DE112018007741B4 - MACHINE LEARNING DEVICE AND DEVICE FOR GENERATING PROGRAMS FOR NUMERICALLY CONTROLLED MACHINING - Google Patents

Abstract

Machine learning device (50), comprising: a state observation unit (51) configured to process shape data as state variables, which includes information about a final machining shape of a machined object and information about a material of a blank, and a method for generating a program for numerically controlled machining to automatically generate a numerically controlled machining program comprising multiple machining operations for machining a workpiece into the machined object; and a learning unit (52) which is designed to learn the method for generating a program for numerically controlled processing based on a data set generated on the basis of the state variables, the learning unit (52) having: a reward calculation unit (53) which is designed to to calculate a reward; and a function update unit (54), which is designed to update a function for determining the method for generating a program for numerically controlled processing based on the calculated reward; characterized in that the status observation unit (51) is further designed to edit a first as a status variable observe a numerically controlled machining program created with reference to the method for generating a numerically controlled machining program; and the reward calculation unit (53) is designed to calculate the reward based on a method for generating an editing-integrated program for numerically controlled editing, which results from the inclusion of the editing in the method for generating the program for numerically controlled editing.

Description

Area

The present invention relates to a machine learning apparatus and a numerically controlled machining program generating apparatus, each for learning a method for generating a numerically controlled machining program used in automatically generating a program for numerically controlled machining is to be used for the numerical control of a machine tool.

background

Recently, in the field of numerically controlled (NC) machine tools, multifunctional machines in which various types of tools are interchangeably arranged on the tool rest are increasingly used to enable various types of machining processes such as turning, drilling and milling, which provide automatic machining Making it possible to produce products with a variety of complex shapes.

It is known from the essay “Generation of Modified Cutting Condition Using Neural Network for an Operation Planning System” by Mr. Myon-Woong Park et al., which appeared in the “Annals of the CIRP Vol. 45/1/1996”, a neural Network to use when determining control data for a lathe.

The paper “Machine-learning techniques and their applications in manufacturing” by D. T. Pham and A. A. Afify, which is included in the “Proc. IMechE Vol. 219 Part B: J. Engineering Manufacture” published under the number DOI: 10. 1243/095440505X32274 proposes to use machine learning algorithms and neural networks for the purpose of automation in the production of goods.

In the Japanese publication JP 2002 - 132 313 A discloses a device for generating a program for numerically controlled machining which stores in a storage unit the change in machining conditions made by a user with respect to an automatically generated machining program so that the changed machining conditions can be used again when another user at a later time would like to create a similar editing program.

Short description

Technical problem

The apparatus for generating programs for numerically controlled machining according to the above-mentioned Japanese Patent Application Laid-Open JP 2002 - 132 313 A However, only allows the reuse of cutting conditions such as the rotation speed of the main shaft for rotating the blank to be machined, the relative cutting feed rate between the turning tool and the blank to be machined and the chamfer, and does not automatically generate a machining program in terms of the machining order and machining method supports.

The present invention has been made in view of the above, and an object of the present invention is to provide a machine learning apparatus which can automatically learn a method of generating a program for numerically controlled machining in terms of the machining order and/or machining method enables the automatic generation of a machining program.

Solution to the problem

In order to solve the above problem and achieve the object, a machine learning device according to the present invention comprises the combination of the features of independent claim 1. A preferred development can be found in dependent claim 2.

The object is further achieved by a device for generating a program for numerically controlled processing with the combination of the features of independent claim 3.

Advantageous effects of the invention

The present invention offers the advantage of providing a machine learning device which can automatically learn a method for generating a program for numerically controlled machining, which enables automatic generation of a machining program with respect to the machining order and/or machining method.

Short description of the characters

• 1 Fig. 12 is a block diagram illustrating an example of the configuration of a numerical controller according to a first embodiment of the present invention.
• 2 shows a representation to illustrate the principle of an action value table.
• 3 Fig. 12 is a diagram illustrating an example of editing knowledge and editing-integrated editing knowledge stored in the function update unit of the machine learning device according to the first embodiment of the present invention.
• 4 shows a flowchart for illustrating the flow of a method for updating an action value function Q(s, a) in the first embodiment of the present invention.
• 5 Fig. 12 is a flowchart illustrating a detail of a procedure for generating an NC machining program performed in the NC machining program generating unit of the NC machining program generating apparatus according to the first embodiment of the present invention.
• 6 shows a perspective view for illustrating an example of a three-dimensional model of the final machining shape represented by CAD (Computer-Aided Design) data, according to the first embodiment of the present invention.
• 7 Fig. 12 is a perspective view illustrating an example of a blank shape surrounding the final machining shape represented by CAD data according to the first embodiment of the present invention.
• 8th Fig. 12 is a perspective view illustrating an example of a cutting shape generated from the final machining shape and the blank shape represented by CAD data according to the first embodiment of the present invention.
• 9 1 is a diagram showing the blank shape, a turning cross-sectional shape generated from the machining shape, and coordinate values according to the first embodiment of the present invention.
• 10 shows a representation to illustrate a turning cross-sectional shape in a facing step, according to the first embodiment of the present invention.
• 11 1 is a diagram illustrating a turning cross-sectional shape in a drilling turning step according to the first embodiment of the present invention.
• 12 Fig. 12 is a diagram illustrating a turning cross-sectional shape in a bar material turning processing step according to the first embodiment of the present invention.
• 13 Fig. 12 is a diagram illustrating a turning cross-sectional shape in a bar material turning processing step according to the first embodiment of the present invention.
• 14 1 is a diagram showing an example of a machining start point and a machining end point in the facing step according to the first embodiment of the present invention.
• 15 Fig. 12 is a diagram showing an example of a machining start point and a machining end point in the drilling rotation step according to the first embodiment of the present invention.
• 16 Fig. 12 is a diagram showing an example of a cutting point, machining start point and machining end point in the bar material turning machining step according to the first embodiment of the present invention.
• 17 Fig. 12 is a diagram showing an example of a cutting point, machining start point and machining end point in the bar material turning machining step according to the first embodiment of the present invention.
• 18 Fig. 12 is a diagram illustrating a turning cross-sectional shape in the drilling turning step according to the first embodiment of the present invention.
• 19 1 shows a diagram for illustrating a turning cross-sectional shape in the facing step according to the first embodiment of the present invention.
• 20 Fig. 12 is a diagram illustrating a turning cross-sectional shape in the bar material turning processing step according to the first embodiment of the present invention.
• 21 Fig. 12 is a diagram illustrating a turning cross-sectional shape in the bar material turning processing step according to the first embodiment of the present invention.
• 22 Fig. 12 is a diagram showing an example of a machining start point and a machining end point in the drilling rotation step according to the first embodiment of the present invention.
• 23 1 is a diagram showing an example of a machining start point and a machining end point in the facing step according to the first embodiment of the present invention.
• 24 Fig. 12 is a diagram illustrating an example of a cutting point, machining start point and machining end point in a bar end turning step according to the first embodiment of the present invention.
• 25 1 is a diagram showing an example of a cutting point, machining start point and machining end point in a bar grooving turning step according to the first embodiment of the present invention.
• 26 Fig. 12 is a schematic diagram illustrating an example of a neural network model used by the machine learning device according to the first embodiment of the present invention.
• 27 Fig. 12 is a diagram showing a hardware configuration of the NC machining program generating apparatus according to the first embodiment of the present invention.

Description of an embodiment

Below, a machine learning device, a device for generating a program for numerically controlled machining and a method for machine learning according to an embodiment of the present invention will be described in detail with reference to the figures. It is noted that this embodiment is not intended to limit the scope of the invention.

First embodiment

1 Fig. 12 shows a block diagram for illustrating an example of a configuration of a numerical controller 1 according to a first embodiment of the present invention. As in 1 As shown, the numerical controller 1 includes an interactive operation processing unit 10, a display unit 20, an instruction input unit 30, and an NC machining program generating device 40. The NC machining program generating device 40 includes a machine learning device 50. The machine learning device 50 includes a state observation unit 51 and a learning unit 52. The numerical controller 1 is attached to or connected to a numerically controlled machine tool (not shown) to numerically control the operation of the numerically controlled machine tool using an NC machining program.

The interactive operation processing unit 10 is an interface unit between the numerical controller 1 and a user and an interface unit between the NC machining program generating device 40 and the user. The interactive operation processing unit 10 transmits instruction information inputted by the user via the instruction input unit 30 to the NC machining program generating device 40. The interactive operation processing unit 10 also displays the instruction information entered by the user via the instruction input unit 30 on the display unit 20.

The display unit 20 is a display terminal such as a liquid crystal display, and the display unit 20 displays CAD data 100, an NC machining program, and the instruction information input by the user via the instruction input unit 30. The display unit 20 also displays various types of information related to the processing performed by the numerical controller 1 and the NC machining program generating device 40.

The instruction input unit 30 is realized by input devices such as a mouse and a keyboard, receives instruction information from the user, and transmits the instruction information to the interactive operation processing unit 10.

The NC machining program generating device 40 generates an NC machining program used for machining a workpiece into the shape of a machined object represented by the CAD data 100 input from outside the numerical controller 1 NC machining program generating device 40 can be entered. An NC machining program is a program for numerically controlled machining that is used to operate an NC machine tool (not shown). In addition, the numerical controller 1 controls the NC machine tool according to the NC machining program to perform machining of the workpiece to produce the machined object.

The NC machining program generating device 40 includes a machining shape data input unit 41, a machining shape data storing unit 42, an NC machining program generating unit 43, an NC machining program storing unit 44, an NC machining program editing unit 45 and a decision making unit 46. The components the NC machining program generating device 40 are designed so that information can be sent to and received from each other.

