Thermal displacement compensation systems, numerical control, methods, and learning model set

Abstract

System (1) for compensating for a thermal displacement, which performs a compensation of a thermal displacement of a machine, comprising the system (1) for compensating for a thermal displacement: a condition specification section (110) that specifies a condition of an individual difference of the machine; a state quantity detection section (140) that detects a state quantity which indicates a state of the machine; a derivation calculation section (220) that derives a compensation measure of a thermal displacement of the machine from the state quantity; a compensation execution section (400) that performs a compensation of a thermal displacement of the machine based on the compensation measure of a thermal displacement of the machine, which is derived by the derivation calculation section (220); a learning model generation section (500) that generates or updates a learning model through machine learning that uses the state quantity; and a learning model storage section (300) that stores at least one learning model generated by the learning model generation section (500) in conjunction with a combination of conditions specified by the condition specification section (110), wherein The derivation calculation section (220) derives the compensation measure of a thermal displacement of the machine by selectively using, on the basis of a condition of an individual difference of the machine, which is specified by the condition specification section (110), at least one learning model from the learning models which are stored in the learning model storage section (300).

Description

[BACKGROUND OF THE INVENTION] 1. Field of the invention

[0001] The present invention relates to a system for compensating for a thermal displacement and in particular a system for compensating for a thermal displacement which performs a compensation while learning models are switched according to an individual difference of a machine in a factory. 2. Description of the state of the art

[0002] Since a feed screw and a spindle are driven by a motor in a machine, the feed screw and spindle expand, and the machine's position changes due to heat generated by the motor, frictional heat from a rotating bearing, and frictional heat in the contact area between a ball screw and a ball screw nut of the feed screw. Furthermore, changes in the machine's ambient temperature and the use of coolant also cause changes in column and bed temperatures, resulting in a change in the machine's position due to extension or tilt caused by such temperature changes. In other words, a deviation occurs in the relative positional relationship between a workpiece being positioned and a tool.Such a change in machine position due to heat becomes a problem when machining is performed with high accuracy.

[0003] To eliminate such heat-induced displacement in a machine position, a technique for compensating a command position using a displacement sensor, a technique for compensating a command position by predicting a thermal displacement of operating conditions, such as the number of revolutions of a spindle, and a structure that circumvents the effect of thermal expansion by applying an initial stress to a feed screw were used (see Japanese Patent Application JP 2002 - 086 329 A, Japanese Patent Application JP 2006 - 055 919A, and Japanese Patent Application JP 2008 - 183 653 A).

[0004] DE 10 2016 011 532 A1 describes a machine learning device and a machine learning method for optimizing a point in time at which a tool in a machine tool needs to be corrected.

[0005] However, a machine's thermal displacement state changes significantly not only according to the machine's operating condition but also according to individual machine characteristics. For example, a machine's thermal displacement state can change depending on the production batch, frequency of use, cumulative operating time, maintenance / modification history, and similar factors, even if the operating condition remains the same. Therefore, it is essential to appropriately account for individual machine characteristics when performing thermal displacement compensation.

[0006] While a machine learning device that takes into account the operating state and individual differences of a machine can certainly be introduced to compensate for thermal displacement of the machine, creating a general-purpose machine learning device (a general-purpose learning model) capable of accommodating a range of situations as described above requires a large volume of state information detected in different situations, and additionally, known problems such as overlearning can arise due to the need for many parameters containing data relating to such diverse situations. [BRIEF DESCRIPTION OF THE INVENTION]

[0007] In light of this, an object of the present invention is to provide systems for compensating for thermal displacement that enable the performance of appropriate thermal displacement compensation, taking into account the individual differences of a machine. A further object of the invention is to provide various methods, a numerical control system, and a learning model set for compensating for thermal displacement. This object is achieved by a system according to claim 1 or 5, a method according to claim 8 or 10, a numerical control system according to claim 7, and a learning model set according to claim 12.

[0008] The thermal displacement compensation system according to the present invention solves the problem described above by providing a mechanism for switching learning models used to determine thermal displacement compensation for a machine in a factory, based on an individual difference of the machine. The thermal displacement compensation system according to the present invention has a plurality of learning models, selects a learning model based on an individual difference or the like of a machine in a factory, performs machine learning based on a state quantity detected at the machine with respect to the selected learning model, and performs thermal displacement compensation for the machine in a factory by selectively using learning models created in this way, based on an individual difference or the like of the machine.

[0009] A system for compensating for a thermal displacement according to one aspect of the present invention performs a compensation of a thermal displacement of a machine, the system for compensating for a thermal displacement comprising: a condition specification section that specifies a condition of an individual difference of the machine; a state quantity detection section that detects a state quantity that indicates a state of the machine; a derivation calculation section that derives a compensation measure of a thermal displacement of the machine from the state quantity; a compensation execution section that performs a compensation of a thermal displacement of the machine based on the compensation measure of a thermal displacement of the machine derived by the derivation calculation section;a learning model generation section that generates or updates a learning model through machine learning, which uses the state quantity; and a learning model storage section that stores at least one learning model generated by the learning model generation section in association with a combination of conditions specified by the condition specification section. Additionally, the derivation computation section derives the compensation measure of a thermal displacement of the machine by selectively using, based on a condition of an individual difference of the machine specified by the condition specification section, at least one learning model from the learning models stored in the learning model storage section.

[0010] The system for compensating for a thermal displacement may further include a feature quantity creation section that creates a feature quantity characterizing a thermal state of the machine from the state quantity detected by the state quantity detection section, wherein the derivation calculation section can derive the compensation measure of a thermal displacement of the machine from the feature quantity, and the learning model generation section can generate or update a learning model by machine learning that uses the feature quantity.

