Multi-task real-time inference scheduling system machine tool and method thereof

The multi-task real-time inferential scheduling system addresses data and equipment variability issues in machine tools by using a neural network to create adaptive learning models, enhancing reliability and reducing maintenance costs through real-time task inference and scheduling.

WO2024237500A1PCT designated stage expired Publication Date: 2024-11-21DN SOLUTIONS CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
PCT/KR2024/005187
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-12
Filing Date
2024-04-17
Publication Date
2024-11-21

AI Technical Summary

Technical Problem

Conventional multi-task learning models for machine tools face challenges such as data imbalance, overfitting, heterogeneity in data distribution, and difficulty in creating reliable models due to equipment variations, leading to suboptimal scheduling and increased maintenance costs.

Method used

A multi-task real-time inferential scheduling system and method that connects central and individual control units through a network, using a neural network to create a multi-task learning model based on real-time usage context, considering dependencies and environments, to infer and schedule tasks effectively.

Benefits of technology

Improves stability and reliability of machine tools, reduces maintenance costs, and enhances worker safety by automatically creating and updating learning models, enabling optimal task scheduling and preventing defective products.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure KR2024005187_21112024_PF_FP_ABST
    Figure KR2024005187_21112024_PF_FP_ABST
Patent Text Reader

Abstract

The present invention relates to a multi-task real-time inference scheduling system and real-time inference scheduling method of a machine tool, wherein a central control unit is connected to each of one or more individual control units through a network, receives a use context of each machine tool through each individual control unit, generates a multi-task learning model through a neural network, infers multiple tasks required to be performed by the individual control unit of each machine tool through machine learning by using real-time use contexts collected during operation of the machine tool by a use scenario, and schedules the multiple tasks of the machine tool through machine learning.
Need to check novelty before this filing date? Find Prior Art

Description

Multi-task real-time inference scheduling system and real-time inference scheduling method for machine tools

[0001] The present invention relates to a multi-task real-time inference scheduling system and a real-time inference scheduling method for a machine tool, and more particularly, to a multi-task real-time inference scheduling system and a real-time inference scheduling method for a machine tool, in which a central control unit is connected to one or more individual control units through a network, the usage context of each machine tool is transmitted through each individual control unit, a multi-task learning model is created through a neural network, and the multi-task to be performed by the individual control unit of each machine tool is inferred through machine learning through the real-time usage context collected during operation according to the usage scenario for each machine tool, and the multi-task of the corresponding machine tool is scheduled through machine learning.

[0002] In general, a machine tool refers to a machine used for the purpose of processing a metal / non-metal workpiece into a desired shape and size using an appropriate tool using various cutting or non-cutting processing methods.

[0003] Various types of machine tools, including turning centers, vertical / horizontal machining centers, gate-type machining centers, Swiss turns, electrical discharge machines, horizontal NC boring machines, and CNC lathes, are widely used in various industrial fields for their respective tasks.

[0004] In addition, the machine tool is equipped with a table on which the workpiece material is placed and transferred for processing the workpiece, a pallet for preparing the workpiece before processing, a spindle on which a tool or workpiece is combined and rotates, a tailstock for supporting the workpiece during processing, a vibration damper, etc.

[0005] Typically, in machine tools, the table, tool rest, spindle, tailstock, and oscillation unit are equipped with a transfer unit that moves along the transfer axis to perform various processing operations.

[0006] Additionally, machine tools generally use multiple tools for various processing operations, and tool magazines or turrets are used as tool storage spaces for storing multiple tools.

[0007] These machine tools use multiple tools for various processing tasks, and a tool magazine is used as a tool storage space to store multiple tools.

[0008] Additionally, machine tools are generally equipped with an automatic tool changer (ATC) to retrieve or retract a specific tool from the tool magazine under the command of the numerical control unit in order to improve the productivity of the machine tool.

[0009] Additionally, machine tools are typically equipped with an automatic pallet changer (APC) to minimize downtime. The APC automatically exchanges pallets between the workpiece processing area and the workpiece installation area. Pallets can be loaded with workpieces.

[0010] Numerical control (NC) is rapidly advancing in various types of machine tools currently in use.

[0011]

[0012] *Recently, as computer numerical control (NC) is used to automatically control machine tools, machine tools that use computer numerical control (CNC) technology, which is a further development of numerical control (NC), are rapidly becoming widespread.

[0013] Numerical control (NC) or computer numerical control (CNC) machine tools are equipped with an operator panel. This panel contains various function switches or buttons and a monitor.

[0014] In this way, machine tools that use numerical control or computer numerical control must have an operator who creates, manages, or operates a machining program to perform machining, and must create various control programs such as machining programs or NC codes for individual control commands of the machine tool and input them into the NC control unit of the machine tool.

[0015] Unlike general computers, many special-purpose home appliances and machine tools have embedded systems or edge devices built into them to perform a given task.

[0016] Recent machine tools are equipped with a numerical control system called CNC (Computer Numerical Control), and these CNCs can control the operation of the machine tool in various ways as desired.

[0017] Today, more and more machine tool controls are required to perform machining of diverse and complex shapes or to perform rapid and precise machining of workpieces for mass production.

[0018] However, unlike the multi-task through conventional machine learning using general neural networks, in the case of machine tools, the problem of imbalance in anomaly detection data occurs, which limits the distribution of collected data and may result in neural network overfitting.

[0019] Furthermore, machine tools struggle to achieve uniformity and repeatability in data acquisition, posing challenges for machine learning. For example, even when data is acquired using the same equipment, materials, tools, machining programs, and processing conditions, the resulting data distribution can vary depending on the equipment's installation, age, and degree of wear.

[0020] Furthermore, due to equipment-related inhomogeneity in data distribution, even machine tools of the same model may have structural differences due to design updates from unit to unit. This can lead to differences in equipment stiffness, resulting in varying sensor signal responsiveness even when the same cutting force is applied under the same machining conditions. Furthermore, equipment control parameters, such as acceleration and deceleration parameters, can influence dynamic behavior depending on the user's preferred processing, resulting in different feed-axis stiffness settings.

[0021] Moreover, since the multi-task of conventional machine tools proceeds without considering the dependencies between tasks for processing multiple operations, it is difficult to create a multi-task learning model that matches the characteristics of the equipment, and even if it is created, there is a problem of low reliability.

[0022] In addition, the multi-task of conventional machine tools is performed only with fixed tasks without considering the multi-task priority or dependency relationship due to the constraints of the limited resources embedded in each machine tool. As described above, it is not possible to create an optimal multi-task learning model according to the diversity and status of the machine tool and to perform optimal multi-tasking scheduling for the machine tool through inference based on the model. As a result, there are problems such as limitations in abnormal diagnosis or monitoring of the machine tool, reduction in reliability and safety, inconvenience to the operator, and increase in productivity and maintenance costs.

[0023] Accordingly, there is an urgent need to develop a multi-task real-time inference scheduling system and a real-time inference scheduling method for machine tools that infer multi-tasks to be performed by the individual control unit of each machine tool through machine learning based on real-time usage contexts collected during operation according to usage scenarios for each machine tool and perform scheduling of the multi-tasks of the machine tool through machine learning.