The machining shape data input unit 41 receives the CAD data 100, which is machining shape data input into the NC machining program generating device 40 from an external device outside the numerical controller 1, and transmits the CAD data 100 to the machining shape data input unit 41 Machining shape data storage unit 42. The machining shape data includes information about the final machining shape of the machined object and information about the material of the blank. The final machining shape is a three-dimensional shape to be created by machining to obtain the machined object. The blank is the workpiece that is to be machined into the final machining shape, which is represented by the CAD data 100. However, the machining shape data is not limited to the CAD data 100. The machining shape data may be any data that can be interpreted by the NC machining program generating device 40.

The machining shape data storage unit 42 stores the machining shape data transmitted from the machining shape data input unit 41.

The NC machining program generating unit 43 generates an NC machining program for machining the blank into the final machining shape represented by the CAD data 100. Specifically, based on the machining shape data of the machined object, the NC machining program generating unit 43 sets a machining order and a machining method regarding a plurality of machining processes with reference to machining knowledge or editing-integrated machining knowledge obtained from the function updating unit 54 via the decision making unit 46. The NC machining program generating unit 43 then generates a first NC machining program 111, which is the NC machining program for machining the workpiece into the machining object, based on the specified machining order and machining method. Further, the NC machining program generating unit 43 links the CAD data 100, which is the machining shape data, the machining knowledge referred to in generating the first NC machining program 111, and the first NC machining program 111 with each other and transfers the CAD data 100, the first NC machining program 111 and the machining knowledge to the NC machining program storage unit 44 and stores them in the NC machining program storage unit 44.

Machining knowledge is a set of determination conditions that are referenced to determine the order of machining the blank and/or the method of machining the blank with respect to multiple machining operations for machining the blank into the machined object, if the NC machining program generating unit 43 automatically generates an NC machining program to machine a blank into the final machining shape of the machined object represented by the CAD data 100. A machining knowledge is also information that specifies the procedure for generating an NC machining program.

More specifically, a machining knowledge is a set of conditions that specifies the order of machining the blank and/or the method of machining the blank based on the dimensions of the cutting shape, i.e. H. the shape that is to be machined in order to bring the blank into the machined object. The order of machining the blank is the order of machining operations when several different machining operations are combined to bring the blank into the machined object. The material processing method is a processing method that includes the processing direction.

As will be described below, by editing the first NC machining program 111 automatically generated in the NC machining program generating unit 43 by the user, an editing-integrated machining knowledge, which is a new machining knowledge that includes the editing, is generated. Accordingly, the new editing knowledge integrates and accumulates the editing that reflects the user's knowledge and experience. Therefore, generation of an NC machining program by the NC machining program generating unit 43 with reference to the machining knowledge enables rapid automatic generation of a user-desired NC machining program similar to an NC machining program generated by the user himself.

The NC machining program storage unit 44 stores the CAD data 100, which is the machining shape data, the first NC machining program 111, which is the NC machining program generated in the NC machining program generating unit 43, and the machining knowledge that was referred to when generating the first NC machining program 111 is linked together. The NC machining program storage unit 44 further stores a second NC machining program 112, which is an NC machining program edited in the NC machining program editing unit 45. It is noted that the first NC machining program 111 can be updated by the second NC machining program 112.

The NC machining program editing unit 45 reads the CAD data 100, the first NC machining program 111 and the machining knowledge associated with the first NC machining program 111 from the NC machining program storage unit 44 and transmits the first NC machining program 111 to the Interactive operation processing unit 10. The NC machining program editing unit 45 edits the first NC machining program 111 based on the editing instruction information inputted by the user via the instruction input unit 30 and generates the second NC machining program 112. The NC machining program editing unit 45 transmits the generated second NC machining program 112 to the NC machining program storage unit 44 and stores it therein. In addition, the NC machining program editing unit 45 transmits the CAD data 100, which is the machining shape data, the first NC machining program 111, the machining knowledge associated with the first NC machining program 111, and the second NC machining program 112 the status observation unit 51 described later.

If no editing operation was performed for the first NC machining program 111, i.e. H. If no second NC machining program 112 has been generated, the NC machining program editing unit 45 transmits the CAD data 100, the first NC machining program 111 and the machining knowledge linked to the first NC machining program 111 to the state observation unit 51.

The decision making unit 46 searches through a plurality of different machining knowledge stored in the function updating unit 54 and determines and obtains the machining knowledge most suitable for the CAD data 100 in response to a request from the NC machining program generating unit 43. The machining knowledge required for the CAD data 100 is most suitable is the machining knowledge obtained after the first NC machining program 111 is generated Machining knowledge that corresponds to the CAD data 100 stored in the function update unit 54 has required the smallest amount of editing operations on the first NC machining program 111. This means that the machining knowledge that is most suitable for the CAD data 100 is the machining knowledge that best reflects the user's knowledge and experience among the machining knowledge that corresponds to the CAD data 100.

The decision making unit 46 obtains the machining knowledge with the highest action value Q among the plurality of machining knowledge corresponding to the CAD data 100 stored in the function updating unit 54, and transmits this machining knowledge to the NC machining program generating unit 43. It is noted that if If only machining knowledge corresponding to the CAD data 100 is stored in the function update unit 54, the decision-making unit 46 determines that the machining knowledge stored in the function update unit 54 is the most suitable machining knowledge for the CAD data 100.

Machining knowledge is linked to the machined object, for example by assigning a reference number that is used together with the machined object. The machining knowledge is also linked, for example, to a combination of the final machining shape represented by the CAD data 100 and the material of the blank. Thereby, the decision making unit 46 can search for the machining knowledge corresponding to the CAD data 100 and specified by the NC machining program generating unit 43, and further select the machining knowledge with the highest action value Q among these machining knowledge using, as the search condition, the reference number or Combination of information about the final machining shape and information about the material of the blank is used.

The machine learning device 50, which includes the state observation unit 51 and the learning unit 52, can use any learning algorithm. An example of an application case for reinforcement learning is described below.

Reinforcement learning is carried out in such a way that an agent, which is supposed to perform actions in a certain environment, observes the current state and determines which action to perform. The agent learns how to obtain the greatest reward through a series of actions, where selecting an action results in obtaining a reward in the environment. The typical well-known reinforcement learning algorithms include Q-Learning and TD-Learning. For example, in Q-learning, a typical update equation of the action-value function Q(s, a) can be expressed by the following mathematical formula (1).
[Formula 1] $$Q\left(s_t,a_t\right)\leftarrow Q\left(s_t,a_t\right)+\alpha \left(r_{t+1}+\gamma \underset{a}{{\displaystyle \text{Max}}}Q\left(s_{t+1},a_{t+1}\right)-Q\left(s_t,a_t\right)\right)$$

In the mathematical formula (1), s _t means the state at time t; at is the action performed at time t; s _t+1 means the state at time t+1 or, in other words, the state resulting from the execution of the action at; and r _t+1 means the reward given depending on the environment that has changed as a result of the action at, that is, r _t+1 means the reward given by the change in state. In the mathematical formula (1), the factor “max” represents the maximum value of the action value Q in the state s _t+1 , that is, the action value related to the best action at _t+1 . The parameter γ is the discount factor, where it is set so that it fulfills the condition 0<γ≤ 1. For example, the parameter γ is set in the range from 0.9 to 0.99 (limits included). The parameter α is the learning rate, where it is set to satisfy the condition 0<α≤ 1. For example, the parameter α is set in the range from 0.05 to 0.2 (limits included).

The update equation, which is expressed by the mathematical formula (1) in Q-learning, increases the action value Q at time t if the action value of the best action a _t+1 at time t+1 is greater than the action value Q of the one at time t+1 t performed action at, but otherwise reduces the action value Q at time t. In other words, the action value function Q(s _t , a _t ) is updated to bring the action value Q of action a _t at time t closer to the best action value at time t+1. In this way, the best action value in a given environment is sequentially transferred to the action values in previous environments.

The machine learning device 50 according to the first embodiment is therefore capable of performing reinforcement learning according to a previously known Q-learning algorithm. The action value Q assigned to a combination of a particular state variable and an action “a” is unknown. The machine learning device 50 randomly selects actions “a” for different state variables, executes the actions “a,” and accumulates the rewards awarded as a result of the actions “a,” thereby forming the action value function Q(s, a). to update. Note that in the NC machining program generating apparatus 40 according to the first embodiment, an “action a” corresponds to machining knowledge, that is, information indicating the method of generating an NC machining program. In addition, a “state s” corresponds to the editing of the first NC machining program 111.

Note that the action value function Q(s, a) is also called an action value table. 2 shows a representation to illustrate the principle of an action value table. An action value table is a data set stored in tabular form in which an arbitrary combination of an action "a" and a state "s" is associated with the action value Q. In Q-learning, a data set represents a so-called Q table. A Q table in the form of a two-dimensional table includes various states “s” in rows and possible actions “a” in columns, with an action value Q entered in each cell , which represents the evaluation value for the combination of action “a” and state “s”. However, due to the limitation of the number of states "s" and actions "a" that can be stored when managing the action value function Q(s, a) using a Q table, the action values Q are calculated using the above mathematical formula (1) and not calculated using a table format when there are many states "s" and many actions "a".

In the first embodiment, when a Q-table format is adopted, a Q-table is stored for each set of editing shape data, each Q-table containing editing knowledge corresponding to the action "a", which is information that provides a method for Specify generation of an NC machining program, and the editing of the first NC machining program 111 corresponding to the state “s”, both of which are linked to the machining form data. That is, in the first embodiment, a data set is a set of data of machining shape data, data of machining knowledge which is information indicating a method of generating an NC machining program, and data of editing the first NC machining program 111.

The numerical controller 1 has machine learning functionality to learn the method of generating an NC machining program. That is, the numerical controller 1 includes the machine learning device 50 that learns the method of generating an NC machining program. The machine learning device 50 includes the state observation unit 51 and the learning unit 52. The learning unit 52 learns the method for generating an NC machining program based on the data set generated based on state variables. Accordingly, the learning unit 52 learns a relationship between the machining shape data, the method of generating an NC machining program and the editing of the first NC machining program 111 according to the data set generated based on state variables, thereby learning the method of generating an NC machining program. The learning unit 52 includes a reward calculation unit 53 that calculates the reward in connection with learning the machining knowledge corresponding to the NC machining program generation method, and a function updating unit 54 that calculates the function in connection with learning the machining knowledge corresponding to the NC machining program generation method updated and determines the action value Q, which is the evaluation value. Note that the machine learning device 50 may be integrated into the NC machining program generating device 40 or installed in a cloud server that can communicate with the NC machining program generating device 40.