[0011] The learning model generation section can generate a new learning model by modifying an existing learning model stored in the learning model memory section.

[0012] The learning model storage section can encrypt and store a learning model generated by the learning model generation section and decrypt the encrypted learning model when a learning model is read by the derivative computation section.

[0013] A system for compensating for a thermal displacement according to another aspect of the present invention performs a compensation of a thermal displacement of a machine, the system for compensating for a thermal displacement comprising: a condition specification section that specifies a condition of an individual difference of the machine; a state quantity detection section that detects a state quantity that indicates a state of the machine; a derivation calculation section that derives a compensation measure of a thermal displacement of the machine from the state quantity; a compensation execution section that performs a compensation of a thermal displacement of the machine based on the compensation measure of a thermal displacement of the machine derived by the derivation calculation section;and a learning model memory section that stores at least one learning model that has been pre-linked to a combination of machine operations. Additionally, the derivation calculation section derives the compensation measure for a thermal displacement of the machine by selectively using, based on a condition of an individual difference of the machine specified by the condition specification section, at least one learning model from the learning models stored in the learning model memory section.

[0014] The system for compensating for a thermal displacement may further include a feature quantity creation section that creates a feature quantity characterizing a thermal state of the machine from the state quantity detected by the state quantity detection section, and the derivation calculation section can derive the compensation measure of a thermal displacement of the machine from the feature quantity.

[0015] A numerical control according to one aspect of the present invention includes the condition specification section and the state quantity detection section described above.

[0016] A method for compensating for a thermal displacement according to one aspect of the present invention comprises the steps of: specifying a condition of individual difference of a machine; detecting a state quantity that indicates a state of the machine; deriving a compensation measure of a thermal displacement of the machine from the state quantity; performing a compensation of a thermal displacement of the machine based on the compensation measure of a thermal displacement; and generating or updating a machine learning model that uses the state quantity.Additionally, in the derivation step, a learning model to be used based on the condition specified in the condition specification step can be selected from at least one learning model that is linked in advance to a combination of operating conditions of the machine, and the compensation measure of a thermal displacement of the machine can be derived using the selected learning model.

[0017] The procedure for compensating for a thermal displacement may further include a step to create a feature quantity that characterizes a thermal state of the machine from the state quantity, wherein in the derivation step the compensation measure of a thermal displacement of the machine can be derived from the feature quantity, and in the step to generate or update a learning model a learning model by machine learning that uses the feature quantity can be generated or updated.

[0018] A method for compensating for a thermal displacement according to another aspect of the present invention comprises the steps of: specifying a condition of individual difference of a machine; detecting a state quantity that indicates a state of the machine; deriving a compensating measure of thermal displacement of the machine from the state quantity; and performing a compensating measure of thermal displacement of the machine based on the compensating measure of thermal displacement. Additionally, in the derivation step, a learning model to be used based on the condition specified in the condition specification step is selected from at least one learning model that is pre-associated with a combination of conditions of individual difference of the machine, and the compensating measure of thermal displacement of the machine is derived using the selected learning model.

[0019] The procedure for compensating for a thermal displacement may further include a step to create a feature quantity that characterizes a thermal state due to an individual difference of the machine from the state quantity, wherein in the derivation step the compensation measure of a thermal displacement of the machine can be derived from the feature quantity.

[0020] A learning model set configured according to one aspect of the present invention is a learning model set obtained by linking each of a plurality of learning models with a combination of conditions under which compensation for a thermal displacement of a machine is to be performed, wherein each of the plurality of learning models is a learning model that is generated or updated based on a state quantity that specifies a state of the machine under a condition of an individual difference of the machine. Additionally, a learning model is selected from the plurality of learning models based on the condition of an individual difference of the machine, and the selected learning model is used in a process to derive a compensation measure for a thermal displacement of the machine.

[0021] Since, according to the present invention, machine learning can be performed based on a state quantity of a machine in a factory, which is detected in every situation with respect to a learning model selected according to an individual difference of the machine, highly efficient machine learning can be performed while preventing excessive learning, and since compensation for a thermal displacement of a machine is performed using a learning model selected according to an individual difference or the like of the machine, the accuracy of compensation for a thermal displacement of the machine is improved. [BRIEF DESCRIPTION OF THE DRAWINGS] Fig. Figure 1 is a schematic functional block diagram of a system for compensating for a thermal displacement according to a first embodiment; Fig. Figure 2 is a schematic representation showing the conditions of an individual difference of a machine; Fig. Figure 3 is a schematic representation showing a classification of conditions of an individual difference of a machine; Fig. Figure 4 is a schematic functional block diagram of a system for compensating for a thermal displacement according to a second embodiment; Fig. Figure 5 is a schematic functional block diagram of a system for compensating for a thermal displacement according to a third embodiment; Fig. Figure 6 is a schematic functional block diagram of a system for compensating for a thermal displacement according to a fourth embodiment; Fig. Figure 7 is a schematic functional block diagram of a system for compensating for a thermal displacement according to a fifth embodiment; Fig. Figure 8 is a schematic functional block diagram of a modification of the system for compensating for a thermal displacement according to the fifth embodiment; Fig. Figure 9 is a schematic functional block diagram of a system for compensating for a thermal displacement according to a sixth embodiment; Fig. 10 is a schematic flow diagram of a process carried out on a system for compensating for a thermal displacement according to the present invention; Fig. Figure 11 is a schematic flow diagram of a process carried out on a system for compensating for a thermal displacement according to the present invention; and Fig. Figure 12 is a schematic hardware configuration diagram showing main parts of a numerical control and a machine learning apparatus according to an embodiment of the present invention. [DETAILED DESCRIPTION OF PREFERRED EXECUTION FORMS]

[0022] Fig. Figure 1 is a schematic functional block diagram of a system for compensating a thermal displacement 1 according to a first embodiment.