[0024] The present invention is to solve the above problems, and the purpose of the present invention is to provide a multi-task real-time inference scheduling system and a real-time inference scheduling method for a machine tool, in which a central control unit is connected to one or more individual control units through a network, and the usage context of each machine tool is transmitted through each individual control unit, a multi-task learning model is created through a neural network, and the multi-task to be performed by the individual control unit of each machine tool is inferred through machine learning through real-time usage context collected during operation according to the usage scenario for each machine tool, and the multi-task of the corresponding machine tool is scheduled through machine learning.

[0025] In order to achieve the object of the present invention, a multi-task real-time inference scheduling system for a machine tool according to the present invention comprises: an individual control unit for controlling the operation of each machine tool; and a central control unit connected to one or more of the individual control units through a network, and receiving a usage context of each of the machine tools through each of the individual control units and generating a multi-task learning model through a neural network; wherein the central control unit can infer a multi-task to be performed by each of the individual control units of the machine tools through machine learning based on real-time usage contexts collected during operation according to usage scenarios for each of the machine tools, and schedule the multi-task of the corresponding machine tools through machine learning.

[0026] In addition, in another preferred embodiment of the multi-task real-time inference scheduling system for a machine tool according to the present invention, the usage context for generating a multi-task learning model through the neural network of the multi-task real-time inference scheduling system for a machine tool and the real-time usage context collected during operation according to the usage scenario for each of the machine tools may be composed of information related to the usage scenario occurring in the entire usage cycle of each of the machine tools.

[0027] In addition, in another preferred embodiment of the multi-task real-time inference scheduling system for a machine tool according to the present invention, the usage context for generating a multi-task learning model through the neural network of the multi-task real-time inference scheduling system for a machine tool and the real-time usage context collected during operation according to the usage scenario for each of the above machines may include at least one of processing material information, equipment information, tool information, NC information, processing information, process information, maintenance information, surrounding information, measurement information, geometric information, and technical information.

[0028] In addition, in another preferred embodiment of the multi-task real-time inference scheduling system for a machine tool according to the present invention, the central control unit of the multi-task real-time inference scheduling system for a machine tool may include: a communication module for communicating with each of the individual control units via a network, either wired or remotely; a database module for storing usage contexts of each of the machine tools received from each of the individual control units via the communication module; a classification module for classifying the usage contexts of each of the machine tools received from each of the individual control units stored in the database module into common usage contexts and individual usage contexts according to dependencies; a featuring module for providing featuring according to characteristics of the common usage contexts and individual usage contexts of each of the machine tools classified in the classification module; and a generation module for generating a common multi-task learning model and an individual multi-task learning model through machine learning based on the common usage contexts and individual usage contexts of each of the machine tools featured through the featuring module.

[0029] In addition, in another preferred embodiment of the multi-task real-time inference scheduling system for a machine tool according to the present invention, the classification module of the multi-task real-time inference scheduling system for a machine tool can classify the usage contexts collected from each of the machine tools into common usage contexts and individual usage contexts through clustering and leveling by similar machine tool usage contexts through machine learning.

[0030] In addition, in another preferred embodiment of the multi-task real-time inference scheduling system for a machine tool according to the present invention, the featuring module of the multi-task real-time inference scheduling system for a machine tool can be performed step by step through a neural network configured by layers.

[0031] In addition, in another preferred embodiment of the multi-task real-time inference scheduling system for a machine tool according to the present invention, the generation module of the multi-task real-time inference scheduling system for a machine tool can generate a common multi-task learning model and an individual multi-task learning model by dividing them by layer according to the dependency relationship between the common usage context and the individual usage context of each machine tool featured by layer.

[0032] In addition, in another preferred embodiment of the multi-task real-time inference scheduling system for machine tools according to the present invention, the central control unit of the multi-task real-time inference scheduling system for machine tools comprises: a confirmation module for confirming whether each of the machine tools is operating according to a usage scenario for each of the machine tools; a collation module for collating and storing real-time usage contexts of machine tools operating according to usage scenarios for each of the machine tools when each of the machine tools is operating as a result of the confirmation by the confirmation module; a selection module for selecting whether the real-time usage contexts stored in the collation module are useful; an analysis module for extracting useful real-time usage contexts from among the real-time usage contexts according to a selection result of the selection module, classifying the extracted real-time usage contexts into real-time common usage contexts and real-time individual usage contexts, and analyzing features of the corresponding usage contexts; a judgment module for determining whether to proceed with common multi-task or individual multi-task according to the dependency relationship and features between the real-time common usage context and the real-time individual usage context according to the analysis result of the analysis module; The system may further include an inference module that automatically infers in real time the multi-tasks that must be performed in the individual control units of each machine tool currently in operation according to the corresponding usage scenario through machine learning using the common multi-task learning model and individual multi-task learning models generated by the generation module based on the judgment result of the judgment module.

[0033] In addition, in another preferred embodiment of the multi-task real-time inference scheduling system for a machine tool according to the present invention, the central control unit of the multi-task real-time inference scheduling system for a machine tool may further include a scheduling module that schedules multi-tasks to be performed in individual control units of each of the machine tools currently in operation according to a usage scenario based on the inference result of the inference module.

[0034] In addition, in another preferred embodiment of the multi-task real-time inference scheduling system for a machine tool according to the present invention, the scheduling module of the multi-task real-time inference scheduling system for a machine tool can derive scheduling to be performed in each of the machine tools in a time series manner through machine learning by a common multi-task learning model generated by the generation module and an individual multi-task learning model, reflecting priority, real-time responsiveness, and / or periodicity.

[0035] In addition, in another preferred embodiment of the multi-task real-time inference scheduling system for a machine tool according to the present invention, the central control unit of the multi-task real-time inference scheduling system for a machine tool may further include a feedback module that receives and stores information generated when performing multi-tasks in individual control units of each of the currently operating machine tools according to the scheduling created by the scheduling module, and continuously modifies and changes the common multi-task learning model and individual multi-task learning models generated by the generation module by reflecting the feedback information through machine learning to upgrade them.

[0036] In addition, in another preferred embodiment of the multi-task real-time inference scheduling system for a machine tool according to the present invention, the central control unit of the multi-task real-time inference scheduling system for a machine tool may further include an output module that converts data or results of the communication module, the database module, the classification module, the featuring module, the generation module, the confirmation module, the collation module, the selection module, the analysis module, the inference module, and the scheduling module into graphics, text, or images and displays them or transmits them to the individual control unit using a user graphic guidance program.

[0037] In addition, in another preferred embodiment of the multi-task real-time inference scheduling system for a machine tool according to the present invention, the individual control unit of the multi-task real-time inference scheduling system for a machine tool may include a memory unit that stores information for controlling the operation of each machine tool; an operating unit that operates the machine tool according to a usage scenario through the information stored in the memory unit; a collecting unit that collects real-time usage context of the machine tool operated by the operating unit; and a transmitting and receiving unit that transmits the real-time usage context collected by the collecting unit or feedback information by scheduling to the central control unit or receives a newly generated or modified multi-task inference result or multi-task scheduling from the central control unit.

[0038] In addition, in another preferred embodiment of the multi-task real-time inference scheduling system for a machine tool according to the present invention, the individual control unit of the multi-task real-time inference scheduling system for a machine tool may further include a display unit that displays data or results of the memory unit, the operation unit, the collection unit, and the transmission / reception unit.