The state observation unit 51 observes the CAD data 100 as state variables, the CAD data 100 being the machining shape data, the machining knowledge, which is information indicating the method for generating a numerically controlled machining program and which is referred to by the NC Machining program generation unit 43 was referred to during the generation of the first NC machining program 111, and the editing of the first NC machining program 111. That is, for each set of the CAD data 100, which is the machining shape data, the state observation unit 51 observes that from the decision making unit 46 certain machining knowledge and the editing of the first NC machining program 111, which is used in NC machining processing program editing unit 45 was carried out during the generation of the second NC machining program 112.

The state observation unit 51 incorporates the editing of the first NC machining program 111 into the machining knowledge referred to in the generation of the first NC machining program 111, and thus generates editing-integrated machining knowledge. An editing-integrated machining knowledge is information that specifies a method for generating an editing-integrated program for numerically controlled machining, which results from the integration of the editing of the first NC machining program 111 into the method for generating a numerically controlled machining program, which is accessed during the Generation of the first NC machining program 111 was referenced. The editing-integrated machining knowledge takes over the reward “r” of the machining knowledge linked to the first NC machining program 111. The state observation unit 51 transmits the machining knowledge referred to during the generation of the first NC machining program 111 and the editing-integrated machining knowledge as an observation result to the reward calculation unit 53.

When the state observation unit 51 generates an editing-integrated machining knowledge, the reference number associated with the original machining knowledge or the combination of the machining final shape information and the blank material information is adopted and assigned. The editing-integrated machining knowledge, which is generated based on the machining knowledge, and the machining knowledge both correspond to the final machining shape of a machined object. The editing-integrated machining knowledge therefore corresponds to the final machining form of a machined object linked to the machining knowledge from which the editing-integrated machining knowledge was generated. Therefore, the reference number associated with the machining knowledge from which the editing-integrated machining knowledge was generated, or the combination of the information about the final machining shape and the information about the material of the blank is transmitted and assigned to the editing-integrated machining knowledge. This means that the link with the CAD data 100 that is assigned to it is transferred to the editing-integrated processing knowledge.

When the CAD data 100, the first NC machining program 111, the machining knowledge associated with the first NC machining program 111 and the second NC machining program 112 are transferred from the NC machining program editing unit 45, the state observation unit 51 can edit the first Extract NC machining program 111. Thus, in this case, the editing-integrated machining knowledge becomes the new machining knowledge that results from the inclusion of the editing of the first NC machining program 111 in conjunction with the CAD data 100 in the machining knowledge referred to when generating the first NC machining program 111 was referred to.

On the other hand, if only the CAD data 100, the first NC machining program 111 and the machining knowledge associated with the first NC machining program 111 are transmitted from the NC machining program editing unit 45 and thus the second NC machining program 112 is not transmitted, there is no editing of the first NC machining program 111. In this case, the editing-integrated machining knowledge is the machining knowledge itself that was referred to when generating the first NC machining program 111.

Accordingly, observing the CAD data 100, which is the machining shape data, means the machining knowledge, which is information indicating the method of generating a program for numerically controlled machining and to which the NC machining program -Generation unit 43 referred to during the generation of the first NC machining program 111, and the editing of the first NC machining program 111 as state variables: the editing of the first NC machining program 111 in conjunction with the CAD data 100 is extracted to as Observation result to obtain the machining knowledge referred to during the generation of the first NC machining program 111 and the editing-integrated machining knowledge for each set of the CAD data 100.

The learning unit 52 learns the relationship between the machining shape data, the method of generating an NC machining program, and the editing of the first NC machining program 111 based on the data set of the CAD data 100, which is the machining shape data, of the machining knowledge which is information that determines the process for generating it a program for numerically controlled machining and to which the NC machining program generating unit 43 has referred during the generation of the first NC machining program 111, and the editing of the first NC machining program 111, thereby learning the method for NC machining program generation . This means that the learning unit 52 learns the machining knowledge based on the state variables observed by the state observation unit 51. In one embodiment, the learning unit 52 learns the editing knowledge according to the principle of reinforcement learning.

The reward calculation unit 53 calculates the reward r based on the editing-integrated editing knowledge, which is the observation result transmitted from the state observation unit 51. Specifically, the reward calculation unit 53 calculates the reward r of the editing-integrated machining knowledge based on the machining knowledge referred to in generating the first NC machining program 111 and the edit-integrated machining knowledge. The reward calculation unit 53 calculates the reward r of the edit-integrated machining knowledge based on the degree of equality between the machining knowledge referred to during the generation of the first NC machining program 111 and the edit-integrated machining knowledge, with the degree of equality as a reward allocation criterion for the allocation the reward r serves. The reward calculation unit 53 associates the reward r with the CAD data 100 and the edit-integrated machining knowledge, and then transmits the calculated reward r together with the CAD data 100 and the edit-integrated machining knowledge to the function update unit 54.

If the machining knowledge referred to in generating the first NC machining program 111 is identical to the editing-integrated machining knowledge, the reward calculation unit 53 increases the reward r of the editing-integrated machining knowledge. If the machining knowledge referred to when generating the first NC machining program 111 is identical to the editing-integrated machining knowledge, the reward calculation unit 53 assigns the editing-integrated machining knowledge, for example, a “positive” reward r, such as. B. a reward r of “+1”. The reward r of the editing-integrated machining knowledge is therefore calculated by adding “+1” to the reward r adopted during generation from the machining knowledge referred to when generating the first NC machining program 111. Thus, the reward r for the editing-integrated machining knowledge becomes greater than the reward r acquired from the machining knowledge referred to during the generation of the first NC machining program 111.

If the machining knowledge referred to in generating the first NC machining program 111 is different from the editing-integrated machining knowledge, the reward calculation unit 53 reduces the reward r of the machining knowledge referred to in generating the first NC machining program 111 . When the machining knowledge referred to in the generation of the first NC machining program 111 is different from the editing-integrated machining knowledge, the reward calculation unit 53 assigns the machining knowledge referred to in the generation of the first NC machining program 111, for example “negative” reward r, such as B. a reward r of “-1”. For example, when a first NC machining program 111 generated so that facing is performed before drilling is edited so that drilling is performed before facing, the reward calculation unit 53 reduces the reward r and assigns to the machining knowledge referenced during the Generation of the first NC machining program 111 was referenced, for example a reward r of “-1”. The reward r of the machining knowledge that was referred to when generating the first NC machining program 111 is therefore calculated by adding “-1” to the current reward r. The reward r of the machining knowledge that was referred to during the generation of the first NC machining program 111 is therefore reduced to a value that is below the reward r at the reference time during the generation of the first NC machining program 111.

In addition, as described above, when the machining knowledge referred to in the generation of the first NC machining program 111 is different from the editing-integrated machining knowledge, the reward calculation unit 53 reduces the reward r of the machining knowledge to that in the generation of the first NC machining program 111 was referred to, and then increases the reward r of the editing-integrated editing knowledge. If the machining knowledge referred to when generating the first NC machining program 111 menu was distinguished from the editing-integrated editing knowledge, the reward calculation unit 53 assigns, for example, the editing-integrated editing knowledge a “positive” reward r, such as a reward r of “+1”. The reward r of the editing-integrated machining knowledge is therefore calculated by adding “+1” to the reward r that was adopted during the generation of the machining knowledge that was referred to when generating the first NC machining program 111. The reward r for the editing-integrated machining knowledge is therefore greater than the reward r that was inherited from the machining knowledge that was referenced during the generation of the first NC machining program 111.

The numerical value of the reward r is not limited to “+1” or “-1”. In addition, the reward r can vary depending on the editing of the first NC machining program 111, i.e. H. A different value can be assigned to the editing to generate the editing-integrated machining knowledge from the machining knowledge to which reference was made when generating the first NC machining program 111.

The function update unit 54 updates the function for determining the method of NC machining program generation based on the reward r of the editing-integrated machining knowledge calculated by the reward calculation unit 53. For example, the function update unit 54 updates the action value function Q(s, a) and determines the action value Q, which is the evaluation value. The function update unit 54 links the determined action value Q with the editing-integrated editing knowledge and stores the action value Q together with the editing-integrated editing knowledge. Thus, in addition to the initially stored machining knowledge, the edit-integrated machining knowledge and the action value Q associated with the edit-integrated machining knowledge are sequentially stored in the function update unit 54 as the machining knowledge to be referred to in generating the first NC machining program 111 . 3 12 is a diagram illustrating an example of editing knowledge and editing-integrated editing knowledge stored in the function update unit 54 of the machine learning device 50 according to the first embodiment of the present invention. 3 illustrates an example of storing an editing knowledge A 121, an editing knowledge B 122, an editing knowledge C 123, an editing-integrated editing knowledge A 131, an editing-integrated editing knowledge B 132 and an editing-integrated editing knowledge C 133. It is noted that an editing knowledge storage unit is provided separately can, which stores the editing knowledge, the editing-integrated editing knowledge and the action values Q that are linked to the editing-integrated editing knowledge.

Next, a reinforcement learning method for updating the action value function Q(s, a) is described. 4 Fig. 12 is a flowchart illustrating a procedure for updating the action value function Q(s, a) according to the first embodiment of the present invention.

At step ST110, the NC machining program generating unit 43 reads the CAD data 100 stored in the machining shape data storage unit 42 and stores the CAD data 100 in a shape storage area of a storage unit (not shown) of the NC machining program generating unit 43 .