[0023] The corresponding ones, in Fig. The functional blocks shown in Figure 1 are implemented as a processor such as a CPU or a GPU, which is provided in a computer such as a numerical control system, a cell computer, a host computer or a cloud server, which controls the operation of different parts of an apparatus according to each system program.

[0024] System 1 for compensating a thermal displacement according to the present embodiment comprises a numerical control section 100 as an edge device, a derivation processing section 200, and a learning model storage section 300. The numerical control section 100 serves as at least one observation / derivation object of a state. It performs a derivation with respect to a state of the edge device. The learning model storage section 300 stores and manages a plurality of learning models. System 1 for compensating a thermal displacement according to the present embodiment further comprises a compensation execution section 400 and a learning model generation section 500. The compensation execution section 400 performs a compensation of a thermal displacement based on the result of a derivation by the derivation processing section 200 with respect to the state of the edge device.The learning model generation section 500 creates and updates a learning model 300, which is stored in the learning model storage section.

[0025] The numerical control section 100 according to the present embodiment controls a machine by executing blocks of a work program stored in a memory (not shown). For example, the numerical control section 100 is implemented as a numerical controller, reads and analyzes blocks of the work program stored in the memory (not shown), calculates a movement range of a motor 120 per control period based on the result of the analysis, and controls the motor 120 according to the calculated movement range per control period. The machine controlled by the numerical control section 100 includes a mechanical section 130 driven by the motor 120, and when the mechanical section 130 is driven, for example, a tool and a workpiece move relative to each other, and the workpiece is machined. Furthermore, while in Fig. If 1 is omitted, the number of motors provided is 120, which corresponds to the number of axes provided in the mechanical section 130 of a machine tool.

[0026] A condition specification section 110, provided in the numerical control section 100, specifies a condition of individual difference of the machine controlled by the numerical control section 100. Examples of a condition of individual difference include a production lot, a frequency of use, a cumulative operating time, and a maintenance / modification history of the machine.The condition specification section 110 specifies, if necessary, a condition set, a condition obtained from the machine by the numerical control section 100, a condition obtained by the numerical control section 100 via a network, a condition recommended by the work program, and the like, with respect to each section of the numerical control section 100 for the numerical control section 100 (outputs it to this) via an input device (not shown) by a worker, and simultaneously specifies the conditions for the learning model storage section 300 and the learning model generation section 500 (outputs these).The condition specification section 110 has a task to inform each section of the system 1 for compensating a thermal displacement about a condition in a current machining operation of the numerical control section 100 as a boundary device as a condition for selecting a learning model.

[0027] For example, in Fig. As shown in Figure 2, a condition of individual difference of the machine controlled by the numerical control section 100 can be specified by a combination of points, including a production lot, a frequency of use, a cumulative operating time, a maintenance history (which can be represented by a set of elapsed time since a previous maintenance and a maintenance frequency for each part of the machine being maintained), and a modification history (which can be represented by a customer-ordered component being introduced in relation to a modified part of the machine). Additionally, any point can be included besides those described above, as long as the point specifies an individual difference that significantly affects a thermal displacement.

[0028] A state quantity detection section 140, provided in the numerical control section 100, detects a state of the machine controlled by the numerical control section 100 as a state quantity. Examples of a state quantity of the machine controlled by the numerical control section 100 include the ambient temperature of the machine, the temperature of any point in the machine, and the operation of a spindle, feed axis, or the like of the machine. For example, the state quantity detection section 140 detects as a state quantity the value of a current flowing through the numerical control section 100 or the motor 120, driving the mechanical section 130 of a machine tool controlled by the numerical control section 100, or a detection value detected by a device such as a sensor, which is provided separately in each section.The state quantity detected by the state quantity detection section 140 is output to the derivation processing section 200 and the learning model generation section 500.

[0029] The derivation processing section 200 according to the present embodiment observes a state of the numerical control section 100 as an edge device (and of the machine controlled by the numerical control section 100) and derives a compensation measure of a thermal displacement of each axis of the machine controlled by the numerical control section 100 based on the result of the observation. For example, the derivation processing section 200 can be implemented as a numerical controller, a cell computer, a host computer, a cloud server, a machine learning device, or the like.

[0030] A feature quantity creation section 210, provided in the derivation processing section 200, creates a feature quantity that specifies properties of a thermal state of the machine controlled by the numerical control section 100, based on the state quantity detected by the state quantity detection section 140. The feature quantity created by the feature quantity creation section 210, which specifies properties of the thermal state of the machine controlled by the numerical control section 100, is information that is useful as a criterion for deriving a thermal displacement magnitude of the machine controlled by the numerical control section 100.Additionally, the feature quantity created by the feature quantity creation section 210, which specifies properties of the thermal state of the machine controlled by the numerical control section 100, is used as input data when a derivation calculation section 220 (which will be described later) performs a derivation using a learning model.The feature quantity created by the feature quantity creation section 210, which specifies properties of the thermal state of the machine controlled by the numerical control section 100, can be, for example, a load on a spindle detected by the state quantity detection section 140, sampled at prescribed sampling intervals during a prescribed period in the past, or a peak value during a prescribed period in the past of the rotational speed of motor 120, also detected by the state quantity detection section 140. The feature quantity creation section 210 performs preprocessing and normalization of the state quantity detected by the state quantity detection section 140 to enable the derivation calculation section 220 to handle the state quantity.