[0039] In order to achieve another object of the present invention, a multi-task real-time inference scheduling method of a machine tool according to the present invention may include the steps of: connecting a central control unit and an individual control unit for controlling the operation of each machine tool by wire or remotely through a network; collecting and storing basic data and usage contexts of each machine tool received from each of the individual control units; classifying the usage contexts of each machine tool received from each of the individual control units into common usage contexts and individual usage contexts according to dependencies; assigning features according to the characteristics of the common usage contexts and individual usage contexts classified into the common usage contexts and individual usage contexts of each machine tool according to the classification results; and generating a common multi-task learning model and an individual multi-task learning model by classifying them by layer according to the dependencies between the common usage contexts and individual usage contexts of each machine tool featured by layer through machine learning based on the common usage contexts and individual usage contexts of each machine tool featured by the featuring results.

[0040] In addition, in another preferred embodiment of the multi-task real-time inference scheduling method for a machine tool according to the present invention, the multi-task real-time inference scheduling method for a machine tool comprises: a step of checking whether each of the machine tools is operated according to a usage scenario for each of the machine tools after the generating step; a step of collecting and storing real-time usage contexts of machine tools that are operated according to usage scenarios for each of the machine tools if the checking result indicates that each of the machine tools is operating; a step of selecting whether the collected and stored real-time usage contexts are useful; a step of extracting useful real-time usage contexts from the real-time usage contexts according to the selection result, classifying the extracted real-time usage contexts into real-time common usage contexts and real-time individual usage contexts, and analyzing features of the corresponding usage contexts; a step of determining whether to proceed with a common multi-task or an individual multi-task according to the dependency relationship and features between the real-time common usage contexts and the real-time individual usage contexts according to the analysis result; And, based on the judgment result, a step of automatically inferring in real time the multi-tasks that must be performed in the individual control unit of each machine tool currently in operation according to the corresponding usage scenario through machine learning using the generated common multi-task learning model and the individual multi-task learning model may be further included.

[0041] In addition, in another preferred embodiment of the multi-task real-time inference scheduling method for a machine tool according to the present invention, the multi-task real-time inference scheduling method for a machine tool may further include a step of scheduling multi-tasks to be performed in each individual control unit of each machine tool currently in operation in a time series manner by reflecting priority, real-time responsiveness, and / or periodicity based on the inference result after the inference step, by using machine learning by a common multi-task learning model generated by the generation module and an individual multi-task learning model.

[0042] In addition, in another preferred embodiment of the multi-task real-time inference scheduling method of a machine tool according to the present invention, the multi-task real-time inference scheduling method of a machine tool may further include a step of receiving and storing feedback information generated when performing multi-tasks in individual control units of each of the machine tools currently in operation according to scheduling after the scheduling step, and continuously modifying and changing the common multi-task learning model and individual multi-task learning models generated by reflecting the feedback information through machine learning to upgrade them.

[0043] In addition, in another preferred embodiment of the multi-task real-time inference scheduling method for a machine tool according to the present invention, the usage context and the real-time usage context of the multi-task real-time inference scheduling method for a machine tool may be at least one of processing material information, equipment information, tool information, NC information, processing information, process information, maintenance information, surrounding information, measurement information, geometric information, and technical information.

[0044] The multi-task real-time inference scheduling system and the real-time inference scheduling method of a machine tool according to the present invention have a central control unit connected to one or more individual control units through a network, and the usage context of each machine tool is transmitted through each individual control unit, and a multi-task learning model is created by considering dependencies according to various environments and conditions through a neural network, thereby improving the stability and reliability of the machine tool, and by automatically continuously creating and updating the multi-task learning model regardless of the skill level, experience and knowledge of the operator, thereby promoting the convenience of the operator and preventing safety accidents in advance.

[0045] In addition, the multi-task real-time inference scheduling system and the real-time inference scheduling method of the machine tool according to the present invention infer the multi-tasks that should be performed in the individual control unit of each machine tool through machine learning through real-time usage contexts collected during operation according to usage scenarios for each machine tool, and schedule the multi-tasks of the corresponding machine tool through machine learning to schedule the optimal multi-task performance for each individual machine tool with each limited embedded environment, thereby easily and quickly performing expansion and usability expansion, promoting miniaturization of the machine tool and reducing space utilization, and reducing maintenance costs and time due to damage or destruction of the machine tool or workpiece by continuously and accurately and quickly making judgments through updates according to new situations of individual machine tools with changed conditions, ratios, or environments as well as existing machine tools, thereby increasing consumer satisfaction and enhancing exports.

[0046] Furthermore, the multi-task real-time inference scheduling system and real-time inference scheduling method of a machine tool according to the present invention automatically reflect the optimal scheduling to each machine tool to perform processing of a workpiece, thereby improving processing precision and productivity, preventing unnecessary occurrence of defective products, thereby conserving resources and increasing convenience for users and workers.

[0047] Figure 1 shows a conceptual diagram of a multi-task real-time inference scheduling system for a machine tool according to the present invention.

[0048] Figure 2 shows a configuration diagram of an individual control unit of a multi-task real-time inference scheduling system for a machine tool according to the present invention.

[0049] Figure 3 shows a configuration diagram of a central control unit of a multi-task real-time inference scheduling system for a machine tool according to the present invention.

[0050] FIG. 4 shows a configuration concept diagram for a usage context in a multi-task real-time inference scheduling system and real-time inference scheduling method for a machine tool according to one embodiment of the present invention.

[0051] FIG. 5 shows a configuration concept diagram for a tool holder spindle equipment context in a multi-task real-time inference scheduling system and real-time inference scheduling method for a machine tool according to another embodiment of the present invention.

[0052] FIG. 6 illustrates a conceptual diagram for generating a multi-task learning method in a multi-task real-time inference scheduling system and real-time inference scheduling method of a machine tool according to one embodiment of the present invention.

[0053] FIG. 7 is a conceptual diagram showing classification and featuring of usage contexts in a multi-task real-time inference scheduling system and real-time inference scheduling method for a machine tool according to one embodiment of the present invention.

[0054] FIG. 8 shows a flow chart of a multi-task real-time inference scheduling method for a machine tool according to one embodiment of the present invention.

[0055] FIG. 9 shows a detailed procedure diagram for multitasking inference in a multitask real-time inference scheduling method for a machine tool according to one embodiment of the present invention.

[0056] FIG. 10 shows a detailed procedure diagram for multitasking scheduling in a multitask real-time inference scheduling method for a machine tool according to one embodiment of the present invention.

[0057] Hereinafter, a multi-task real-time inference scheduling system and a real-time inference scheduling method for a machine tool according to an embodiment of the present invention will be described in detail with reference to drawings. The embodiments introduced below are provided as examples to ensure that those skilled in the art can sufficiently convey the spirit of the present invention. Therefore, the present invention is not limited to the embodiments described below and may be embodied in other forms. In addition, in the drawings, the size and thickness of the devices may be exaggerated for convenience. Like reference numbers represent like elements throughout the specification.