Next, in step ST120, the NC machining program generating unit 43 obtains the machining knowledge corresponding to the CAD data 100 via the decision making unit 46 from the function updating unit 54 to generate an NC machining program. The decision making unit 46 determines the machining knowledge based on the learning result of the machine learning device 50. That is, the decision making unit 46 searches a plurality of machining knowledge stored in the function update unit 54, the machining knowledge with the highest action value Q from the plurality of machining knowledge related to the query the NC machining program generation unit 43 relates, determines and relates and transfers the related machining knowledge to the NC machining program generation unit 43.

The following is an example of the machining knowledge transferred to the NC machining program generating unit 43.

(Processing knowledge A 121)

Blank material: S45C

Processing order: The drilling-turning step is carried out first, and then the facing step is carried out when the machining allowance for the end surface is 20mm or more. The facing step is carried out first and then the drill turning step if the blank surface allowance does not meet this condition.

(Processing knowledge B 122)

Blank material: S45C

Part to be machined: The end face is determined as the part to be machined if “lateral length of the shape of the open area/longitudinal length of the shape of the open area” is “less than 1”. The outer circumference is determined as the part to be machined when “lateral length of the open area shape/longitudinal length of the open area shape” is “1 or greater”. The term “open area” refers to the part of the turning cross-sectional shape t that is exposed on the surface of the blank.

(Processing knowledge C 123)

Blank material: S45C

Processing method: A groove turning step is carried out when the longitudinal length of the turning cross-sectional shape is 10 mm or less and the lateral length of the turning cross-sectional shape is 10 mm or less. A bar stock turning processing step is performed when either the longitudinal length of the turning cross-sectional shape or the lateral length does not meet these conditions.

Next, the NC machining program generating unit 43 generates an NC machining program, i.e., in step ST130. H. the first NC machining program 111, with reference to the related machining knowledge.

Next, in step ST140, the NC machining program generating unit 43 associates the CAD data 100, the machining knowledge referred to in generating the first NC machining program 111, and the first NC machining program 111 and transfers the CAD -Data 100, the first NC machining program 111 and the machining knowledge to the NC machining program storage unit 44. The NC machining program storage unit 44 then stores the CAD data 100, the first NC machining program 111 and the machining knowledge.

Next, in step ST150, the NC machining program editing unit 45 reads out the CAD data 100, the first NC machining program 111 and the machining knowledge associated with the first NC machining program 111 from the NC machining program storage unit 44 and transfers the first NC -Machining program 111 to the interactive operation processing unit 10. The interactive operation processing unit 10 then displays the first NC machining program 111 transmitted from the NC machining program editing unit 45 on the display unit 20 and prompts the user to do so first NC machining program 111 to confirm.

When the first NC machining program 111 is displayed on the display unit 20, the user confirms the first NC machining program 111 and inputs the editing instruction information for the first NC machining program 111 into the instruction input unit 30. The editing instruction information is transmitted from the instruction input unit 30 to the NC machining program editing unit 45 of the NC machining program generating device 40 via the interactive operation processing unit 10.

Next, in step ST160, the NC machining program editing unit 45 receives the editing instruction information and edits the first NC machining program 111 based on the editing instruction information to generate the second NC machining program 112. The second NC machining program 112 is an NC machining program generated by editing the first NC machining program 111 based on the editing instruction information. The NC machining program editing unit 45 then transfers the CAD data 100, the first NC machining program 111, the machining knowledge linked to the first NC machining program 111 and the second NC machining program 112 to the status observation unit 51.

Next, in step ST170, the state observation unit 51 compares the first NC machining program 111 and the second NC machining program 112 transmitted from the NC machining program editing unit 45, and obtains the editing of the first NC machining program 111. In this way, the state observation unit 51 as state variables, the CAD data 100, which is the machining shape data, the machining knowledge linked to the first NC machining program 111, and the editing of the first NC machining program 111.

Next, in step ST180, the state observation unit 51 incorporates the editing of the first NC machining program 111 into the machining knowledge referred to in generating the first NC machining program 111, thereby generating the editing-integrated machining knowledge. The state observation unit 51 transmits the machining knowledge linked to the first NC machining program 111 and the editing-integrated machining knowledge as the observation result to the reward calculation unit 53.

Next, in step ST190, the reward calculation unit 53 compares the machining knowledge associated with the first NC machining program 111 and the machining-integrated machining knowledge to determine whether they are identical. If these are identical to each other, i.e. h in step ST 190 Yes, the procedure proceeds to step ST200. If these are different from each other, i.e. H. in step ST190 No, the procedure continues with step ST210.

In step ST200, the reward calculation unit 53 calculates the reward r that increases the editing-integrated editing knowledge reward r to cause an increase in the action value Q for the editing-integrated editing knowledge reward r, and adds the calculated reward r to the editing-integrated editing knowledge reward r, which was adopted from the machining knowledge linked to the first NC machining program 111. This increases the reward r of the editing-integrated machining knowledge to be greater than the reward r acquired from the machining knowledge associated with the first NC machining program 111. This increases the action value Q of the editing-integrated editing knowledge.

If the machining knowledge associated with the first NC machining program 111 is identical to the editing-integrated machining knowledge, the machining knowledge associated with the first NC machining program 111 itself is the editing-integrated machining knowledge. Since no editing of the first NC machining program 111 was carried out, it is assumed that the editing-integrated machining knowledge is high-quality editing-integrated machining knowledge. The machining knowledge associated with the first NC machining program 111 should therefore be high-quality machining knowledge.

For example, if the edit-integrated machining knowledge is as follows, the machining knowledge associated with the first NC machining program 111 are all identical to the corresponding edit-integrated machining knowledge, and accordingly the rewards r of the edit-integrated machining knowledge are increased in step ST200. The editing-integrated editing knowledge A 121a is an editing-integrated editing knowledge that is generated on the basis of the editing knowledge A 121. The editing-integrated editing knowledge B 121b is an editing-integrated editing knowledge that is generated on the basis of the editing knowledge B 122. The editing-integrated editing knowledge C 121c is an editing-integrated editing knowledge that is generated based on the editing knowledge C 123.

(Editing-integrated editing knowledge A 121a)

Blank material: S45C

Processing sequence: The drilling turning step is carried out first, and then the facing step is carried out when the blank area allowance is 20 mm or more. The facing step is carried out first and then the drill turning step if the blank surface allowance does not meet this condition.

(Editing-integrated editing knowledge B 121b)

Blank material: S45C

Part to be machined: The end face is determined as the part to be machined if “lateral length of the shape of the open area/longitudinal length of the shape of the open area” is “less than 1”. The outer circumference is determined as the part to be machined when “lateral length of the open area shape/longitudinal length of the open area shape” is “1 or greater”. The term “open area” refers to the area of a turning cross-sectional shape t that is exposed on the surface of the blank.

(Editing-integrated editing knowledge C 121c)

Blank material: S45C

Processing method: A groove turning step is carried out when the longitudinal length of the turning cross-sectional shape is 10 mm or less and the lateral length of the turning cross-sectional shape is 10 mm or less. A bar material turning processing step is performed when either the longitudinal length or the lateral length of the turning cross-sectional shape does not meet these conditions.

Meanwhile, in step ST210, the reward calculation unit 53 calculates the reward r that reduces the machining knowledge reward r to cause a reduction in the action value Q for the machining knowledge reward r, and adds the calculated reward r to the machining knowledge reward r. This reduces the reward r of the machining knowledge to be smaller than the reward r at the reference time during the generation of the first NC machining program 111. This reduces the action value Q of the machining knowledge.

For example, if the editing-integrated machining knowledges are as follows, the machining knowledges associated with the first NC machining program 111 are all different from the corresponding editing-integrated machining knowledges, and accordingly the rewards r of the machining knowledges are reduced in step ST210. The editing-integrated editing knowledge A 131 is an editing-integrated editing knowledge that is generated on the basis of the editing knowledge A 121. The editing-integrated editing knowledge B 132 is an editing-integrated editing knowledge that is generated on the basis of the editing knowledge B 122. The editing-integrated editing knowledge C 133 is an editing-integrated editing knowledge that is generated on the basis of the editing knowledge C 123.

(Editing-integrated editing knowledge A 131)

Blank material: S45C

Processing sequence: The drilling turning step is carried out first, and then the facing step is carried out when the blank area allowance is 10 mm or more. The facing step is carried out first and then the drilling turning step if the blank surface allowance does not meet this condition.

(Editing-integrated editing knowledge B 132)

Blank material: S45C

Part to be machined: The end face is determined as the part to be machined when “lateral length of the open area shape/longitudinal length of the open area shape” is “1.0 or less”. The outer circumference is determined as the part to be machined if “lateral length of the open area shape/longitudinal length of the open area shape” is “greater than 1.0”.

(Editing-integrated editing knowledge C 133)

Blank material: S45C

Processing method: A groove turning step is carried out when the longitudinal length of the turning cross-sectional shape is 10mm or less and the lateral length of the turning cross-sectional shape is 20mm or less. A bar material turning processing step is performed when either the longitudinal length or the lateral length of the turning cross-sectional shape does not meet these conditions.

If the machining knowledge associated with the first NC machining program 111 is different from the editing-integrated machining knowledge, the editing of the first NC machining program 111 is carried out. Therefore, it is assumed that the machining knowledge associated with the first NC machining program 111 is of lower quality than the editing-integrated machining knowledge.

In contrast, it is assumed that the editing-integrated machining knowledge has a higher quality than the machining knowledge linked to the first NC machining program 111. Therefore, after step ST210, in step ST200, the reward calculation unit 53 calculates the reward r that increases the editing-integrated editing knowledge reward r to cause an increase in the action value Q for the editing-integrated editing knowledge reward r, and adds the calculated reward r to the reward r of the editing-integrated machining knowledge, which was adopted from the machining knowledge linked to the first NC machining program 111. This increases the reward r for the editing-integrated machining knowledge to a value that is greater than the reward r adopted by the machining knowledge linked to the first NC machining program 111. This increases the action value Q of the editing-integrated editing knowledge.