[0031] The derivation calculation section 220, provided in the derivation processing section 200, derives a compensation measure of a thermal displacement of each axis of the machine controlled by the numerical control section 100, based on a learning model selected from the learning model memory section 300 based on a current condition of an individual difference between the machine controlled by the numerical control section 100 and the feature quantity created by the feature quantity creation section 210. The derivation calculation section 220 is implemented by applying a learning model 300, stored in the learning model memory section, to a platform capable of performing derivation processing through machine learning.The derivative processing section 220 can be configured, for example, to perform derivative processing using a multilayer neural network or to perform derivative processing using a known machine learning algorithm such as a Bayesian network, a support vector machine, or a Gaussian mixed model. The derivative processing section 220 can also be configured, for example, to perform derivative processing using a learning algorithm such as supervised learning, unsupervised learning, or augmented learning. Additionally, the derivative processing section 220 can be capable of performing derivative processing based on a variety of learning algorithms simultaneously.The derivation calculation section 220 represents a machine learning device based on a learning model for machine learning, selected from the learning model memory section 300, and derives a compensation measure of a thermal displacement of each axis of the machine controlled by the numerical control section 100 by performing a derivation processing using the feature quantity created by the feature quantity creation section 210 as input data to the machine learning device.

[0032] The learning model memory section 300 according to the present embodiment is capable of storing a plurality of learning models associated with a combination of conditions of an individual difference of the machine controlled by the numerical control section 100, which are specified by the condition specification section 110. The learning model memory section 300 can be implemented, for example, in the numerical control section 100, a cell computer, a host computer, a cloud server, or a database server.

[0033] The learning model memory section 300 stores a multitude of learning models 1, 2, ..., N, which are linked to a combination of conditions of an individual difference of the machine controlled by the numerical control section 100, which are specified by the condition specification section 110.A combination of conditions of an individual difference of the machine controlled by the numerical control section 100 as described here means a combination that relates to values ​​that each condition can take, a range of values ​​of each condition, and an enumeration of values ​​of each condition. For example, conditions can be classified according to combinations of a range of a production lot, a range of a frequency of use, a range of a cumulative operating time, a range of a maintenance history, and a range of a modification history (such as a list of components ordered by the customer that can be installed in the machine), as in . Fig. 3 shown.

[0034] The learning model 300, stored in the learning model memory section, is stored as information capable of constructing a learning model adaptable to derivative processing in the derivative computation section 220. In the case of a learning model using a multilayer neural network learning algorithm, the learning model 300 stored in the learning model memory section can be stored as the number of neurons (perceptrons) in each layer, a weight parameter between neurons (perceptrons) in each layer, and the like. In the case of a learning model using a Bayesian network learning model, the learning model 300 stored in the learning model memory section can be stored as a transition probability between nodes representing a Bayesian network.Each of the learning models stored in learning model memory section 300 can be a learning model that uses the same learning algorithm, or a learning model that uses a different learning algorithm, and the learning models stored in learning model memory section 300 can be a learning model that uses any type of learning algorithm, as long as the learning model can be used in derivative processing by the derivative computation section 220.

[0035] The learning model memory section 300 can store a learning model linked to a combination of conditions of an individual difference of the machine controlled by a numerical control section 100, or it can link or store a learning model that uses two or more different learning algorithms with a combination of conditions of an individual difference of the machine controlled by a numerical control section 100. The learning model memory section 300 can link or store a learning model that uses a different learning algorithm with any of the combinations that have overlapping ranges of conditions of an individual difference of the machine stored by a plurality of numerical control sections 100.At this point, for example, the learning model memory section 300, by defining usage conditions more precisely, such as a required throughput and a type of learning algorithm, with respect to the learning model according to a combination of conditions of an individual difference of the machine controlled by the numerical control section 100, enables the selection of a learning model according to the derivative computation section 220, whose executable derivative processing and processing capacity differ from another derivative computation section 220, with respect to a combination of conditions of an individual difference of the machine controlled by the numerical control section 100.

[0036] When the learning model memory section 300 receives an external read / write request for a learning model containing a combination of conditions of an individual difference of the machine controlled by the numerical control section 100, the learning model memory section 300 performs a read / write operation with respect to the learning model stored in association with the combination of conditions of an individual difference of the machine controlled by the numerical control section 100.At this point, the read / write request for a learning model may contain information about a derivation processing that can be performed by the derivation computation section 220 and the processing capacity of the derivation computation section 220, and in such a case, the learning model memory section 300 performs a read / write operation with respect to the learning model, which is associated with the combination of conditions of an individual difference of the machine controlled by the numerical control section 100, and with the derivation processing that can be performed by the derivation computation section 220, and the processing capacity of the derivation computation section 220.The learning model storage section 300 can be provided with a function that, in response to an external read / write request for a learning model, causes a read / write operation to be performed on a learning model associated with (a combination of) conditions specified by the condition specification section 110, based on those conditions. Providing such a function eliminates the need to provide the derivative computation section 220 or the learning model generation section 500 with a function to request a learning model based on conditions specified by the condition specification section 110.

[0037] Furthermore, the learning model storage section 300 can be configured to encrypt and store a learning model generated by the learning model generation section 500 and to decrypt the encrypted learning model when the learning model is read by the derivative computation section 220.