[0058] The present invention is capable of various modifications and embodiments. Therefore, specific embodiments are illustrated in the drawings and described in detail in the detailed description. The effects and features of the present invention, as well as the methods for achieving them, will become clear with reference to the embodiments described in detail below together with the drawings. However, the present invention is not limited to the embodiments disclosed below and can be implemented in various forms. In the following embodiments, the terms "first," "second," etc. are not used in a limiting sense but are used for the purpose of distinguishing one component from another. Furthermore, the singular expression includes the plural expression unless the context clearly indicates otherwise. Furthermore, terms such as "include" or "have" indicate the presence of a feature or component described in the specification, and do not preemptively exclude the possibility that one or more other features or components may be added. Furthermore, in the drawings, the sizes of components may be exaggerated or reduced for convenience of explanation. For example, the size and thickness of each component shown in the drawings are arbitrarily shown for convenience of explanation, and thus the present invention is not necessarily limited to what is shown.

[0059] The advantages and features of the present invention, and the methods for achieving them, will become clearer with reference to the embodiments described in detail below together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms. These embodiments are provided only to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined only by the scope of the claims. Like reference numerals designate like elements throughout the specification. The sizes and relative sizes of layers and regions in the drawings may be exaggerated for clarity of description.

[0060] The terminology used herein is for the purpose of describing embodiments and is therefore not intended to be limiting of the present invention. In this specification, the singular also includes the plural unless specifically stated otherwise. As used herein, the terms "comprise" and / or "comprising" do not exclude the presence or addition of one or more other components, steps, operations, and / or elements mentioned.

[0061] FIG. 1 is a conceptual diagram of a multi-task real-time inference scheduling system for a machine tool according to the present invention. FIG. 2 is a configuration diagram of individual control units of a multi-task real-time inference scheduling system for a machine tool according to the present invention. FIG. 3 is a configuration diagram of a central control unit of a multi-task real-time inference scheduling system for a machine tool according to the present invention. FIG. 4 is a conceptual diagram of a configuration of a usage context in a multi-task real-time inference scheduling system and a real-time inference scheduling method for a machine tool according to an embodiment of the present invention. FIG. 5 is a conceptual diagram of a configuration of a tool holder spindle equipment context in a usage context in a multi-task real-time inference scheduling system and a real-time inference scheduling method for a machine tool according to another embodiment of the present invention. FIG. 6 is a conceptual diagram for generating a multi-task learning method in a multi-task real-time inference scheduling system and a real-time inference scheduling method for a machine tool according to an embodiment of the present invention. FIG. 7 is a conceptual diagram of classification and featuring of usage contexts in a multi-task real-time inference scheduling system and a real-time inference scheduling method for a machine tool according to an embodiment of the present invention. Fig. 8 is a flowchart illustrating a multi-task real-time inference scheduling method for a machine tool according to an embodiment of the present invention. Fig. 9 is a detailed flowchart illustrating a multi-tasking inference in a multi-task real-time inference scheduling method for a machine tool according to an embodiment of the present invention. Fig. 10 is a detailed flowchart illustrating a multi-tasking scheduling in a multi-task real-time inference scheduling method for a machine tool according to an embodiment of the present invention.

[0062] Referring to FIGS. 1 to 10, a multi-task real-time inference scheduling system and a real-time inference scheduling method for a machine tool according to the present invention are described.

[0063] As illustrated in FIGS. 1 to 7, the multi-task real-time inference scheduling system for a machine tool according to the present invention includes an individual control unit (10) and a central control unit (20).

[0064] The individual control unit (10) controls the operation of each machine tool.

[0065] The central control unit (20) is connected to one or more of the above individual control units through a network, and receives the usage context of each machine tool through each of the above individual control units to create a multi-task learning model through a neural network.

[0066] The central control unit infers the multitasks that must be performed by the individual control units of each machine tool through machine learning based on real-time usage contexts collected during operation according to usage scenarios for each of the above-mentioned machine tools, and schedules the multitasks of the machine tools through machine learning.

[0067] Accordingly, it improves the stability and reliability of machine tools, promotes operator convenience by continuously generating and updating multi-task learning models regardless of the operator's skill level, experience, and knowledge, and has the effect of preventing safety accidents in advance.

[0068] As illustrated in FIG. 1, the central control unit (20) of the multi-task real-time inference scheduling system for a machine tool according to the present invention is connected to one or more individual control units through a network, and, if necessary, one or more of the individual control units can also be integrated and connected through a network.

[0069] As illustrated in Fig. 2, the individual control unit (10) of the multi-task real-time inference scheduling system for a machine tool according to the present invention includes a memory unit (11), an operation unit (12), a collection unit (13), a transmission / reception unit (14), and a display unit (15). The individual control unit includes an NC (numerical control, NC) or a CNC (computerized numerical control), and various numerical control programs are built in.

[0070] In addition, although not shown in the drawing, according to a preferred embodiment of the present invention, the individual control unit includes a main operation unit, and the main operation unit includes a screen display program and a data input program according to screen display selection, and performs a function of displaying a software switch on a display screen according to an output of the screen display program, and recognizing the on / off of the software switch to issue an input / output command for machine operation.

[0071] Additionally, but not necessarily limited to this, the main control unit is installed on the housing, case, or one side of the machine tool and is equipped with various function switches or buttons and a monitor capable of displaying various information.

[0072] The memory unit (11) stores information for controlling the operation of each machine tool. The memory unit may be a variety of storage devices such as ROM, RAM, EPROM, flash drive, hard drive, etc., and may also be web storage that performs the storage function of the memory unit on the Internet.

[0073] The operating unit (12) operates the machine tool according to the usage scenario using the information stored in the memory unit.

[0074] The collection unit (13) collects real-time usage contexts of the machine tool operated by the operation unit. That is, the collection unit collects one or more pieces of information from among processing material information, equipment information, tool information, NC information, processing information, process information, maintenance information, surrounding information, measurement information, geometric information, and technical information for each of the machine tools during operation according to the usage scenario.

[0075] The transmitter / receiver (14) transmits real-time usage context or feedback information based on scheduling collected in the collection unit to the central control unit or receives newly generated or modified multi-task inference results or multi-task scheduling from the central control unit.

[0076] The display unit (15) displays data or results of the memory unit, the operation unit, the collection unit, and the information transmission / reception unit. The display unit may include at least one of a liquid crystal display (LCD), a thin film transistor-liquid crystal display (TFT LCD), an organic light-emitting diode (OLED), a flexible display, a 3D display, and an e-ink display.

[0077] In addition, although not necessarily limited thereto, the display unit may be formed in the form of a touch screen and installed on the housing, case, or one side of the machine tool to display various function switches or buttons and various types of information.

[0078] In addition, the individual control unit can detect the manipulation or input of control commands through a touch screen, keypad, or keyboard on a display unit, etc., by an administrator or user for individual operation control of the machine tool, and manually input them into the memory information collection unit of the individual control unit.

[0079] As illustrated in FIG. 3, the central control unit (20) of the multi-task real-time inference scheduling system for a machine tool according to the present invention includes a communication module (21), a database module (22), a classification module (23), a featuring module (24), a generation module (25), a confirmation module (26), a collation module (27), a selection module (28), an analysis module (29), a judgment module (30), an inference module (31), a scheduling module (32), a feedback module (33), and an output module (34).

