System and method for optimizing metal cutting process path based on artificial intelligence and big data
Patent Information
- Authority / Receiving Office
- KR · KR
- Patent Type
- Patents
- Current Assignee / Owner
- INNOX CORP
- Filing Date
- 2025-12-22
- Publication Date
- 2026-08-03
Smart Images
Figure 112025144898656-PAT00023_ABST
Abstract
Description
Technology Field
[0001] The present invention relates to the field of metal processing technology, and more specifically, to a process path optimization system and method that automatically calculates the optimal cutting sequence and cutting direction capable of minimizing thermal deformation and shortening processing time by training an artificial intelligence (AI) model with metal material, thickness, and past cutting history data. Background Technology
[0003] In metal cutting processes using numerically controlled (NC) machine tools, such as laser cutters and plasma cutters, the most critical factor determining productivity is the 'cutting path.' The cutting path refers to the sequence and direction in which the machining head moves, and it directly affects the total processing time (cycle time) and product quality.
[0004] Conventional CAM (Computer Aided Manufacturing) software generally generated cutting paths by relying on rules set by the operator or simple shortest distance algorithms (e.g., Nearest Neighbor). However, this conventional technology has the following problems.
[0005] First, because the path is generated based solely on geometric distance without considering the thermal conductivity according to the metal's material properties and thickness, thermal distortion occurs due to excessive heat accumulation when cutting is concentrated in a single area. This reduces the dimensional accuracy of the product and causes an increase in the defect rate.
[0006] Second, when processing parts with complex shapes or thin sheet metal, if the cutting sequence is improper, a 'tip-up' phenomenon may occur where already cut parts lose support and tilt, potentially leading to an accident involving a collision with the machining head.
[0007] Third, since routes are often manually modified relying on the experience of skilled workers, productivity varies significantly depending on the worker's proficiency, and it takes a long time for new workers to set the optimal route.
[0008] Therefore, there is a need for intelligent path optimization technology that allows artificial intelligence to learn metal physical property information and past success / failure history data, enabling the minimization of process time while avoiding thermal deformation.
[0009] The information described above is intended to enhance understanding of the background of the present invention and may include information that does not constitute prior art. Prior art literature
[0011] Registered Patent Publication No. 10-2710513 (September 25, 2024) Published Patent Publication No. 10-2025-0151964 (October 22, 2025) The problem to be solved
[0012] The present invention was developed to solve the problems of the aforementioned prior art, and the objective of the present invention is to provide a system that analyzes metal material, thickness, shape information, and cutting history data using artificial intelligence, and automatically generates an optimal cutting sequence that minimizes thermal deformation based thereon.
[0013] In addition, another objective of the present invention is to provide a method for maximizing productivity by deriving a cutting path that minimizes the rapid traverse distance of a processing head while simultaneously dispersing heat concentration by applying a reinforcement learning algorithm.
[0014] In addition, another objective of the present invention is to provide a self-learning system that continuously improves an artificial intelligence model by receiving feedback on error and time data collected after actual processing. means of solving the problem
[0016] An artificial intelligence-based process path optimization system according to an embodiment of the present invention for achieving the above objective comprises: an input unit that receives processing attribute data including drawing data, material information, and thickness information of a workpiece; a database that stores historical data including the cutting sequence, time required, occurrence of thermal deformation, and cutting quality data of a previously performed cutting process; a characteristic analysis module that analyzes the thermal conductivity characteristics and processing difficulty of a workpiece based on the processing attribute data and the historical data; an AI path generation unit that calculates an optimal process path including the movement path of a processing head, the cutting sequence, and the entry direction using a reinforcement learning model based on the analysis results of the characteristic analysis module; and a control unit that generates a Numerical Control (NC) code according to the calculated optimal process path and controls a cutting device.
[0017] Here, the AI path generation unit is characterized by including a deep neural network trained by dividing the area of the workpiece into a plurality of grids or nodes, defining the predicted amount of heat generated when cutting is performed at each node as a state, and setting the uniformity of the heat distribution and the total travel distance as a reward function.
[0018] In addition, the system may further include a simulation module that performs a virtual cutting simulation based on the calculated optimal process path to visualize the thermal deformation risk zone and calculate the estimated process time.
[0019] A process path optimization method according to one embodiment of the present invention comprises: (a) obtaining drawing data and material and thickness information of a workpiece; (b) extracting past cutting history data matching the material and thickness information from a database; (c) an artificial intelligence model extracting shape features of the drawing data and calculating a dispersion cutting pattern to prevent heat accumulation and the shortest movement path of a processing head; (d) performing a thermal deformation prediction simulation on the calculated path to verify it; and (e) generating a processing code for the verified path.