Next, in step ST220, the function update unit 54 updates the action value function Q(s, a) based on the reward r of the edit-integrated editing knowledge calculated in step ST200 or based on the reward r of the edit-integrated editing knowledge calculated in step ST200 and the reward r of the machining knowledge calculated in step ST210 and determines the action values Q, which are the evaluation values of the machining knowledge linked to the first NC machining program 111 and the editing-integrated machining knowledge. The function update unit 54 assigns the action value Q determined for the editing-integrated editing knowledge to the editing-integrated editing knowledge and stores the action value Q as editing knowledge together with the editing-integrated editing knowledge.

The function update unit 54 further associates the action value Q determined for the machining knowledge associated with the first NC machining program 111 with this machining knowledge, and stores the action value Q as machining knowledge together with the machining knowledge. In this context, the function update unit 54 therefore stores the machining knowledge linked to the first NC machining program 111, the machining knowledge whose action value Q was determined in step ST220, and the editing-integrated machining knowledge whose action value Q was determined in step ST220, each of which has a different action value Q, as the machining knowledge corresponding to the CAD data 100. Thereby, when determining the machining knowledge corresponding to the CAD data 100, the decision-making unit 46 can determine the editing-integrated machining knowledge, which is machining knowledge with a relatively high action value Q, as the most appropriate machining knowledge that adequately reflects the user's knowledge and experience.

Note that the machining knowledge stored in the function update unit 54 and referred to during the generation of the first NC machining program 111 can be updated with the machining knowledge associated with the first NC machining program 111 in this case, from which the action value Q was determined in step ST220.

The procedure from step ST110 to step ST220 is then carried out iteratively in conjunction with a further set of newly read CAD data 100. When generating new editing-integrated machining knowledge, associated information, such as the reference number of the original machining knowledge, the information about the final machining shape and the information about the material of the blank as well as the reward r, is adopted.

Note that when iterating the procedure from step ST110 to step ST220, the editing-integrated editing knowledge determined by the decision making unit 46 corresponds to the editing knowledge of step ST120 described above.

In addition, in the function update unit 54, high-quality editing-integrated editing knowledge with an increased action value Q is stored in the procedure of steps ST200 and step ST220. Accordingly, step ST210 may be omitted.

Next will be with reference to 5 a detail of the NC machining program generation method in the NC machining program generation unit 43 is described in detail. 5 Fig. 12 is a flowchart illustrating a detail of the procedure of the NC machining program generating method in the NC machining program generating unit 43 of the NC machining program generating apparatus 40 according to the first embodiment of the present invention. 6 12 shows a perspective view illustrating an example of a three-dimensional model of the final machining shape represented by the CAD data 100 according to the first embodiment of the present invention. 7 Fig. 12 is a perspective view illustrating an example of a blank mold encasing the final machining shape represented by the CAD data 100 according to the first embodiment of the present invention. 8th Fig. 12 is a perspective view illustrating an example of a cutting shape generated from the final machining shape represented by the CAD data 100 and the blank shape according to the first embodiment of the present invention. 9 Fig. 12 is a diagram illustrating a blank shape S, a turning cross-sectional shape produced from the machining shape, and coordinate values according to the first embodiment of the present invention. The following description is given in the context of generating an NC machining program for turning in which a blank is machined while rotating.

In step ST310 of the NC machining program generation process, the NC machining program generation unit 43 reads the CAD data 100 stored in the machining shape data storage unit 42, generates a three-dimensional machining final shape represented by the in 6 shown CAD data 100 is represented, and positions the final machining form in the program origin, which is the machining origin in the program coordinate system. The program coordinate system is defined so that the Z axis of the XYZ axes is aligned with a rotation axis SG, which is the central axis during rotation.

In order to position the machining final shape in the program origin, the NC machining program generating unit 43 positions the machining final shape by moving and rotating the machining final shape about the rotation axis of the cylindrical surface or the rotation axis of the conical surface having the largest diameter in the machining final shape with the Z -align axis; and moves the machining finish mold to position its end surface in the Z-axis direction at the program origin. This means that the NC machining program generating unit 43 extracts, as a rotating end surface, a cylindrical surface or a conical surface whose central rotation axis is aligned with the rotation axis SG from the CAD data 100. In this way, the NC machining program generating unit 43 generates from the CAD data 100 a three-dimensional final machining shape having a central rotation axis, which is the rotation axis.

Next, in step ST320, the NC machining program generating unit 43 generates a three-dimensional cylindrical shape enveloping the final machining shape positioned at the program origin, as shown in 7 is shown. The NC machining program generating unit 43 then positions the generated cylindrical shape as a blank shape in the program coordinate system similarly to the case of the final machining shape. The term “blank shape” refers to the shape of the blank that is to be turned into the final machining shape. This means that the NC machining program generating unit 43 generates a three-dimensional blank shape based on the final machining shape generated from the CAD data 100.

The dimensions of the cylindrical shape enveloping the final machining shape can be determined from the maximum value and the minimum value of the final machining shape in the X-axis direction, Y-axis direction and Z-axis direction. It is noted that with regard to the end face turning machining, the dimensions of the blank mold in the Z-axis direction in accordance with the final machining shape are each larger by 2 to 3 mm than the corresponding dimension of the final machining shape. In addition, the end surface of the blank mold is positioned in the Z-axis direction to protrude from the final machining mold by 2 to 3 mm from the program origin. It is assumed here that the blank mold has a blank outer diameter of 150mm and a blank length of 120mm, and the blank material is S45C.

Next, the NC machining program generating unit 43 generates as in step ST330 8th shown a cutting shape, ie the shape that is to be cut by turning from the blank shape, from the final machining shape and from the blank shape. 8th shows the final machining form in the form of dashed lines. The cutting shape can be obtained by subtraction, where the volume model of the final machining shape is subtracted from the volume model of the blank shape.

Next, in step ST340, the NC machining program generating unit 43 generates a turning cross-sectional shape t, which is the cross-sectional shape of the cutting shape as shown in 9 is shown. As in 9 shown, the rotational cross-sectional shape t is a half-rotational cross-sectional shape obtained by projecting the sectional shape onto the +XZ plane, which is part of the XZ plane limited to the +X region. The rotational cross-sectional shape t can be obtained by obtaining the cross-sectional shape through a multiplication operation of a cross-sectional model formed from the XZ plane and a solid model formed from the machining shape, and limiting this shape to the range of X≥0. In this way, a rotational cross-sectional shape t1 and a rotational cross-sectional shape t2 are generated, which are the rotational cross-sectional shape t, as in 9 is shown.

Next, in step ST350, the NC machining program generating unit 43 obtains the machining knowledge corresponding to the CAD data 100 via the decision making unit 46 from the function updating unit 54 and refers to this machining knowledge. The NC machining program generating unit 43 obtains, for example, the above-described machining knowledge A 121, the machining knowledge B 122 and the machining knowledge C 123 stored in the function updating unit 54 and refers to these machining knowledge.

For example, the NC machining program generating unit 43 obtains the reference number commonly assigned to the CAD data 100 and the machining knowledge based on the CAD data 100, and transmits the reference number to the decision making unit 46. Thereby, the decision making unit 46 can use the reference number as a search condition, search for the desired machining knowledge specified by the NC machining program generating unit 43.

As used herein, “facing step” refers to a step of performing end face turning machining in which the protruding portion on the end face of the blank is removed using an end face turning machining tool. The term “drilling-turning step” refers to a step of performing a drilling-turning operation in which a rotary drill is used to drill along the central axis into a radial central portion of the blank. The term “bar material turning processing step” refers to a step for performing bar turning machining in which an outer circumference, an inner circumference, an end face or a back side of a round bar is turned with a turning tool. The term “grooving step” refers to a step for performing grooving by turning in which an outer circumference, an inner circumference, an end face or a back side of a round bar is profiled using a turning tool. The term “blank surface allowance” refers to the machining allowance in end face turning machining.

Next, in step ST360, the NC machining program generating unit 43 divides the turning cross-sectional shape t of the cutting shape into parts corresponding to the obtained machining knowledge A 121, the machining knowledge B 122 and the machining knowledge C 123 in several different machining steps to generate turning shapes. The term “turning shape” refers to the turning cross-sectional shape t in each machining step when a blank is machined into the final machining shape in several different machining steps.

A method for dividing the rotational cross-sectional shape t is specifically described below. 10 1 is a diagram illustrating a turning cross-sectional shape SH1 of a facing step according to the first embodiment of the present invention. 11 Fig. 12 is a diagram illustrating a turning cross-sectional shape SH2 of a drilling turning step according to the first embodiment of the present invention. 12 Fig. 12 is a diagram illustrating a turning cross-sectional shape SH3 of a bar material turning processing step according to the first embodiment of the present invention. 13 Fig. 12 is a diagram illustrating a turning cross-sectional shape SH4 of a bar material turning processing step according to the first embodiment of the present invention.

The NC machining program generating unit 43 obtains information about the blank area allowance: 10 mm from the turning cross-sectional shape t1. The NC machining program generating unit 43 then refers to the machining knowledge A 121, determines that the facing step is to be performed first because the allowance is not 20 mm or larger, and separates as shown in 10 is shown, the turning cross-sectional shape SH1 for the facing step depends on the turning cross-sectional shape t1.

The NC machining program generating unit 43 then refers to the machining knowledge A 121, determines that the drilling-turning step is to be performed after the facing step, and separates, as in 11 is shown, the turning cross-sectional shape SH2 for the drilling turning step depends on the turning cross-sectional shape t1.

Next, the NC machining program generating unit 43 separates as in 12 is shown, the turning cross-sectional shape SH3 for the bar material turning processing step depends on the turning cross-sectional shape t1. The NC machining program generation unit 43 then separates, as in 13 is shown, the turning cross-sectional shape t2 as turning cross-sectional shape SH4 for the bar material turning processing step.