[0080] The communication module (21) communicates with each of the above individual control units via a network, either wired or remotely. The network with the individual control units of the communication module refers to a connection structure in which information can be exchanged between each node, such as the individual control units and / or the central control unit, and examples of such networks include, but are not limited to, a 3GPP (3rd Generation Partnership Project) network, an LTE (Long Term Evolution) network, a WIMAX (World Interoperability for Microwave Access) network, the Internet, a LAN (Local Area Network), a Wireless LAN (Wireless Local Area Network), a WAN (Wide Area Network), a PAN (Personal Area Network), a Bluetooth network, a satellite broadcasting network, an analog broadcasting network, a DMB (Digital Multimedia Broadcasting) network, etc.

[0081] Additionally, the individual control unit (10) and the central control unit (20) may be connected via Ethernet or serial communication as needed, but are not limited thereto.

[0082] The database module (22) stores the usage context of each machine tool received from each individual control unit via the communication module. The database module may be a variety of storage devices such as ROM, RAM, EPROM, flash drive, hard drive, etc., and may also be a web storage that performs the storage function of the database module on the Internet.

[0083] The classification module (30) classifies the usage contexts of each machine tool received from each of the individual control units stored in the database module into common usage contexts and individual usage contexts according to the dependency relationship.

[0084] For example, machine tools are equipped with a variety of sensors, including acceleration sensors, current sensors, and temperature sensors. Networks must be designed to efficiently process and fuse data from these diverse sensors. Analog sensors are stored in the form of time-series waveforms.

[0085] Each sensor has a different sampling rate and different significant frequency bands and resolutions. Time-series waveform data contains statistically significant variables (mean, moving average, standard deviation, maximum, minimum, etc.).

[0086] Frequency-domain features are also low-level features commonly required for feature extraction from sensor data. Alternatively, preprocessed data generated through signal processing methods such as wavelets, which simultaneously express time and frequency, can be utilized.

[0087] The classification module of the multi-task real-time inference scheduling system for machine tools classifies the usage contexts collected from each machine tool into common usage contexts and individual usage contexts through clustering and leveling by similar machine tool usage contexts through machine learning.

[0088] As shown in Fig. 4 and Table 1 below, the usage context for generating a multi-task learning model through a neural network and the real-time usage context collected during operation according to the usage scenario for each of the above-mentioned machine tools may be composed of information related to the usage scenario occurring in the entire usage cycle of each of the above-mentioned machine tools.

[0089] Specifically, the usage consensus for generating a multi-task learning model through the neural network and the real-time usage context collected during operation according to the usage scenario for each of the above-mentioned space machines may include one or more of processing material information, equipment information, tool information, NC information, processing information, process information, maintenance information, surrounding information, measurement information, geometric information, and technical information.

[0090] As illustrated in Figure 5, geometric information can include geometric information about the tool holder combination. Specifically, this can be used to define the stiffness between the tool, holder, and spindle, and analyze its impact. Furthermore, the data impact of the manufacturing joint configuration between each component can be identified.

[0091] The featuring module (24) provides features according to the characteristics of the common usage context and individual usage context of each machine tool classified in the classification module.

[0092] These feature modules are implemented step-by-step through a neural network structured by layers. This means that as feature processing is performed step-by-step, it is possible to track the features generated at each step. By observing the composition and changes in feature values ​​at each low, middle, and high level, inference and learning can be secured for each task.

[0093] For example, if accelerometer data is represented as an image, low-level features will identify fundamental, localized patterns. Intermediate-level features will combine low-level features to identify shapes and variations. Finally, high-level features can be expressed at a human-understandable level, such as harmonic frequencies or physical meanings.

[0094] In this way, it is possible to simultaneously receive input from sensors with different meanings and sampling, debugging the meaning of features calculated in the neural network of each sensor step and fusion, thereby producing diverse features compared to cases with a single modality.

[0095] In addition, when the feature module is performed step by step through a neural network structured by layers, not all neural networks are activated, but the data distribution is identified and only meaningful data is learned through this, or tasks that need to be inferred only during current use are selected and identified, enabling resource-efficient operation.

[0096] The generation module (25) generates a common multi-task learning model and an individual multi-task learning model through machine learning based on the common usage context and individual usage context of each machine tool featured through the above-mentioned featuring module.

[0097] For example, learning strategies are selected by identifying data feature distributions in real-time data pipelines. Data learning volume is controlled through selective data learning, data learning cycles, and strategies, thereby efficiently utilizing learning time and computing resources. Data used in a common multi-task learning model and data used in individual multi-task learning models may have different feature spaces.

[0098] Figure 6 illustrates an example of multichannel sensor time series data, including materials, equipment, tools, machining results, shapes, and processing conditions, in the context of machine tool usage. This time series can be edited section by section to suit various machine tool usage contexts. This allows for clustering of similar machine tool usage data sections and preparation of data for common and individual multitask learning models.

[0099] From this data, various features are generated, and these features are arranged using correlation analysis and statistical techniques. The distribution of these features in the feature space is then verified for each data type. This allows for the identification of the distribution and similarity of the data used to create the common multi-task learning model and individual multi-task learning models, enabling selection of training data.

[0100] As illustrated in Figure 7, the distribution of data across various machine tool usage contexts can be verified. This allows for the feature space distribution of previously learned data and new data to be verified and meaningful edge cases to be identified. This allows for comparison and analysis of the data distribution used for training the common multi-task learning model and individual multi-task learning models to determine whether additional training is necessary. Furthermore, the appropriate data distribution can be verified and training data selected for individual multi-task learning models with additional algorithm development. Furthermore, fine-tuning data tailored to the machine tool usage context can be utilized to develop a learning model more suitable for the equipment.

[0101] The verification module (26) verifies whether each of the above machine tools is operating according to the usage scenario for each of the above machine tools.

[0102] The collation module (27) collates and stores the real-time usage context of the machine tool operating according to the usage scenario for each machine tool when each machine tool is in operation as a result of the confirmation of the above confirmation module.

[0103] The selection module (28) selects whether the real-time usage context stored in the above collation module is useful.

[0104] The analysis module (29) extracts useful real-time usage contexts from among real-time usage contexts according to the selection result of the above selection module, classifies the extracted real-time usage contexts into real-time common usage contexts and real-time individual usage contexts, and analyzes the features of the corresponding usage contexts.

[0105] The judgment module (30) determines whether to proceed with a common multi-task or an individual multi-task based on the dependency relationship and features between the real-time common usage context and the real-time individual usage context based on the analysis results of the above analysis module.

[0106] This judgment module (900) can be implemented using at least one of ASICs (application specific integrated circuits), DSPs (digital signal processors), DSPDs (digital signal processing devices), PLDs (programmable logic devices), FPGAs (field programmable gate arrays), controllers, micro-controllers, microprocessors, and other electrical units for performing functions.

[0107] For example, individual multitasks should be designed for specific tasks, such as event detection or semantic segmentation in machine tool processes. Tasks applicable to machine tools include fault detection, equipment control parameter optimization, and tool life diagnosis. Fault detection includes tasks requiring real-time control, such as collision detection or tool breakage, as well as chatter detection with missing control elements, tool imbalance detection, and excessive vibration detection. Tasks for optimizing control parameters based on user-defined machining priorities (machining accuracy, machining speed, and tool life priority), and life diagnosis tasks include bearing life (spindle spindle, feed axes), tool life, and more.