[0020] The AI path generation unit includes an AI-based process path optimization system characterized by modeling the region of the metal to be processed into multiple nodes, receiving the current estimated temperature of each node as a state, selecting cutting at a specific node as an action, and including a learned deep neural network with the total process time reduction and heat distribution uniformity resulting from the action as a reward.
[0021] The above AI path generation unit provides an artificial intelligence-based process path optimization system characterized by applying a heat distribution logic that, when the thickness of the metal to be processed is greater than or equal to a preset threshold, excludes adjacent parts within a preset distance from the currently cut part from the next cutting sequence and assigns weights to prioritize the selection of parts located at a distance.
[0022] The present invention provides an artificial intelligence-based process path optimization system characterized by further including a preprocessing unit that identifies an inner loop and an outer loop from the drawing data and performs phase alignment to prioritize cutting the inner loop, and the AI path generation unit that searches for an optimal path within the group aligned by the preprocessing unit.
[0023] The present invention provides an artificial intelligence-based process path optimization system characterized by further including a feedback module that collects processing time data and quality inspection data measured after an actual cutting process is performed by the control unit, updates the database, and retrains the neural network model of the AI path generation unit based on this.
[0024] The objective of certain embodiments of the present invention is to solve, mitigate, or eliminate at least one of the problems and / or disadvantages associated with the prior art, at least partially. Certain embodiments are intended to provide at least one of the advantages described below. Effects of the invention
[0026] According to the present invention, the following effects are achieved.
[0027] First, by automating path setting—which previously relied on the experience of skilled workers—using data-driven artificial intelligence, uniform and high-quality cutting results can be obtained regardless of the worker's skill level.
[0028] Second, by optimizing the cutting sequence (e.g., heat-dispersing cutting) by considering the thermal conductivity characteristics according to the material and thickness, thermal deformation of the material can be significantly reduced and the dimensional accuracy of the product can be improved.
[0029] Third, through reinforcement learning, the non-cutting movement (Rapid Traverse) distance of the machining head can be minimized to reduce the overall process time and increase productivity.
[0030] Fourth, as the system evolves on its own by continuously learning cutting history data, optimization performance is gradually improved even for special materials or difficult shapes.
[0031] The effects of the present disclosure are not limited to those mentioned above, and other unmentioned effects will be clearly understood by a person skilled in the art to which the present disclosure pertains (referred to as "person skilled in the art") from the description in the claims. Brief explanation of the drawing
[0033] The accompanying drawings, included as part of the detailed description to aid in understanding the embodiments, provide various embodiments and explain the technical features of the various embodiments together with the detailed description. FIG. 1 illustrates a process path optimization system according to various embodiments of the present disclosure. FIG. 2 illustrates the configuration of a process path optimization device in a process path optimization system according to various embodiments of the present disclosure. FIG. 3 illustrates a schematic diagram of the process of a process path optimization device generating numerical control code information in a process path optimization system according to various embodiments of the present disclosure. FIG. 4 illustrates a flowchart of the operation method of a process path optimization device in a process path optimization system according to various embodiments of the present disclosure. FIG. 5 illustrates an example that explains the process path optimization process in its entirety in a process path optimization system according to various embodiments of the present disclosure. Specific details for implementing the invention
[0034] The terms used in this disclosure are used merely to describe specific embodiments and are not intended to limit the scope of other embodiments. A singular expression may include a plural expression unless the context clearly indicates otherwise. Terms used herein, including technical or scientific terms, may have the same meaning as generally understood by those skilled in the art described in this disclosure. Terms used in this disclosure that are defined in a general dictionary may be interpreted as having the same or similar meaning as they have in the context of the relevant technology, and are not to be interpreted in an ideal or overly formal sense unless explicitly defined in this disclosure. In some cases, even terms defined in this disclosure are not to be interpreted to exclude the embodiments of this disclosure.
[0035] In the various embodiments of the present disclosure described below, a hardware-based approach is described as an example. However, since the various embodiments of the present disclosure include techniques using both hardware and software, the various embodiments of the present disclosure do not exclude a software-based approach.
[0036] The present disclosure relates to a system for optimizing a process path for metal processing. Specifically, the present disclosure describes an artificial intelligence-based process path optimization technology comprising: an input unit receiving input data including drawing data, material information, and thickness information of a metal to be processed; a database storing cutting history data including cutting sequences, movement paths, time required, and thermal deformation error data by material and thickness of previously performed cutting processes; an AI path generation unit that performs reinforcement learning based on the input data and the cutting history data to calculate an optimal cutting sequence and entry path that minimizes the rapid traverse time of a processing head and disperses local heat accumulation of the metal to be processed; and a control unit that converts the calculated optimal cutting sequence and entry path into a numerical control (NC) code and transmits it to a cutting device.