Next, in step ST370, the NC machining program generating unit 43 assigns turning machining steps to the divided machining shapes, that is, the turning cross-sectional shapes, based on the machining knowledge A 121, the machining knowledge B 122 and the machining knowledge C 123, thereby generating the entire turning machining steps. 14 Fig. 12 is a diagram illustrating an example of the machining start point and the machining end point in the facing step according to the first embodiment of the present invention. 15 Fig. 12 is a diagram illustrating an example of the machining start point and the machining end point in the drilling rotation step according to the first embodiment of the present invention. 16 Fig. 12 is a diagram illustrating an example of the cutting point, the machining start point and the machining end point in the bar material turning machining step according to the first embodiment of the present invention. 17 Fig. 12 is a diagram illustrating an example of the cutting point, the machining start point and the machining end point in the bar material turning machining step according to the first embodiment of the present invention.

The NC machining program generation unit 43 receives, as in 14 is shown, the coordinate values of a machining start point P1 and the coordinate values of a machining end point P2 of the facing step from the turning cross-sectional shape SH1 for the facing step and generates a facing step LC1, which is the facing step.

The NC machining program generation unit 43 then receives, as in 15 is shown, the bore diameter of the bore to be removed during drilling turning, the coordinate values of a machining start point P3 and the coordinate values of a machining end point P4 from the turning cross-sectional shape SH2 for the drilling turning step and generates a drilling turning step LC2, which is the drilling turning step.

Next, the NC machining program generating unit 43 receives as in 16 As shown, information about the longitudinal length of the open area shape: 30 mm and information about the lateral length of the open area shape: 30 mm from the turning cross-sectional shape SH3 for the bar material turning processing step. The NC machining program generating unit 43 then refers to the machining knowledge B 122 and determines that the part to be machined in the bar material turning machining step is the outer circumference because “lateral length of the open area shape/longitudinal length of the open area shape” “1 “0”. Next, the NC machining program generating unit 43 obtains from the turning cross-sectional shape SH3 for the bar material turning machining step the coordinate values of a cutting point P5, the coordinate values of a machining start point P6 and the coordinate values of a machining end point P7 of the turning cross-sectional shape SH3, and generates a bar outer peripheral turning step LC3, which is a bar material turning machining step in which The outer circumference was determined as the part to be machined.

Next, the NC machining program generating unit 43 receives as in 17 is shown, information about the longitudinal length of the turning cross-sectional shape SH4: 10 mm and information about the lateral length of the turning cross-sectional shape SH4: 20 mm from the turning cross-sectional shape SH4 for the bar material turning processing step. The NC machining program generating unit 43 then refers to the machining knowledge C 123 and determines that the bar material turning machining step is to be performed because the longitudinal length of the turning cross-sectional shape SH4 is 10 mm or less and the lateral length of the turning cross-sectional shape SH4 is not 10 mm or less.

Next, from the turning cross-sectional shape SH4 for the bar material turning processing step, the NC machining program generating unit 43 obtains information about the longitudinal length of the open area shape: 10 mm and information about the lateral length of the open area shape: 20 mm. The NC machining program generating unit 43 then refers to the machining knowledge B 122 and determines that the part to be machined is the outer circumference because “lateral length of the shape of the open area/longitudinal length of open area shape” is “2.0” and therefore is not “less than 1.0”. Next, the NC machining program generating unit 43 obtains from the turning cross-sectional shape SH4 for the bar material turning machining step the coordinate values of a cutting point P8, a machining start point P9 and a machining end point P10 of the turning cross-sectional shape SH4, and generates a bar outer peripheral turning step LC4, which is a bar material turning machining step, as a part to be machined the outer circumference was determined.

• Plandrehschritt LC1: Bearbeitungsstartpunkt P1 (75, -10) und Bearbeitungsendpunkt P2 (0, 0)
• Bohrdrehschritt LC2: Bohrungsdurchmesser 30 mm, Bearbeitungsstartpunkt P3 (0, 0) und Bearbeitungsendpunkt P4 (0, 110)
• Stabaußenumfangsdrehschritt LC3: Schneidpunkt P5 (75, 0), Bearbeitungsstartpunkt P6 (45, 0) und Bearbeitungsendpunkt P7 (45, 30)
• Stabaußenumfangsdrehschritt LC4: Schneidpunkt P8 (75, 60), Bearbeitungsstartpunkt P9 (65, 60) und Bearbeitungsendpunkt P10 (65, 80).

Details of the turning processing steps generated in the NC machining program generating unit 43 will be described below.

• Facing step LC1: machining start point P1 (75, -10) and machining end point P2 (0, 0)
• Drilling turning step LC2: hole diameter 30 mm, machining start point P3 (0, 0) and machining end point P4 (0, 110)
• Bar outer circumference turning step LC3: cutting point P5 (75, 0), machining start point P6 (45, 0) and machining end point P7 (45, 30)
• Bar outer circumference turning step LC4: cutting point P8 (75, 60), machining start point P9 (65, 60) and machining end point P10 (65, 80).

In the method described above, the NC machining program generating unit 43 refers to the machining knowledge A 121, the machining knowledge B 122 and the machining knowledge C 123, and can automatically generate the plurality of different turning steps for machining a blank into the final machining shape. In addition, the automatic generation of the turning machining steps, with reference to the machining knowledge that has integrated and accumulated the editing that reflects the user's knowledge and experience, enables the automatic generation of the turning machining steps desired by the user, corresponding to the turning machining steps in the case of generating an NC -Editing program by the user themselves are similar.

Further, the above-mentioned learning action performed by the machine learning device 50 according to the first embodiment is continuously performed every time an NC machining program is generated. Accordingly, the NC machining program generating device 40 according to the first embodiment continuously performs autonomous learning of machining knowledge indicating the method of generating an NC machining program according to the CAD data 100, the information about the final machining shape of the machined object, and information about include the material of the blank. That is, the NC machining program generating device 40, which includes the machine learning device 50, continuously and autonomously generates machining knowledge that provides high machining efficiency and reflects the user's knowledge and experience based on the knowledge in the NC machining program editing unit 45 carried out editing of the NC machining program generated by the NC machining program generating unit 43.

Next, a detail of a method for generating editing-integrated editing knowledge will be described. In this method, the state observation unit 51 incorporates the editing of the first NC machining program 111 into the machining knowledge referred to in the generation of the first NC machining program 111 to generate editing-integrated machining knowledge. First, a case will be described in which the first NC machining program 111 was edited to execute the facing step LC1 according to the editing instruction information input by the user after the drilling turning step LC2. The condition observation unit 51 obtains information about the blank area allowance: 10 mm from the turning cross-sectional shape SH1 for the facing step LC1 of the second NC machining program 112 transmitted from the NC machining program editing unit 45.

In this case, the second NC machining program 112 was generated by editing the first NC machining program 111 to perform the drilling turning step at the blank area allowance criterion: 10 mm first, unlike the case of the machining knowledge criterion A 121. Accordingly, the state observation unit 51 involves the editing of the first NC machining program 111 and generates an editing-integrated machining knowledge A 131 resulting from the change of the machining knowledge A 121 according to the following criteria.

(Editing-integrated editing knowledge A 131)

Blank material: S45C

Processing sequence: The drilling turning step is carried out first and then the facing step is carried out when the blank area allowance is “10mm or larger”. The facing step and then the drilling turning step are carried out if the blank surface allowance does not meet this condition.

Next, a case will be described in which the first NC machining program 111 was edited according to the editing instruction information inputted by the user to change the part to be machined in the bar outer circumference turning step LC3 from outer circumference to end face and thus change the bar outer circumference turning step LC3 into a bar end face turning step LC31. The state observation unit 51 obtains information about the longitudinal length of the open area shape: 30 mm and information about the lateral length of the open area shape: 30 mm from the turning cross-sectional shape SH3 for the rod outer circumference turning step LC3 of the second NC machining program 112 that is from the NC -Editing program editing unit 45 is transmitted.

In this case, the second NC machining program 112 was generated by editing the first NC machining program 111 to produce the part to be machined, unlike the case of the machining knowledge criterion B 122 in the "lateral length of the shape of the open area/longitudinal length of the" criterion Shape of the open area" to "1.0" in front side. Accordingly, the state observation unit 51 involves the editing of the first NC machining program 111 and generates an editing-integrated machining knowledge B 132 obtained when the machining knowledge B 122 is changed to include the following criteria.

(Editing-integrated editing knowledge B 132)

Blank material: S45C

Part to be machined: The part to be machined is determined as the end face when “lateral length of the open area shape/longitudinal length of the open area shape” is “equal to 1.0 or less”. The part to be machined is determined as the outer perimeter when “lateral length of the open area shape/longitudinal length of the open area shape” is “greater than 1.0”.

Next, a case in which the first NC machining program 111 was edited to change the bar outer circumference turning step LC4 to a groove turning step LC41 according to the editing instruction information input by the user will be described. The state observation unit 51 obtains information about the longitudinal length of the turning cross-sectional shape SH4: 10 mm and information about the lateral length of the turning cross-sectional shape SH4: 20 mm from the turning cross-sectional shape SH4 for the rod outer circumference turning step LC4 of the second NC machining program 112 prepared by the NC machining program editing unit 45 was transferred.

In this case, the second NC machining program 112 was generated by editing the first NC machining program 111 to produce, in contrast to the case of the machining knowledge criterion C 123, in a criterion “the longitudinal length of the turning cross-sectional shape SH4 is 10 mm and the lateral length of the turning cross-sectional shape SH4 is 20 mm", to carry out the groove turning step LC41. Accordingly, the state observation unit 51 includes the editing of the first NC machining program 111 and generates an editing-integrated machining knowledge C 133 resulting from the change of the machining knowledge C 123 according to the following criteria.

(Editing-integrated editing knowledge C 133)

Blank material: S45C

Processing method: The groove turning step is carried out when the longitudinal length of the turning cross-sectional shape is “10mm or less” and the lateral length of the turning cross-sectional shape is “20mm or less”. The bar material turning processing step is carried out when either the longitudinal length or the lateral length of the turning cross-sectional shape does not satisfy these conditions.