[0108] For example, a common multitask context is a context in which a network can learn common features across multiple tasks and improve the network's overall performance and efficiency. For a common use case, we define a neural network that processes features continuously generated from analog sensor data. This common neural network requires a neural network design that satisfies the task's defined requirements by exploiting the trade-off between computational efficiency and throughput. It produces statistical features in the time-frequency domain for computing features from analog sensors and mechanical components, as well as features with physical meaning derived from spindles and motor rotors.

[0109] The inference module (31) automatically infers in real time the multi-tasks that must be performed in the individual control units of each machine tool currently in operation according to the corresponding usage scenario through machine learning using the common multi-task learning model and individual multi-task learning model generated by the generation module based on the judgment result of the judgment module.

[0110] The scheduling module (32) schedules multi-tasks to be performed in the individual control units of each machine tool currently in operation according to the usage scenario based on the inference result of the above inference module.

[0111] These scheduling modules derive scheduling to be performed on each of the machine tools in a time series manner through machine learning using a common multi-task learning model generated by the generating module and individual multi-task learning models, reflecting priority, real-time responsiveness, and / or periodicity.

[0112] For example, tasks requiring responsiveness and time determinism should be run continuously at a fixed cycle as the highest priority task. Soft real-time tasks for monitoring tolerate jitter in their execution cycles and have relatively long execution cycles. Tasks requiring high schedule-based accuracy collect data based on trigger events or fixed schedules. During this time, other real-time tasks are not run.

[0113] As illustrated in Figure 9, it is necessary to identify commonly required features through data including the machine tool usage context. At this time, a preliminary check for the operation of the common multi-task learning model or individual multi-task learning models, such as the presence of sensor abnormalities, is performed, and common features for the operation of the overall multi-task learning model or individual multi-task learning models in the time series and frequency domains must be created. Once this analysis is completed, the inference and scheduling of the overall multi-task learning model or individual multi-task learning model are performed. In other words, the trade-off relationship between accuracy and responsiveness is determined, and the multitasks to be performed among the overall multi-task learning model or individual multi-task learning models are inferred in real time for each machine tool usage situation, and a scheduling is derived accordingly.

[0114] As illustrated in Figure 10, for example, in individual multitask scheduling, one of the priorities for individual multitasks that should be driven by machine tools rather than computing resources is real-time responsiveness of inference. This can be divided into low-latency deterministic inference tasks for control, soft real-time inference tasks for monitoring, and tasks that infer and predict the current state according to a set schedule through data collection.

[0115] Low-latency deterministic inference and scheduling for equipment control also require different real-time responsiveness for each individual multitask.

[0116] For example, chatter is a self-induced vibration phenomenon that significantly impacts processing quality due to processing conditions and equipment rigidity. It typically requires rapid response, such as RPM control, as it is typically measured within tens of milliseconds.

[0117] Furthermore, for example, collision detection is a task that significantly impacts the overall lifespan of the equipment and machining quality by disrupting spindle axis alignment due to collisions between equipment and tools. This task requires a typical low-latency, real-time response inference system that must be determined within a few milliseconds and communicated to the equipment's digital I / O.

[0118] For example, tool breakage during machining can result not only in the loss of a portion of the cutting edge, but can also result in the complete destruction of the tool due to a collision caused by user negligence. This impacts the equipment, causing permanent damage. This impact can affect its structural rigidity and alignment, ultimately impacting machining quality.

[0119] For example, in the case of excessive equipment vibration, low-frequency vibrations can occur due to various causes, including incorrect machining settings, processing conditions, and tool settings. This can affect the machine's performance. This can affect machining quality in certain paths and situations, and should be recorded in an event log to help modify machining conditions. This log provides sufficient responsiveness to synchronize event occurrences with positioning along the path.

[0120] For example, in the case of tool imbalance, the task of detecting alignment issues between the tool and toolholder spindle clamping can lead to unintended rotational imbalance. This can impact machining quality. This requires signal monitoring, and post-processing evaluations can be made based on these events.

[0121] For example, in the case of bearing life diagnosis, even for the same equipment, the lifespan can vary depending on the various machine tool usage scenarios for each user. This task involves two tasks: one based on a schedule-based, non-processed air cut that assesses the condition, and the other predicting lifespan from the data obtained through this process. This task prioritizes accuracy over responsiveness, so it's crucial to accurately collect and analyze meaningful data and conduct periodic training to improve the accuracy of the neural network.

[0122] The feedback module (33) receives and stores information generated when performing multi-tasks in the individual control units of each machine tool currently in operation according to the scheduling created by the scheduling module, and continuously modifies and changes the common multi-task learning model and individual multi-task learning model generated by the generation module through machine learning by reflecting the feedback information.

[0123] The output module (34) converts data or results of the communication module, the database module, the classification module, the featuring module, the generation module, the confirmation module, the collation module, the selection module, the analysis module, the inference module, and the scheduling module into graphics, text, or images using a user graphic guidance program and displays them or transmits them to the individual control unit.

[0124] In this way, the multi-task real-time inference scheduling system of the machine tool according to the present invention infers the multi-tasks that must be performed in the individual control unit of each machine tool through machine learning through real-time usage contexts collected during operation according to usage scenarios for each machine tool, and schedules the multi-tasks of the corresponding machine tool through machine learning to schedule the optimal multi-task performance for each individual machine tool with each limited embedded environment, thereby easily and quickly performing expansion and usability expansion, promoting miniaturization of the machine tool and reducing space utilization, and continuously and accurately and quickly making judgments through updates according to new situations of individual machine tools with changed conditions, ratios, or environments as well as existing machine tools, thereby reducing maintenance costs and time due to damage or destruction of machine tools or workpieces, increasing consumer satisfaction, and improving exports.

[0125] The multi-task real-time inference scheduling method of a machine tool according to the present invention will be described with reference to FIGS. 8 to 10. As illustrated in FIG. 8, the multi-task real-time inference scheduling method of a machine tool according to the present invention includes a connection step, a collection step, a classification step, a featuring step, a generation step, a confirmation step, a collation step, a selection step, an analysis step (a judgment step, an inference step, a scheduling step, a feedback step, and an output step).

[0126] In each stage, the specific operation process or operation principle and the configuration or content of the device are the same as the multi-task real-time inference scheduling system of the machine tool of the specification of the present invention, and thus, the following description focuses on the unique points of the multi-task real-time inference scheduling method of the machine tool with reference to FIGS. 8 to 10.

[0127] The individual control unit that controls the operation of each machine tool and the central control unit are connected via a network, either wired or remotely.

[0128] After the connection step, the basic data and the usage context of each machine tool received from each of the above individual control units are collected and stored.

[0129] After the collection step, the usage contexts of each machine tool received from each of the above individual control units are classified into common usage contexts and individual usage contexts according to the dependency relationship.

[0130] The usage context and real-time usage context may be one or more of processing material information, equipment information, tool information, NC information, processing information, process information, maintenance information, surrounding information, measurement information, geometric information, and technical information.

[0131] After the classification step, features are assigned based on the characteristics of the common usage context and individual usage context classified into the common usage context and individual usage context of each machine tool according to the classification results.