[0037] Hereinafter, various embodiments are described in detail with reference to the attached drawings so that those skilled in the art can easily implement the present disclosure. However, since the technical concept of the present disclosure can be modified and implemented in various forms, it is not limited to the embodiments described in this specification. In describing the embodiments disclosed in this specification, if it is determined that specifically describing related prior art could obscure the essence of the technical concept of the present disclosure, such specific description of prior art is omitted. Identical or similar components are assigned the same reference number, and redundant descriptions thereof are omitted.
[0038] When an element is described in this specification as being "connected" to another element, this includes not only cases where they are "directly connected" but also cases where they are "indirectly connected" with another element in between. When an element is described as "comprising" another element, this means that, unless specifically stated otherwise, it does not exclude other elements in addition to the other elements but may include additional elements.
[0039] Some embodiments may be described by functional block configurations and various processing steps. Some or all of these functional blocks may be implemented by various numbers of hardware and / or software configurations that perform specific functions. For example, the functional blocks of the present disclosure may be implemented by one or more microprocessors or by circuit configurations for a specific function. The functional blocks of the present disclosure may be implemented in various programming or scripting languages. The functional blocks of the present disclosure may be implemented as algorithms executed on one or more processors. The functions performed by the functional blocks of the present disclosure may be performed by a plurality of functional blocks, or the functions performed by a plurality of functional blocks in the present disclosure may be performed by a single functional block. Additionally, the present disclosure may employ prior art for electronic configuration, signal processing, and / or data processing, etc.
[0040] Additionally, in this disclosure, expressions such as "greater than" or "less than" have been used to determine whether specific conditions are satisfied or fulfilled; however, this is merely for illustrative purposes and does not exclude descriptions of "greater than" or "less than." Conditions described as "greater than" may be replaced with "greater than," conditions described as "less than" with "less than," and conditions described as "greater than and less than" with "greater than and less than."
[0041] Additionally, in the present disclosure, an image refers to an image conceptualized to visually capture information by a light beam or medium arranged in a plane or space. According to one embodiment of the present disclosure, an image may include still images and video images.
[0043] FIG. 1 illustrates a process path optimization system according to various embodiments of the present disclosure.
[0044] A process path optimization system (100) derives an optimized path within a process from input data including drawing data, material information, and thickness information of a metal to be processed based on deep learning, and directs a system to proceed with a process so that a user can select and execute the optimized process path. Specifically, the process path optimization system (100) derives an optimized path within a process from input data including drawing data, material information, and thickness information of a metal to be processed, and directs a system to proceed with a process so that a user can select and execute the optimized process path. According to one embodiment of the present disclosure, the process path optimization system (100) may include a user terminal (110), an external server (130), a process path optimization device (150), and a network (170).
[0045] The user terminal (110) refers to an electronic device operated by a user that captures or checks images. The user terminal (110) may be equipped with a camera and may capture images of a specific area based on user input and transmit them externally. According to one embodiment of the present disclosure, the user terminal (110) may include a fixed terminal implemented as a computer device or a mobile terminal. Specifically, the user terminal (110) may refer to a smartphone, a mobile phone, a navigation system, a computer, a laptop, a digital broadcasting terminal, a PDA (personal digital assistant), a PMP (portable multimedia player), or a tablet PC.
[0046] The external server (130) indicates a server device that stores data necessary during the process of analyzing deep learning-based modal data. The external server (130) may store data related to a video or the environment in which the video was captured in advance, and may transmit the stored data upon an external request. According to one embodiment of the present disclosure, the external server (130) may be implemented as a computer device or a plurality of computer devices that communicate with a user terminal (110) via a network to provide commands, codes, files, content, services, etc.
[0047] The process path optimization device (150) indicates a self-operating electronic device that provides a process path optimization system. The process path optimization device (150) can perform the function of analyzing an optimized process path from drawing data of a metal to be processed and existing cutting history data based on deep learning and carrying it out. According to one embodiment of the present disclosure, the process path optimization device (150) may also include a fixed terminal or a mobile terminal implemented as a computer device.