A method for dividing the turning cross-sectional shape t will be specifically described below for a case in which the NC machining program generating unit 43 refers to the editing-integrated machining knowledge A 131, the editing-integrated machining knowledge B 132 and the editing-integrated machining knowledge C 133 and an NC machining program generated. 18 Fig. 12 is a diagram illustrating a turning cross-sectional shape SH5 of the drilling turning step according to the first embodiment of the present invention. 19 Fig. 12 is a diagram illustrating a turning cross-sectional shape SH6 of the facing step according to the first embodiment of the present invention. 20 Fig. 12 is a diagram illustrating a turning cross-sectional shape SH7 of the bar material turning processing step according to the first embodiment of the present invention. 21 Fig. 12 is a diagram illustrating a turning cross-sectional shape SH8 of the bar material turning processing step according to the first embodiment of the present invention.

The NC machining program generating unit 43 obtains information about the length of the blank: 120 mm and information about the outer diameter of the blank: 150 mm from the blank mold and information about the blank area allowance: 10 mm from the turning cross-section mold t1. The NC machining program generation unit 43 then refers to the editing-integrated machining knowledge A 131, determines that the drilling turning step is to be performed first because the allowance is 10 mm or larger, and separates as shown in 18 is shown, the turning cross-sectional shape SH5 for the drilling turning step depends on the turning cross-sectional shape t1.

Next, the NC machining program generating unit 43 refers to the editing-integrated machining knowledge A 131, determines that the facing step is to be performed after the drilling-turning step, and separates as shown in 19 is shown, the turning cross-sectional shape SH6 for the facing step depends on the turning cross-sectional shape t1.

Next, the NC machining program generating unit 43 separates as in 20 is shown, the turning cross-sectional shape SH7 for the bar material turning processing step depends on the turning cross-sectional shape t1. The NC machining program generation unit 43 then separates, as in 21 is shown, the turning cross-sectional shape t2 as turning cross-sectional shape SH8 for the bar material turning processing step.

A method for assigning the turning machining steps is described below specifically for the case that the NC machining program generating unit 43 refers to the editing-integrated machining knowledge A 131, the editing-integrated machining knowledge B 132 and the editing-integrated machining knowledge C 133 and generates an NC machining program. 22 Fig. 12 is a diagram illustrating an example of the machining start point and the machining end point in the drilling rotation step according to the first embodiment of the present invention. 23 Fig. 12 is a diagram illustrating an example of the machining start point and the machining end point in the facing step according to the first embodiment of the present invention. 24 Fig. 12 is a diagram illustrating an example of the cutting point, the machining start point and the machining end point in the bar end turning step according to the first embodiment of the present invention. 25 Fig. 10 is a diagram illustrating an example of the cutting point, the machining start point and the machining end point in the bar grooving turning step according to the first embodiment of the present invention.

The NC machining program generation unit 43 receives, as in 22 is shown, the hole diameter to be excluded during drilling turning, the coordinate values of a machining start point P11 and the coordinate values of a machining end point P12 from the rotary cross-sectional shape SH5 for the rotary drilling step and generates a rotary drilling step LC11, which is the rotary drilling step.

The NC machining program generation unit 43 then receives, as in 23 is shown, from the turning cross-sectional shape SH6 for the facing step, the coordinate values for a machining start point P13 and the coordinate values for a machining end point P14 of the facing step and generates a facing step LC12, which is the facing step.

Next, the NC machining program generating unit 43 receives as in 24 is shown, from the turning cross-sectional shape SH7 for the bar material turning processing step, information about the longitudinal length of the open area shape: 30 mm and information about the lateral length of the open area shape: 30 mm. The NC machining program generating unit 43 refers then refer to the editing-integrated machining knowledge B 132 and determine that the part to be machined in the bar material turning machining step is the end face because “lateral length of the shape of the open area/longitudinal length of the shape of the open area” is “1.0”, i.e. “1, 0 or less”. Next, the NC machining program generating unit 43 obtains from the turning cross-sectional shape SH7 for the bar material turning machining step the coordinate values of a cutting point P15, the coordinate values of a machining start point P16 and the coordinate values of a machining end point P17 of the turning cross-sectional shape SH7, and generates a bar end face turning step LC31, which is a bar material turning machining step in which The end face was determined as the part to be machined.

Next, the NC machining program generating unit 43 receives as in 25 is shown, information about the longitudinal length of the turning cross-sectional shape SH8: 10 mm and information about the lateral length of the turning cross-sectional shape SH8: 20 mm from the turning cross-sectional shape SH8 for the bar material turning processing step. The NC machining program generating unit 43 then refers to the editing-integrated machining knowledge C 133 and determines that the groove turning step is to be performed because the longitudinal length of the turning cross-sectional shape SH4 is 10 mm or less and the lateral length of the turning cross-sectional shape SH4 is 20 mm or less.

Next, from the turning cross-sectional shape SH8 for the bar material turning processing step, the NC machining program generating unit 43 obtains information about the longitudinal length of the open area shape: 10 mm and information about the lateral length of the open area shape: 20 mm. The NC machining program generating unit 43 obtains the information about the groove depth: 10 mm from the information about the longitudinal length of the shape of the open area: 10 mm and the information about the groove width: 20 mm from the information about the lateral length of the shape of the open area: 20 mm. The NC machining program generating unit 43 then obtains the coordinate values of a machining start point P18 from the turning cross-sectional shape SH8 for the bar material turning machining step and generates a groove turning step LC14.

• • Bohrdrehschritt LC11: Bohrungsdurchmesser 30 mm, Bearbeitungsstartpunkt P11 (0, -10) und Bearbeitungsendpunkt P12 (0, 110)
• • Plandrehschritt LC12: Bearbeitungsstartpunkt P13 (75, -10) und Bearbeitungsendpunkt P14 (15, 0)
• • Stabstirnflächendrehschritt LC13: Schneidpunkt P15 (75, 0), Bearbeitungsstartpunkt P16 (75, 30) und Bearbeitungsendpunkt P17 (45, 30)
• • Nutdrehschritt LC14: Nutbreite 20 mm, Nuttiefe 10 mm und Bearbeitungsstartpunkt P18 (75, 60).

The turning processing steps generated in the NC machining program generating unit 43 will be described in detail below with reference to the editing-integrated machining knowledge A 131, the editing-integrated machining knowledge B 132 and the editing-integrated machining knowledge C 133.

• • Drilling and turning step LC11: hole diameter 30 mm, machining start point P11 (0, -10) and machining end point P12 (0, 110)
• • Facing step LC12: machining start point P13 (75, -10) and machining end point P14 (15, 0)
• • Bar end face turning step LC13: cutting point P15 (75, 0), machining start point P16 (75, 30) and machining end point P17 (45, 30)
• • Groove turning step LC14: Groove width 20 mm, groove depth 10 mm and machining start point P18 (75, 60).

In the method described above, the NC machining program generating unit 43 refers to the editing-integrated machining knowledge A131, the editing-integrated machining knowledge B132 and the editing-integrated machining knowledge C133, and can thus automatically generate the plurality of different turning steps for machining a blank into the final machining shape. Further, the automatic generation of turning machining steps with reference to the editing-integrated machining knowledge that has integrated and accumulated the editing reflecting the user's knowledge and experience enables the automatic generation of the turning machining steps desired by the user corresponding to the turning machining steps in the case of generating a NC machining program by the user themselves are similar.

The decision making unit 46 determines the machining knowledge or the editing-integrated machining knowledge for the request of the NC machining program generating unit 43 using the learning results of the machine learning device 50. The NC machining program generating unit 43 may be required to edit the NC machining program after the NC -Processing program was created. Since the machine learning device 50 according to the present embodiment has already learned the editing-integrated machining knowledge, the decision-making unit 46 can, if necessary, determine the editing-integrated machining knowledge most suitable for the request of the NC machining program generating unit 43 from the machining knowledge stored in the function updating unit 54.

The machine learning device 50 is also capable of learning a relationship between the material of the blank, the CAD data 100 and machining knowledge. If it is intended that the NC machining program is to be edited with regard to the material of the blank and the CAD data 100, the numerical control can be designed such that the decision-making unit 46 determines to select another machining knowledge or editing-integrated machining knowledge.

It should be noted that editing-integrated editing knowledge can be stored such that the CAD data 100 itself is linked to it. For example, the CAD data 100 is linked to the editing-integrated machining knowledge A131 and stored in the function update unit 54 together with the editing-integrated machining knowledge A131.

In addition, editing-integrated machining knowledge can also be stored in such a way that information relating to the CAD data 100 is linked to the editing-integrated machining knowledge. For example, information related to the CAD data 100 is linked to the editing-integrated machining knowledge A131 and stored together with the editing-integrated machining knowledge A131. Examples of information related to the CAD data 100 include an image of the CAD data and an element of the CAD data. It is noted that “information relating to the CAD data 100” is, in other words, “information relating to a combination of information about the final machining shape of the machined object and information about the material of the blank”.

Thereby, the decision-making unit 46 can search and obtain a desired editing-integrated editing knowledge among a plurality of editing knowledge stored in the function updating unit 54 using the CAD data 100 or information related to the CAD data 100 as a search condition.

Further, a different set of machining knowledge and editing-integrated machining knowledge may be stored for each of the NC machine tools controlled by executing the NC machining program. When multiple NC machine tools are used, the NC machine tools often have different functions from each other, for example, in view of a situation in which different types of NC machine tools are used, a situation in which NC machine tools of the same type are used, but too different optional functions are added to different NC machine tools, or the like. In these cases, there are appropriate criteria for machining knowledge on a per NC machine tool basis.

Accordingly, by storing a different set of machining knowledge and editing-integrated machining knowledge for each of the NC machine tools, the editing that reflects the user's knowledge and experience and is suitable for each of the NC machine tools is integrated and accumulated into an editing-integrated machining knowledge. In this way, the NC machining program generating unit 43 generates an NC machining program with reference to machining knowledge or editing-integrated machining knowledge managed on a per NC machine tool basis. This enables the rapid automatic generation of an NC machining program desired by the user and suitable for any NC machine tool, which is similar to an NC machining program generated by the user himself.