[0038] The compensation execution section 400 compensates for a thermal displacement of the machine controlled by the numerical control section 100 based on a compensation measure for the thermal displacement of each axis of the machine controlled by the numerical control section 100, derived by the derivation processing section 200. The compensation execution section 400 compensates for a thermal displacement of the machine controlled by the numerical control section 100, for example, by instructing a compensation measure for each axis to the numerical control section 100.

[0039] The learning model generation section 500 generates or updates (performs machine learning) a learning model 300, which is stored in the learning model memory section, based on a condition of an individual difference of the machine controlled by the numerical control section 100, which is specified by the condition specification section 110, and a feature quantity created by the feature quantity creation section 210, which specifies properties of a thermal state of the machine controlled by the numerical control section 100.The learning model generation section 500 selects a learning model as an object to generate or update based on the condition of an individual difference of the machine controlled by the numerical control section 100, which is specified by the condition specification section 110, and performs machine learning with respect to the selected learning model according to the feature quantity created by the feature quantity creation section 210, which specifies properties of a thermal state of the machine controlled by the numerical control section 100.The learning model generation section 500 performs learning when, for example, a worker manually sets a compensation measure for a thermal displacement of each axis of the machine with respect to the numerical control section 100, or when the compensation measure for a thermal displacement of each axis of the machine is set by another compensation means for a thermal displacement.In this case, if the machine is operated normally based on the set compensation measure of a thermal displacement (for example, if a worker finds that a workpiece has been machined with prescribed accuracy), the learning model generation section 500, with respect to a learning model selected on the basis of the condition of an individual difference of the machine controlled by the numerical control section 100, generates or updates (machine learns) the learning model using the feature quantity created by the feature quantity creation section 210, which specifies properties of a thermal state of the machine controlled by the numerical control section 100 as a state variable and a set compensation measure of a thermal displacement of each axis as labeled data.

[0040] If a learning model associated with (a combination of) conditions of an individual difference of the machine controlled by the numerical control section 100, specified by the condition specification section 110, is not the one stored in the learning model memory section 300, the learning model generation section 500 generates a new learning model associated with (the combination of) conditions; but if a learning model associated with (a combination of) conditions of an individual difference of the machine controlled by the numerical control section 100, specified by the condition specification section 110, is stored in the learning model memory section 300, the learning model generation section 500 updates the learning model by performing machine learning with respect to the learning model.If a large number of learning models are linked to (a combination of) conditions of an individual difference of the machine controlled by the numerical control section 100, which are specified by the condition specification section 110 and stored in the learning model memory section 300, the learning model generation section 500 can perform machine learning with respect to each of the learning models or can perform machine learning only with respect to a part of the learning models based on a learning process that can be executed by the learning model generation section 500 or the processing capacity of the learning model generation section 500.

[0041] The learning model generation section 500 can modify a learning model 300 stored in the learning model memory section to generate a new learning model. Examples of modification of a learning model by the learning model generation section 500 include the generation of a distilled model. A distilled model refers to a learned model obtained by performing learning from scratch in another machine learning device using an output in response to an input into a machine learning device with a built-in learned model. The learning model generation section 500 is able to store a distilled model obtained through such a process (referred to as the distillation process) as a new learning model in the learning model memory section 300 and use the distilled model.Since a distilled model is generally capable of producing a degree of accuracy comparable to that of an original learned model, despite being smaller in size, a distilled model is better suited for distribution to other computers over a network and the like. Other examples of modification of a learning model by the learning model generation section 500 involve the integration of learning models.If the structures of two or more learning models, which are stored in association with (a combination of) conditions of an individual difference of the machine controlled by the numerical control section 100, are similar to each other, and if, for example, a value of each weight parameter is within a prescribed threshold that is preset, the learning model generation section 500 can first integrate (combinations of) conditions of an individual difference of the machine controlled by the numerical control section 100 linked to the learning models, and then store any of the two or more learning models with similar structures in association with the integrated (combinations of) conditions.

[0042] Fig. Figure 4 is a schematic functional block diagram of system 1 for compensating a thermal displacement according to a second embodiment.

[0043] In the thermal displacement compensation system 1 according to the present embodiment, each functional block is mounted on a single numerical controller 2. By adopting such a configuration, the thermal displacement compensation system 1 according to the present embodiment derives a compensation measure for the thermal displacement of each axis of the machine controlled by the numerical controller 100 using a different learning model according to an individual difference of a machine controlled by the numerical controller 2, and performs a compensation of the thermal displacement of each axis of the machine based on the result of the derivation. Additionally, each learning model can be generated / updated by a numerical controller 2 according to a condition of an individual difference of the machine controlled by the numerical controller 100.

[0044] Fig. Figure 5 is a schematic functional block diagram of system 1 for compensating a thermal displacement according to a third embodiment.

[0045] In the system 1 for compensating for thermal displacement according to the present embodiment, the numerical control section 100, the derivation processing section 200, and the compensation execution section 400 are mounted on the numerical controller 2, and the learning model storage section 300 and the learning model generation section 500 are mounted on a machine learning device 3, which is connected to the numerical controller 2 via a standard interface or a network. The machine learning device 3 can be mounted on a cell computer, a host computer, a cloud server, or a database server.By adopting such a configuration, since derivation processing using a learned model that performs relatively easy processing can be performed on the numerical controller 2 and generation / update processing of a learning model that performs relatively heavy processing can be performed on the machine learning apparatus 3, operations of system 1 to compensate for a thermal shift can be carried out without disturbing intrinsic operations of the numerical controller 2.

[0046] Fig. Figure 6 is a schematic functional block diagram of system 1 for compensating a thermal displacement according to a fourth embodiment.