[0132] After the featuring step, based on the common usage context and individual usage context of each machine tool featured according to the featuring result, a common multi-task learning model and an individual multi-task learning model are created by classifying them into layers according to the dependency relationship between the common usage context and individual usage context of each machine tool featured according to the feature result.

[0133] After the generation step, it is checked whether each of the above machine tools operates according to the usage scenario for each of the above machine tools.

[0134] After the verification step, if each of the above machine tools is in operation as a result of the verification, the real-time usage context of the machine tool operating according to the usage scenario for each machine tool is collected and stored.

[0135] After the collation step, you can choose whether the collated saved real-time usage context is useful or not.

[0136] After the selection step, useful real-time usage contexts are extracted from the real-time usage contexts based on the selection results, the extracted real-time usage contexts are classified into real-time common usage contexts and real-time individual usage contexts, and the features of the corresponding usage contexts are analyzed.

[0137] After the analysis step, based on the analysis results, it is determined whether to proceed with common multi-task or individual multi-task depending on the dependencies and features between the real-time common usage context and the real-time individual usage context.

[0138] After the judgment step, the multi-tasks that must be performed in the individual control units of each machine tool currently in operation according to the corresponding usage scenario are automatically inferred in real time through machine learning using supervised learning, unsupervised learning, and reinforcement learning based on the common multi-task learning model and individual multi-task learning models generated based on the judgment results.

[0139] After the inference step, based on the inference result, the multi-tasks to be performed in the individual control units of each machine tool currently in operation are scheduled to be performed in time series on each machine tool through machine learning by the common multi-task learning model generated by the generation module and the individual multi-task learning models, reflecting the priority, real-time responsiveness, and / or periodicity.

[0140] After the scheduling step, the information generated when performing multi-tasks in the individual control units of each machine tool currently in operation according to the scheduling is received as feedback and stored, and the common multi-task learning model and individual multi-task learning model that are generated by reflecting the feedback information are continuously modified and changed through machine learning to upgrade them.

[0141] In this way, the multi-task real-time inference scheduling system and the real-time inference scheduling method of the machine tool according to the present invention have a central control unit connected to one or more individual control units through a network, and the usage context of each machine tool is transmitted through each individual control unit, and a multi-task learning model is created by considering dependencies according to various environments and conditions through a neural network, thereby improving the stability and reliability of the machine tool, and by automatically continuously creating and updating the multi-task learning model regardless of the skill level, experience, and knowledge of the operator, there is an effect of promoting the convenience of the operator and preventing safety accidents in advance.

[0142] In addition, the multi-task real-time inference scheduling system and the real-time inference scheduling method of the machine tool according to the present invention infer the multi-tasks that should be performed in the individual control unit of each machine tool through machine learning through real-time usage contexts collected during operation according to usage scenarios for each machine tool, and schedule the multi-tasks of the corresponding machine tool through machine learning to schedule the optimal multi-task performance for each individual machine tool with each limited embedded environment, thereby easily and quickly performing expansion and usability expansion, promoting miniaturization of the machine tool and reducing space utilization, and reducing maintenance costs and time due to damage or destruction of the machine tool or workpiece by continuously and accurately and quickly making judgments through updates according to new situations of individual machine tools with changed conditions, ratios, or environments as well as existing machine tools, thereby increasing consumer satisfaction and enhancing exports.

[0143] Furthermore, the multi-task real-time inference scheduling system and real-time inference scheduling method of a machine tool according to the present invention automatically reflect the optimal scheduling to each machine tool to perform processing of a workpiece, thereby improving processing precision and productivity, preventing unnecessary occurrence of defective products, thereby conserving resources and increasing convenience for users and workers.

[0144] In addition, the embodiments of the present invention described above may be implemented in the form of program commands that can be executed through various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include program commands, data files, data structures, etc., either singly or in combination. The program commands recorded on the computer-readable recording medium may be specially designed and configured for the present invention or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specifically configured to store and execute program commands, such as ROMs, RAMs, and flash memories. Examples of program commands include not only machine language codes generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc. Hardware devices may be changed into one or more software modules to perform processing according to the present invention, and vice versa.

[0145] The specific implementations described in the present invention are exemplary embodiments and do not limit the scope of the present invention in any way. For the sake of brevity, descriptions of conventional electronic components, control systems, software, and other functional aspects of the systems may be omitted. In addition, the lines connecting or connecting members between components illustrated in the drawings are merely representative of functional connections and / or physical or circuit connections, and may be replaced or represented as various additional functional connections, physical connections, or circuit connections in an actual device. In addition, unless specifically mentioned as “essential,” “important,” etc., a component may not be absolutely necessary for the application of the present invention.

[0146] Furthermore, although the detailed description of the present invention has been described with reference to preferred embodiments of the present invention, it will be understood by those skilled in the art or having ordinary knowledge in the art that various modifications and changes can be made to the present invention without departing from the spirit and technical scope of the present invention as set forth in the claims below. Accordingly, the technical scope of the present invention should not be limited to the contents described in the detailed description of the specification, but should be defined by the claims.

[0147] <Explanation of symbols>

[0148] 10: Individual control unit

[0149] 20: Central Control Unit

Claims

1. Individual control unit for controlling the operation of each machine tool; and A central control unit, each of which is connected to one or more of the above individual control units through a network, receives the usage context of each of the above machine tools through each of the above individual control units, and generates a multi-task learning model through a neural network; A multi-task real-time inference scheduling system for a machine tool, characterized in that the central control unit infers, through machine learning, multi-tasks to be performed by the individual control units of each machine tool through real-time usage contexts collected during operation according to usage scenarios for each of the machine tools, and schedules the multi-tasks of the machine tool through machine learning.

2. In paragraph 1, A real-time inference scheduling system for driving a multi-task learning model of a machine tool, characterized in that the usage context for generating a multi-task learning model through the above neural network and the real-time usage context collected during operation according to the usage scenario for each of the above-mentioned machine tools are composed of information related to the usage scenario occurring in the entire usage cycle of each of the above-mentioned machine tools.

3. In paragraph 1, A real-time inference scheduling system for driving a multi-task learning model of a machine tool, characterized in that the usage context for generating a multi-task learning model through the neural network and the real-time usage context collected during operation according to the usage scenario for each of the above-mentioned machines include at least one of processing material information, equipment information, tool information, NC information, processing information, process information, maintenance information, surrounding information, measurement information, geometric information, and technical information.

4. In paragraph 1 or 2, The above central control unit, A communication module that communicates with each of the above individual control units via a network, either wired or remotely; A database module that stores the usage context of each machine tool received from each of the individual control units through the communication module; A classification module that classifies the usage contexts of each machine tool received from each of the individual control units stored in the database module into common usage contexts and individual usage contexts according to the dependency relationship; A featuring module that provides features according to the characteristics of the common usage context and individual usage context of each machine tool classified in the above classification module; and A real-time inference scheduling system for driving a multi-task learning model of a machine tool, characterized in that it includes a generation module that generates a common multi-task learning model and an individual multi-task learning model through machine learning based on the common usage context and individual usage context of each of the machine tools featured through the above-mentioned featuring module.

