 8 (b), from the left to the right of the <second page>, "E" is set at position P21 and "E" is set at P22 in the order of the number of operations. "D", P23, "G", P24, "H", and P25, "page return" icons.

  In the above description, the operation history (the number of times of access to the operation menu and the transition information) and the operation menu have been simplified, but the state quantities (operation history etc.) input to the machine learning device 2 and the output from the machine learning device 2 are described. It goes without saying that the amount of operation to be performed (learned operation menu, etc.), and the position and authority level of the operator 6 can be variously modified and changed.

  FIG. 9 is a block diagram schematically showing another embodiment of the machine tool system according to the present invention, to which supervised learning is applied. As is clear from the comparison between FIG. 9 and FIG. 1 described above, the machine tool system to which the supervised learning shown in FIG. 9 is applied is a machine tool system to which the Q learning (reinforcement learning) shown in FIG. 1 is applied. Data (data with result (label)) is provided.

  As shown in FIG. 9, the machine learning device 4 in the machine tool system to which the supervised learning is applied includes a state observing section 41, a learning section 42, and a decision making section 45. The learning unit 42 includes an error calculation unit 43 and a learning model updating unit (error model updating unit) 44. Also in the machine tool system of the present embodiment, the machine learning device 4 learns the display of the operation menu based on the operation history of the operation menu.

  That is, the state observing unit 41 observes the operation history of the operation menu, for example, the state quantities such as the number of times of accessing the operation menu and the transition information of the operation menu, as in the state observing unit 21 in FIG. 1. Further, the state observing unit 41 is in the process of editing the information of the currently selected operation menu, the information indicating whether the machine tool is in the machining operation, the alarm information of the numerical control device and the machine tool, and the program being edited. At least one piece of information indicating whether or not there is a state quantity can be observed.

  As shown in FIG. 9, the learning unit 42 includes an error calculation unit 43 and a learning model updating unit 44, and the error calculation unit 43 and the learning model updating unit 44 respectively perform the operation to which the Q learning shown in FIG. 1 is applied. It corresponds to the reward calculation unit 23 and the value function update unit 24 in the mechanical system. However, the learning model updating unit 44 updates the learning model so that the difference between the training data and the learning model (error model) is input to the error calculating unit 43 in the present embodiment from the outside. The configuration and the like are different from those described with reference to FIG.

  That is, the error calculation unit 43 receives the output of the state observation unit 41 and the teacher data, and calculates the error between the result (labeled) data and the learning model implemented in the learning unit 42. Here, as the teacher data, for example, when the same work is performed by the same machine tool system, the labeled data obtained up to the day before the predetermined day to actually perform the work is held, and the predetermined day, It can be provided to the error calculation unit 43 as teacher data.

  Alternatively, the data obtained by a simulation performed outside the machine tool system (numerical control device, machine tool, etc.) or the labeled data of another machine tool system can be processed by a memory card or communication line. It is also possible to provide it as error data to the error calculation unit 43 of the mechanical system. Further, the teacher data (labeled data) is held in, for example, a nonvolatile memory such as a flash memory (Flash Memory) built in the learning unit 42, and the labeled data held in the nonvolatile memory is directly used as the learning unit. It can also be used at 42.

  In the above, when considering a manufacturing system including a plurality of machine tool systems, for example, the machine learning device 2 (4) is provided for each machine tool system, and a plurality of machine learning devices provided for a plurality of machine tool systems are provided. 2 (4) can share or exchange data with each other via a communication medium. Further, the machine learning device 2 (4) can be made to exist on the cloud server.

  As described above, as the machine learning device according to the present invention, not only “reinforcement learning” but also “supervised learning” or various machine learning methods such as “unsupervised learning” and “semi-supervised learning” Can be applied.