[0047] In the system 1 for compensating for a thermal displacement according to the present embodiment, the numerical control section 100 is mounted on the numerical controller 2, and the derivation calculation section 220, the learning model storage section 300, and the learning model generation section 500 are mounted on the machine learning device 3, which is connected to the numerical controller 2 via a standard interface or a network. Additionally, the compensation execution section 400 is provided separately.Furthermore, in the present embodiment of System 1 for compensating for a thermal shift, the feature quantity creation section 210 is omitted, assuming that a state quantity detected by the state quantity detection section 140 is data that can be used without modification in a derivation processing by the derivation calculation section 220 and in a generation / update processing of a learning mode by the learning model generation section 500. Since, by adopting such a configuration, derivation processing using a learned model and generation / update processing of a learning model can be performed on the machine learning apparatus 3, operations of System 1 for compensating for a thermal shift can be carried out without disturbing intrinsic operations of the numerical control 2.

[0048] Fig. Figure 7 is a schematic functional block diagram of system 1 for compensating a thermal displacement according to a fifth embodiment.

[0049] In the system 1 for compensating for a thermal displacement according to the present embodiment, each functional block is mounted on a single numerical control 2. Furthermore, in the system 1 for compensating for a thermal displacement according to the present embodiment, a plurality of learned learning models, which are linked to combinations of conditions of an individual difference of the machine controlled by the numerical control section 100, are already stored in the learning model memory section 300, and the configuration of the learning model generation section 500 is omitted under the assumption that no generation / update of learning models is performed.By adopting such a configuration, system 1, in order to compensate for a thermal displacement according to the present embodiment, derives a compensation measure for the thermal displacement of each axis of the machine controlled by the numerical control section 100 using a different learning model, for example, based on the individual differences of a machine controlled by the numerical control 2, and performs a compensation of the thermal displacement of each axis of the machine based on the result of the derivation. Additionally, since no arbitrary update of a learning model is performed, the configuration described above can, for example, be assumed to be a configuration of the numerical control 2 that is delivered to a customer.

[0050] Fig. Figure 8 is a schematic functional block diagram of a modification of system 1 to compensate for a thermal displacement ( Fig. 7) according to the fifth embodiment.

[0051] System 1 for compensating for a thermal displacement according to the present embodiment represents an example in which the learning model memory section 300 according to the fifth embodiment ( Fig. 7) is mounted on an external data storage device 4, which is connected to the numerical control 2.

[0052] Since the present modification allows the storage of large-capacity learning models in external data storage 4, enabling the use of a large number of learning models and simultaneously allowing the reading of learning models without involving a network and the like, the present modification is effective when derivation processing is to be performed in real time.

[0053] Fig. Figure 9 is a schematic functional block diagram of system 1 for compensating a thermal displacement according to a sixth embodiment.

[0054] In the system 1 for compensating for thermal displacement according to the present embodiment, the numerical control section 100 is mounted on the numerical controller 2, and the derivative calculation section 220 and the learning model storage section 300 are mounted on the machine learning device 3, which is connected to the numerical controller 2 via a standard interface or a network. The machine learning device 3 can be mounted on a cell computer, a host computer, a cloud server, or a database server.Furthermore, in System 1 for compensating for a thermal shift according to the present embodiment, a plurality of learned learning models, which are linked to combinations of conditions of an individual difference of the machine controlled by the numerical control section 100, are already stored in the learning model memory section 300, and the configuration of the learning model generation section 500 is not carried out, assuming that no generation / update of the learning models is performed. Additionally, in System 1 for compensating for a thermal shift according to the present embodiment, the configuration of the feature quantity creation section 210 is omitted, assuming that a state quantity detected by the state quantity detection section 140 is data that can be used without modification in derivative processing by the derivative calculation section 220.By adopting such a configuration, for example, system 1, according to the present embodiment, derives a compensation measure for the thermal displacement of each axis of the machine controlled by the numerical control section 100 using a different learning model, based, for example, on the individual differences of a machine controlled by the numerical control 2, and performs a compensation of the thermal displacement of each axis of the machine based on the result of the derivation. Since, in addition, no arbitrary updating of a learning model is performed, the configuration described above can, for example, be assumed to be a configuration of the numerical control 2 that is delivered to a customer.

[0055] Fig. Figure 10 is a schematic flow diagram of a process carried out by the system 1 to compensate for a thermal displacement according to the present invention.

[0056] The flowchart that is in Fig. Figure 10 shows a process if an update of a learning model in system 1 to compensate for a thermal shift is not carried out (fifth and sixth embodiments). • [Step SA01] The condition specification section 110 specifies a condition of an individual difference of the machine, which is controlled by the numerical control section 100. • [Step SA02] The state quantity detection section 140 detects a state of the machine controlled by the numerical control section 100 as a state quantity. • [Step SA03] The feature quantity creation section 210 creates a feature quantity that specifies properties of a thermal state of the machine controlled by the numerical control section 100, based on the state quantity detected in step SA02. • [Step SA04] The derivation calculation section 220 selects a learning model corresponding to the condition of an individual difference of the machine controlled by the numerical control section 100, specified in step SA01, from the learning model memory section 300 as the learning model to be used for derivation and reads it. • [Step SA05] The derivation calculation section 220 derives a compensation measure of a thermal displacement of each axis of the machine controlled by the numerical control section 100 based on the learning model read in step SA04 and the feature quantity created in step SA03. • [Step SA06] The compensation execution section 400 performs a compensation of a thermal displacement based on the compensation measure of a thermal displacement of each axis of the machine, which was derived in step SA05.