5. In paragraph 4, A real-time inference scheduling system for a machine tool, characterized in that the above classification module performs classification into common usage contexts and individual usage contexts through clustering and leveling of usage contexts collected from each machine tool through machine learning according to similar machine tool usage contexts.

6. In paragraph 4, A real-time inference scheduling system for a multi-task learning model of a machine tool, characterized in that the above-mentioned feature module is performed step by step through a neural network configured by layers.

7. In paragraph 4, A real-time inference scheduling system for driving a multi-task learning model of a machine tool, characterized in that the above generation module generates a common multi-task learning model and an individual multi-task learning model by dividing them by layer according to the dependency between the common usage context and the individual usage context of each machine tool featured by layer.

8. In paragraph 7, The above central control unit, A verification module that verifies whether each of the above machine tools is operating according to the usage scenario for each of the above machine tools; A collation module that collates and stores real-time usage contexts of machine tools operated according to usage scenarios for each machine tool when each machine tool is in operation as a result of the verification of the above verification module; A selection module that selects whether the real-time usage context stored in the above collation module is useful; An analysis module that extracts useful real-time usage contexts from real-time usage contexts according to the selection result of the above selection module, classifies the extracted real-time usage contexts into real-time common usage contexts and real-time individual usage contexts, and analyzes the features of the corresponding usage contexts; A judgment module that determines whether to proceed with a common multi-task or an individual multi-task based on the dependency relationship and features between the real-time common usage context and the real-time individual usage context based on the analysis results of the above analysis module; A real-time inference scheduling system for a machine tool driven by a multi-task learning model, characterized in that it further includes an inference module that automatically infers in real time the multi-tasks that must be performed in the individual control unit of each machine tool currently in operation according to the corresponding usage scenario through machine learning by the common multi-task learning model and the individual multi-task learning model generated by the generation module according to the judgment result of the judgment module.

9. In paragraph 8, The above central control unit, A real-time inference scheduling system for a machine tool, characterized in that it further includes a scheduling module that schedules multi-tasks to be performed in the individual control units of each machine tool currently in operation according to a usage scenario based on the inference results of the above inference module.

10. In paragraph 9, A real-time inference scheduling system for a machine tool driven by a multi-task learning model, characterized in that the scheduling module derives a scheduling to be performed in each of the machine tools in a time series manner through machine learning by a common multi-task learning model generated by the generation module and an individual multi-task learning model, reflecting priority, real-time responsiveness, and / or periodicity.

11. In paragraph 9, The above central control unit, A real-time inference scheduling system for driving a multi-task learning model of a machine tool, characterized in that it further includes a feedback module that receives and stores information generated when performing multi-tasks in the individual control units of each currently operating machine tool according to the scheduling created by the scheduling module, and continuously modifies and changes the common multi-task learning model and the individual multi-task learning model generated by the generation module by reflecting the feedback information through machine learning.

12. In paragraph 11, The above central control unit, A multi-task learning model-driven real-time inference scheduling system for a machine tool, characterized in that it further includes an output module that converts data or results of the communication module, the database module, the classification module, the featuring module, the generation module, the confirmation module, the collation module, the selection module, the analysis module, the inference module, and the scheduling module into graphics, text, or images and displays them as such using a user graphic guidance program, or transmits them to the individual control unit.

13. In paragraph 1, The above individual control unit, A memory unit for storing information for controlling the operation of each machine tool; An operating unit that operates the machine tool according to a usage scenario through information stored in the memory unit; A collection unit that collects real-time usage context of the machine tool operated by the above operation unit; and A real-time inference scheduling system for a machine tool driven by a multi-task learning model, characterized in that it includes a transmission / reception unit that transmits feedback information based on real-time usage context or scheduling collected in the above-mentioned collection unit to the central control unit or receives a newly generated or modified multi-task inference result or multi-task scheduling from the central control unit.

14. In paragraph 13, The above individual control unit, A real-time inference scheduling system for a multi-task learning model of a machine tool, characterized in that it further includes a display unit for displaying data or results of the memory unit, the operation unit, the collection unit, and the transmission / reception unit.

15. A step of connecting the individual control unit that controls the operation of each machine tool and the central control unit through a network, either wired or remotely; A step of collecting and storing basic data and the usage context of each machine tool received from each of the above individual control units; A step of classifying the usage contexts of each machine tool received from each of the above individual control units into common usage contexts and individual usage contexts according to the dependency relationship; A step of assigning features according to the characteristics of the common usage context and individual usage context classified into the common usage context and individual usage context of each machine tool according to the classification result; A method for real-time inference scheduling of a machine tool multi-task learning model, characterized in that it comprises a step of generating a common multi-task learning model and an individual multi-task learning model by classifying them by layer based on the dependency relationship between the common usage context and the individual usage context of each machine tool featured by layer through machine learning based on the common usage context and the individual usage context of each machine tool featured by the featuring result; 16. In paragraph 15, After the above creation step, A step of checking whether each of the above machine tools is operating according to the usage scenario for each of the above machine tools; A step of collecting and storing real-time usage contexts of machine tools operated according to usage scenarios for each machine tool when each of the above machine tools is in operation as a result of verification; A step for selecting whether the collected and saved real-time usage context is useful; A step of extracting useful real-time usage contexts from real-time usage contexts based on the selection results, classifying the extracted real-time usage contexts into real-time common usage contexts and real-time individual usage contexts, and analyzing the features of the corresponding usage contexts; A step of determining whether to proceed with common multi-task or individual multi-task based on the dependency relationship and features between the real-time common usage context and the real-time individual usage context based on the analysis results; and A method for real-time inference scheduling using a multi-task learning model of a machine tool, characterized in that it further includes a step of automatically inferring in real time the multi-tasks that should be performed in the individual control unit of each machine tool currently in operation according to the corresponding usage scenario through machine learning using a common multi-task learning model and an individual multi-task learning model generated according to the judgment result.

17. In paragraph 16, After the above inference step, A method for real-time inference scheduling of a machine tool using a multi-task learning model, characterized in that it further includes a step of scheduling multi-tasks to be performed in each of the machine tools in a time series manner by reflecting the priority, real-time responsiveness, and / or periodicity of the multi-tasks to be performed in each of the individual control units of the machine tools currently in operation according to the usage scenario based on the inference result, and machine learning by the common multi-task learning model generated by the generation module and the individual multi-task learning models.

18. In paragraph 17, After the above scheduling step, A real-time inference scheduling method for driving a multi-task learning model of a machine tool, characterized by including a step of receiving and storing information generated when performing multi-tasks in individual control units of each machine tool currently in operation according to scheduling, and continuously modifying and changing the generated common multi-task learning model and individual multi-task learning model through machine learning by reflecting the feedback information.

19. In paragraph 18, A real-time inference scheduling method for a multi-task learning model of a machine tool, characterized in that the usage context and real-time usage context are at least one of processing material information, equipment information, tool information, NC information, processing information, process information, maintenance information, surrounding information, measurement information, geometric information, and technical information.

Citation Information

Patent Citations

  • Scheduling device and learning device

    JP6929484B1

  • Method and Apparatus for automatically scheduling jobs in Computer Numerical Control machines using machine learning approaches

    KR101984460B1

  • A folding table

    KR102215871B1

  • Method and control apparatus for optimized control of a machine tool

    US20180314227A1

  • KR20220160309A