  Although the embodiments have been described above, all the examples and conditions described here are described for the purpose of helping understanding of the concept of the invention applied to the invention and the technology, and the described examples and conditions are It is not intended to limit the scope of the invention. Nor does such a description in the specification indicate the advantages and disadvantages of the invention. While the embodiments of the invention have been described in detail, it should be understood that various changes, substitutions, and changes can be made without departing from the spirit and scope of the invention.

1 Machine tool 2, 4 Machine learning device 3 Numerical control device 5 Database 6 Operator 21, 41 State observation part 22, 42 Learning part 23 Reward calculation part 24 Value function updating part 25, 45 Decision making part 30 Display device 31 Detection part 32 communication unit 43 error calculation unit 44 learning model updating unit

Claims (15)

A machine learning device that detects an operator, communicates with a database in which information of the operator is registered, and learns to display an operation menu based on the information of the operator,
A state observation unit for observing the operation history of the operation menu,
A learning unit for learning the display of the operation menu based on the operation history of the operation menu observed by the state observation unit,
A machine learning device characterized by that. The operation history of the operation menu is
Including the number of times the operation menu is accessed and transition information of the operation menu,
The machine learning device according to claim 1, wherein: The state observation unit further includes
Information on the currently selected operation menu, information indicating whether or not the machine tool is in machining operation, alarm information for the numerical control device and the machine tool, and information indicating whether or not a program is being edited. Including at least one of,
The machine learning device according to claim 1 or 2, characterized in that. further,
The learning unit is provided with a decision making unit that decides the position and order of the operation menu displayed on the display device with reference to the display of the operation menu learned
The machine learning device according to claim 1, wherein the machine learning device is a machine learning device. The learning unit is
A reward calculation unit that calculates a reward based on the output of the state observation unit,
A value function updating unit that updates the value function that determines the value of the position and order of the operation menu displayed on the display device based on the output of the state observing unit and the reward calculating unit, and a value function updating unit that updates according to the reward.
The machine learning device according to any one of claims 1 to 4, wherein: The reward calculator is
Give a positive reward when operated from the position of the operation menu that is easy to access and the menu arranged in order,
Give a negative reward when operated from the position of the operation menu that is difficult to access and the menu arranged in order,
The machine learning device according to claim 5, wherein: The learning unit is
An output of the state observation unit, and an error calculation unit that calculates an error based on the input teacher data,
A learning model updating unit that updates a learning model that determines an error in the position and order of the operation menu displayed on the display device based on the outputs of the state observing unit and the error calculating unit,
The machine learning device according to any one of claims 1 to 4, wherein: The machine learning device comprises a neural network,
The machine learning device according to claim 1, wherein the machine learning device is a machine learning device. The operator information includes information on the operator's position or authority level,
The operation menu based on the information of the operator changes based on information of the position or authority level of the operator,
9. The machine learning device according to claim 1, wherein the machine learning device is a machine learning device. A detection unit for detecting the operator,
A communication unit that communicates with a database in which the information of the operator is registered,
A machine learning device according to any one of claims 1 to 9,
A display device for displaying an operation menu learned by the machine learning device,
Numerical control device characterized in that. A numerical controller, a machine tool controlled by the numerical controller, and the machine learning device according to claim 1.
Machine tool system characterized by A manufacturing system comprising a plurality of machine tool systems according to claim 11,
The machine learning device is provided in each of the machine tool systems,
The plurality of machine learning devices provided in the plurality of machine tool systems are configured to share or exchange data with each other via a communication medium.
A manufacturing system characterized by the above. The machine learning device exists on a cloud server,
The manufacturing system according to claim 12, wherein: A machine learning method of detecting an operator, communicating with a database in which information of the operator is registered, and learning to display an operation menu based on the information of the operator,
Observe the operation history of the operation menu,
Learning the display of the operation menu based on the operation history of the observed operation menu,
A machine learning method characterized by the above. The operation history of the operation menu is
Including the number of times the operation menu is accessed and transition information of the operation menu,
15. The machine learning method according to claim 14, wherein:

Images