In addition, the numerical control can be designed such that the machining knowledge and the editing-integrated machining knowledge are stored in a storage unit provided in a network server.

As a variant of the machine learning device 50 according to the first embodiment, the machine learning device 50 may perform machine learning according to a neural network model. 26 Fig. 12 is a schematic diagram illustrating an example of a neural network model followed by the machine learning device 50 according to the first embodiment of the present invention. A neural network is constructed, for example, from an input layer that includes an input neuron, an intermediate layer that includes a number of m neurons, and an output layer that includes a number of n output neurons. 26 illustrates an example of a neural network model constructed from an input layer with four input neurons, an intermediate layer with five neurons, and an output layer with one output neuron. It is noted that 26 only one layer is illustrated as an intermediate layer, but two or more intermediate layers may be present.

A neural network learns the relationship between a state variable and an environmental change, e.g. B. when an optimal action is learned intuitively based on a person's successful experience or an unsuccessful experience. A neural network learns the relationship between a state variable and an environmental change based on a data set generated based on a state variable or variables observed by the state observation unit 51. This means that a neural network learns the relationship between a state variable and an environmental change through so-called supervised learning. In this case, the machine learning device 50 includes the output layer to output, in response to the machining shape data input to the input layer of the neural network and the NC machining program generation method, machining knowledge that corresponds to the machining shape data and better reflects the user's knowledge and experience. Then, the decision making unit 46 is operated to determine an optimal NC machining program generation method. Therefore, even in the case of using a neural network model, the machine learning device 50 can learn the relationship between the machining shape data, the NC machining program generation method and the editing of the first NC machining program 111, thereby learning the method of generating an NC machining program.

Note that although an embodiment in which the machine learning device 50 performs machine learning using reinforcement learning or using a neural network has been described above, the machine learning device 50 may also perform machine learning according to another known method, such as for example genetic programming, inductive logic programming or a support vector machine.

Further, according to an embodiment of learning the learning unit 52, the learning unit 52 may learn machining knowledge to be applied to an NC machining program used in another numerical controller. That is, the learning unit 52 may be configured to acquire machining knowledge to be applied to an NC machining program used in another numerical control with respect to specific machining shape data, based on the machining knowledge used in generating the NC machining program should be used, which was created for another numerical control, and learned on the basis of editing the NC machining program. The learning unit 52 may learn machining knowledge to be applied to an NC machining program used in another numerical control operated at the same location, or it may learn machining knowledge to be applied to an NC machining program that is used in another numerical control that operates independently at a different location.

27 12 is a diagram illustrating a hardware configuration of the NC machining program generating apparatus 40 according to the first embodiment of the present invention. In the 27 Hardware shown includes a processor 201 that performs arithmetic operations, a memory 202 used by the processor 201 as a work area, a storage device 203 that stores a program for operation as a numerical controller or a program conversion device, an input device 204 that provides an input interface for represents a user, a display device 205 which displays information for the user, and a communication device 206 whose function is to communicate with a device to be controlled or with another numerical control and/or with various other devices. The processor 201, the memory 202, the storage device 203, the input device 204, the display device 205 and the communication device 206 are connected to one another via a data bus 207. The processor 201 can be a processing unit, a computing unit, a microprocessor, a microcomputer, a central processing unit (CPU), a digital signal processor (DSP) or similar. Further, the memory 202 is a non-volatile or volatile semiconductor memory, such as a random access memory (RAM), a read-only memory (ROM), a flash memory, an erasable programmable ROM (EPROM), or an electrically erasable programmable ROM (EEPROM) (registered trademark). ; a magnetic disk, a floppy disk, an optical disk, a compact disc, a minidisc, a digital versatile disc (DVD), or the like.

The NC machining program generating unit 43 is, for example, executed by executing a program described in FIG 27 shown memory 202 is stored, implemented by the processor 201. The above functionality can be implemented by multiple processors in conjunction with multiple memories. Furthermore, the functionality of the NC machining program generation unit 43 can be partially implemented as an electronic circuit, with the remaining functionality being implemented by the processor 201 and the memory 202.

Further, the NC machining program generating device 40 may be configured so that the machining shape data input unit 41, the NC machining program editing unit 45, the decision making unit 46, the state observation unit 51, the reward calculation unit 53 and the function update unit 54 are similarly executed by executing an im Memory 202 stored program can be implemented by the processor 201. In addition, the above functionality can be implemented by multiple processors in conjunction with multiple memories. Further, the functionality of the machining shape data input unit 41, the NC machining program editing unit 45, the decision making unit 46, the state observation unit 51, the reward calculation unit 53 and the function update unit 54 can be partially implemented as an electronic circuit, with the remaining functionality being provided by the processor 201 and the Memory 202 can be implemented. Furthermore, the processor and memory for implementing the functionality of the machining shape data input unit 41, the NC machining program editing unit 45, the decision making unit 46, the condition observation unit 51, the reward calculation unit 53 and the function update unit 54 may be the same processor and the same memory act or another processor and another memory different from the processor and memory for implementing the NC machining program generating unit 43.

The NC machining program generating device 40 according to the first embodiment described above refers to machining knowledge that has integrated and accumulated the editing reflecting the user's knowledge and experience, and automatically generates turning machining steps based on the CAD data 100 which are the processing form data. This enables the automatic generation of the turning machining steps desired by the user, which are similar to the turning machining steps in the case of the creation of an NC machining program by the user himself.

In addition, according to the first embodiment, the NC machining program generating device 40, which includes the machine learning device 50, independently learns machining knowledge corresponding to the method for generating an NC machining program and the CAD data 100, the information about the final machining shape of the machined object and information about the material of the blank. That is, the NC machining program generating device 40 comprising the machine learning device 50 independently learns machining knowledge that provides high machining efficiency and reflects the user's knowledge and experience based on the editing of the NC machining program performed in the NC machining program editing unit 45. Machining program generation unit 43 generated NC machining program. Thereby, based on the learning result, the decision making unit 46 can determine the machining knowledge that corresponds to the CAD data 100, has the highest action value Q and offers the highest machining efficiency from the machining knowledge stored in the function update unit 54 when the CAD data 100 is re-inputted NC machining program generating device 40 can be entered.

Accordingly, the NC machining program generating device 40 according to the first embodiment enables automatic generation of a high-quality NC machining program that provides high machining efficiency and performs turning machining steps close to the turning machining steps desired by the user without the need to accumulate the knowledge and skills of an expert . Since the user eliminates the need to generate an NC machining program on a trial-and-error basis, the work efficiency in generating an NC machining program is improved. Therefore, the NC machining program generating device 40 according to the first embodiment is effective in reducing the time required for automatically generating an NC machining program and in reducing the machining time for a workpiece.

The NC machining program generating device 40 according to the first embodiment further learns the editing of the NC machining program performed by the user and incorporates the editing into the machining knowledge, thereby increasing the extent of the editing operation of the NC machining program by the user on a next occasion of the NC Program generation can be reduced.

Reference symbol list

1

numerical control;

10

processing unit for interactive operation;

20

display unit;

30

instruction input unit;

40

NC machining program generating device;

41

edit shape data input unit;

42

edit shape data storage unit;

43

NC machining program generation unit;

44

NC machining program storage unit;

45

NC machining program editing unit;

46

decision making unit;

50

machine learning device;

51

condition observation unit;

52

learning unit;

53

reward calculation unit;

54

Function update unit;

100

CAD data;

111

first NC machining program;

112

second NC machining program;

121

Machining knowledge A;

122

Machining knowledge B;

123

Machining knowledge C;

131

editing-integrated editing knowledge A;

132

editing-integrated editing knowledge B;

133

editing-integrated editing knowledge C;

201

Processor;

202

Storage;

203

storage device;

204

input device;

205

display device;

206

communication device;

207

Data bus.

Claims (3)

Machine learning device (50), comprising: a state observation unit (51) which is designed to process shape data as state variables, which includes information about a final machining shape of a machined object and information about a material of a blank, and a method for generating a program for a numerically controlled machining for automatically generating a program for numerically controlled machining comprising a plurality of machining operations for machining a workpiece into the machined object; and a learning unit (52) which is designed to learn the method for generating a program for numerically controlled processing based on a data set generated on the basis of the state variables, wherein the learning unit (52) comprises: a reward calculation unit (53) configured to calculate a reward; and a function update unit (54) configured to update a function for determining the method for generating a program for numerically controlled machining based on the calculated reward; characterized in that the state observation unit (51) is further configured to observe, as state variables, an editing of a first program for numerically controlled machining which was generated with reference to the method for generating a program for numerically controlled machining; and the reward calculation unit (53) is designed to calculate the reward based on a method for generating an editing-integrated program for numerically controlled machining, which results from the inclusion of the editing in the method for generating the program for numerically controlled machining. Machine learning device (50). Claim 1 , wherein the reward calculation unit (53) is designed to increase a reward of the method for generating an editing-integrated program for numerically controlled machining if the method for generating a program for numerically controlled machining is identical to the method for generating an editing-integrated program for numerically controlled machining controlled machining, and to reduce a reward of the method of generating a program for numerically controlled machining and to increase the reward of the method of generating an editing-integrated program for numerically controlled machining when the method of generating a program for numerically controlled machining Editing differs from the method for generating an editing-integrated program for numerically controlled editing. Device for generating a program for numerically controlled machining, comprising: the machine learning device (50). Claim 1 or 2 ; a numerically controlled machining program generating unit (43) for generating the first numerically controlled machining program with reference to the method for generating a numerically controlled machining program based on a result of learning of the machine learning device (50) is determined; a numerical control machining program editing unit (45) for editing the first numerical control machining program to produce a second numerical control machining program; and a decision making unit (46) for determining a method of generating a numerically controlled machining program corresponding to the machining shape data based on the result of learning of the learning unit (52) of the machine learning device (50).

Images