[0057] Fig. Figure 11 is a schematic flow diagram of a process carried out by the system 1 to compensate for a thermal displacement according to the present invention.

[0058] The in Fig. Figure 11 shows a processing sequence if a learning model is generated / updated in system 1 to compensate for a thermal shift (first to fourth embodiments). • [Step SB01] The condition specification section 110 specifies a condition of an individual difference of the machine, which is controlled by the numerical control section 100. • [Step SB02] The state quantity detection section 140 detects a state of the machine controlled by the numerical control section 100 as a state quantity. • [Step SB03] The feature quantity creation section 210 creates a feature quantity that specifies properties of a thermal state of the machine controlled by the numerical control section 100, based on the state quantity detected in step SB02. • [Step SB04] The derivation calculation section 220 selects and reads a learning model from the learning model memory section 300 as a learning model to be used for derivation, according to the condition of an individual difference of the machine controlled by the numerical control section 100, specified in step SB01. • [Step SB05] The learning model generation section 500 determines whether a learned learning model has been generated in the learning model memory section 300 according to the condition of an individual difference of the machine controlled by the numerical control section 100, specified in step SB01. If a learned learning model has been generated, processing continues with step SB07, but if a learned learning model has not been generated, processing continues with step SB06. • [Step SB06] The learning model generation section 500 generates / updates the learning model according to the condition of an individual difference of the machine controlled by the numerical control section 100, specified in step SB01, based on the feature quantity created in step SB03, and processing continues with step SB01. • [Step SB07] The derivation calculation section 220 derives a compensation measure of a thermal displacement of each axis of the machine controlled by the numerical control section 100 based on the learning model read in step SB04 and the feature quantity created in step SB03. • [Step SB08] The compensation execution section 400 performs a compensation of a thermal displacement based on the compensation measure of a thermal displacement of each axis of the machine, which was derived in step SB07.

[0059] Fig. Figure 12 is a schematic hardware configuration diagram showing main parts of a numerical control and a machine learning apparatus according to an embodiment of the present invention.

[0060] A CPU 11, provided in the numerical control 2 according to the present embodiment, is a processor that controls the entire numerical control 2. The CPU 11 reads a system program stored in a ROM 12 via a bus 20 and controls the entire numerical control 2 according to the system program. A RAM 13 temporarily stores computational and display data, various data entered by an operator via an input section (not shown), and the like.

[0061] A non-volatile memory 14 is configured as a memory whose data state is retained even when the power supply to the numerical control 2, supported by a battery (not shown) or the like, is switched off. The non-volatile memory 14 stores a work program read via an interface 15, a work program entered via a display / MDI unit 70 (which will be described later), and the like. The work program relating to a repetitive control stored in the non-volatile memory 14 can be accessed on the RAM 13 during operation. In addition, various system programs (including a system program for controlling communication with the machine learning device 3, which will be described later) necessary for the operation of the numerical control 2 are pre-written in the ROM 12.

[0062] Interface 15 is an interface for connecting the numerical control 2 to an external device 72, such as an adapter. Work programs, various parameters, and the like are read from the external device 72. Additionally, the program relating to repetitive control, various parameters, and the like, which are prepared in the numerical control 2, can be stored in the external data storage device by the external device 72. A programmable machine control (PMC) 16 outputs a signal via an I / O unit 17 to peripheral devices (for example, an actuator, such as a robot hand for tool changes) of a machine tool in order to control the peripheral devices with a sequence program that is built into the numerical control 2.In addition, the programmable machine control receives 16 signals from various switches on a control panel and sensors located in a main body of the machine tool, and, after performing necessary signal processing, passes the processed signals to the CPU 11.

[0063] The display / MDI unit 70 is a manual data input device provided with a display, a keyboard, and the like, and an interface 18 receives commands and data from the keyboard of the display / MDI unit 70 and transmits the commands and data to the CPU 11. An interface 19 is connected to a control panel 71, which is provided with a manual pulse generator and the like, used in the manual drive of each axis.

[0064] The display / MDI unit 70 is also capable of displaying an intermediate value and a trend of derivative evaluation values ​​representing a thermal displacement state of a machine. While a final result can be obtained through various methods, such as a threshold discrimination method, a trend graph determination method, and an outlier detection method in embodiments of the proposed system, visualizing a portion of the process in which the result is obtained provides a result that aligns with the industrial intuition of a worker actually operating a machine tool on a production line.

[0065] Upon receiving a motion command for an axis from the CPU 11, an axis control circuit 30, used to control an axis provided in the machine tool, issues an axis command to a servo amplifier 40. The servo amplifier 40 receives the command and drives the motor 120, which moves an axis provided in the machine tool. The axis motor 120 has a built-in position / speed detector and performs position / speed feedback control by feeding a position / speed feedback signal from the position / speed detector back to the axis control circuit 30. The hardware configuration diagram in Fig. Figure 12 shows only one of each of axis control circuit 30, servo amplifier 40 and motor 120, in fact there are as many components to prepare as the number of axes provided in the machine tool which is a control object.

[0066] Interface 21 is an interface for connecting the numerical control 2 and the machine learning device 3. The machine learning device 3 is equipped with a processor 80, which controls the entire machine learning device 3, a ROM 81, which stores a system program and the like, a RAM 82, which performs temporary data storage during each machine learning process, and a non-volatile memory 83, which stores learning models and the like. The machine learning device 3 exchanges various data with the numerical control 2 via an interface 84 and interface 21.

[0067] While embodiments of the present invention have been described previously, it is clear that the present invention is not limited to the examples shown in the embodiments and can be carried out in various ways by making suitable modifications.

Claims

[1] System (1) for compensating for a thermal displacement, which performs a compensation of a thermal displacement of a machine, comprising the system (1) for compensating for a thermal displacement: a condition specification section (110) that specifies a condition of an individual difference of the machine; a state quantity detection section (140) that detects a state quantity which indicates a state of the machine; a derivation calculation section (220) that derives a compensation measure of a thermal displacement of the machine from the state quantity; a compensation execution section (400) that performs a compensation of a thermal displacement of the machine based on the compensation measure of a thermal displacement of the machine, which is derived by the derivation calculation section (220); a learning model generation section (500) that generates or updates a learning model through machine learning that uses the state quantity; and a learning model storage section (300) that stores at least one learning model generated by the learning model generation section (500) in conjunction with a combination of conditions specified by the condition specification section (110), wherein The derivation calculation section (220) derives the compensation measure of a thermal displacement of the machine by selectively using, on the basis of a condition of an individual difference of the machine, which is specified by the condition specification section (110), at least one learning model from the learning models which are stored in the learning model storage section (300). [2] System (1) for compensating a thermal displacement according to claim 1, further comprising a feature quantity creation section (210) which creates a feature quantity characterizing a thermal state of the machine from the state quantity detected by the state quantity detection section (140), wherein the derivation calculation section (220) derives the compensation measure of a thermal displacement of the machine from the feature quantity and The learning model generation section (500) generates or updates a learning model by machine learning that uses feature quantity. [3] System (1) for compensating a thermal displacement according to claim 1 or 2, wherein the learning model generation section (500) generates a new learning model by performing a modification of an existing learning model stored in the learning model storage section (300). [4] System (1) for compensating a thermal displacement according to any one of claims 1 to 3, wherein the learning model storage section (300) encrypts and stores a learning model generated by the learning model generation section (500) and decrypts the encrypted learning model when a learning model is read by the derivative calculation section (220). [5] System (1) for compensating for a thermal displacement, which performs a compensation of a thermal displacement of a machine, comprising the system (1) for compensating for a thermal displacement: a condition specification section (110) that specifies a condition of an individual difference of the machine; a state quantity detection section (140) that detects a state quantity which indicates a state of the machine; a derivation calculation section (220) that derives a compensation measure of a thermal displacement of the machine from the state quantity; a compensation execution section (400) that performs a compensation of a thermal displacement of the machine based on the compensation measure of a thermal displacement of the machine, which is derived by the derivation calculation section (220); and a learning model memory section (300) which stores at least one learning model that has been pre-associated with a combination of operations of the machine, wherein The derivation calculation section (220) derives the compensation measure of a thermal displacement of the machine by selectively using, on the basis of a condition of an individual difference of the machine, which is specified by the condition specification section (110), at least one learning model from the learning models which are stored in the learning model storage section (300). [6] System (1) for compensating a thermal displacement according to claim 5, further comprising a feature quantity creation section (210) which creates a feature quantity characterizing a thermal state of the machine from the state quantity detected by the state quantity detection section (140), wherein The derivation calculation section (220) derives the compensation measure of a thermal displacement of the machine from the feature quantity. [7] Numerical control (2) comprising the condition specification section (110) and the state quantity detection section (140) of a system according to one of claims 1 to 4 or of a system according to one of claims 5 to 6. [8] Method for compensating for a thermal displacement, comprising the steps: Specifying a condition of an individual difference of a machine; Detecting a state quantity that indicates a state of the machine; Deriving a compensation measure for a thermal displacement of the machine from the quantity of state; Performing a compensation for thermal displacement of the machine based on the compensation measure for thermal displacement; and Generating or updating a learning model through machine learning that uses state quantity, where In the derivation step, a learning model to be used is selected based on the condition specified in the condition specification step. This learning model is selected from at least one learning model that is linked in advance to a combination of operating conditions of the machine, and the compensation measure for a thermal displacement of the machine is derived using the selected learning model. [9] Method for compensating for a thermal displacement according to claim 8, further comprising a step towards creating a feature quantity that characterizes a thermal state of the machine from the state quantity, wherein In the derivation step, the compensation measure of a thermal displacement of the machine is derived from the feature quantity and In the step of generating or updating a learning model, the learning model is generated or updated by machine learning, which uses the feature quantity. [10] Method for compensating for a thermal displacement, comprising the steps: Specifying a condition of an individual difference of a machine; Detecting a state quantity that indicates a state of the machine; Deriving a compensation measure for a thermal displacement of the machine from the quantity of state; and Performing a compensation for a thermal displacement of the machine based on the compensation measure of a thermal displacement, wherein In the derivation step, a learning model to be used is selected from at least one learning model based on the condition specified in the condition specification step, which is linked in advance to a combination of conditions of an individual difference of the machine, and the compensation measure of a thermal displacement of the machine is derived using the selected learning model. [11] Method for compensating for a thermal displacement according to claim 10, further comprising a step towards creating a feature quantity that characterizes a thermal state, due to an individual difference of the machine, from the state quantity, whereby In the derivation step, the compensation measure for a thermal displacement of the machine is derived from the feature quantity. [12] Learning model set formed by linking each of a plurality of learning models with a combination of conditions under which compensation for a thermal displacement of a machine is to be carried out, wherein Each of the multitude of learning models is a learning model that is generated or updated based on a state quantity that specifies a state of the machine under a condition of an individual difference of the machine, and A learning model is selected from the multitude of learning models based on the condition of an individual difference of the machine, and the selected learning model is to be used in a process to derive a compensation measure for a thermal displacement of the machine.

Images