[0048] As illustrated in FIG. 1, the components of the process path optimization system (100) can be connected via a network (170). According to one embodiment of the present disclosure, the network (170) refers to a connection structure capable of exchanging information between each node, such as a plurality of terminals and servers. Examples of such networks include, but are not limited to, RF, 3GPP (3rd generation partnership project) networks, LTE (long term evolution) networks, 5GPP (5th generation partnership project) networks, WIMAX (world interoperability for microwave access) networks, the internet, LAN (local area network), Wireless LAN (wireless local area network), WAN (wide area network), PAN (personal area network), Bluetooth networks, NFC networks, satellite broadcasting networks, analog broadcasting networks, DMB (digital multimedia broadcasting) networks, etc.
[0049] According to the process path optimization system (100), the process path optimization device (150) and the artificial intelligence-based process path optimization system (200) instruct a system that calculates an optimal cutting sequence and entry path to minimize the non-cutting movement (Rapid Traverse) time of the processing head in the metal processing process and to disperse local heat accumulation in the metal to be processed, converts this into a numerical control (NC) code, and transmits it to a cutting device. Below, the specific configuration and operation of the process path optimization device (150) are described in detail.
[0051] FIG. 2 illustrates the configuration of a process path optimization device (150) in a process path optimization system (200) according to various embodiments of the present disclosure.
[0052] Terms such as '...part', '...unit' used below refer to a unit that processes at least one function or operation, and this may be implemented as hardware or software, or a combination of hardware and software. The image event analysis device (150) may include an input unit (210), a database (220), a path generation unit (230), and a control unit (240).
[0053] The database (220) temporarily or permanently stores data such as basic programs, application programs, and configuration information for the operation of the process path optimization device (150). The memory (210) may include a non-perishable permanent mass storage device such as RAM (random access memory), ROM (read only memory), and a disk drive, but the present invention is not limited thereto. These software components may be loaded from a computer-readable recording medium separate from the database (220) using a drive mechanism. This separate computer-readable recording medium may include computer-readable recording media such as a floppy drive, disk, tape, DVD / CD-ROM drive, or memory card. According to an embodiment, the software components may be loaded into the database (220) through a communication unit (not shown) rather than a computer-readable recording medium. Additionally, the database (220) may provide stored data upon request from the control unit (240).
[0054] The control unit (240) controls the overall operations of the process path optimization device (150). For example, the control unit (240) can control the transmission and reception of signals through a communication unit (not shown). Additionally, the control unit (240) can be configured to process commands of a computer program by performing basic arithmetic, logic, and input / output operations. Commands may be provided to the control unit (240) by a database (220) or a communication unit (not shown). For example, the control unit (240) can be configured to execute commands received according to program code stored in a recording device such as a database (220). For example, the control unit (240) can control the process path optimization device (150) to perform operations according to various embodiments described below.
[0055] According to one embodiment of the present disclosure, the control unit (240) can calculate an optimal cutting sequence and entry path to minimize the non-cutting movement (Rapid Traverse) time of the processing head in a metal processing process and to disperse local heat accumulation in the metal to be processed, generate numerical control (NC) code information, and control the transmission of numerical control code information to a user terminal.
[0056] Additionally, the control unit (240) can implement an artificial neural network model structure. That is, the artificial neural network model can be implemented in hardware or software through the control unit (240). The artificial neural network can be trained using big data regarding images. Such training may be performed, for example, in the process path optimization device (150) itself to which the artificial intelligence model is applied, or through a separate training server. The learning algorithm may include, for example, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but is not limited to the examples described above.
[0057] An artificial neural network model may include multiple artificial neural network layers. The artificial neural network may be one of a deep neural network (DNN), a convolutional neural network (CNN), a recurrent neural network (RNN), a restricted Boltzmann machine (RBM), a deep belief network (DBN), a bidirectional recurrent deep neural network (BRDNN), deep Q-networks, or a combination of two or more of these, but is not limited to the examples described above. In addition to the hardware structure, the artificial intelligence model may include a software structure, either additionally or substantially.
[0058] Throughout this specification, a neural network may be composed of a set of interconnected computational units referred to as nodes. These nodes may also be referred to as neurons. An artificial neural network is composed of a plurality of nodes, and the nodes may be interconnected by one or more links. Within the neural network, two or more nodes connected via links may form a relative relationship between an input node and an output node. The concepts of input and output nodes are relative; any node in an output node relationship with respect to one node may be in an input node relationship with respect to another node, and vice versa. As previously described, the input node versus output node relationship may be generated around links. One or more output nodes may be connected to a single input node via links. In an input node and output node relationship connected via a single link, the value of the output node may be determined based on data input to the input node. Here, the nodes interconnecting the input node and the output node may have weights. The weights may be variable and may be varied by a user or an algorithm to enable the neural network to perform the desired function. The artificial neural network model according to the present disclosure may indicate a model for detecting events within an image, and the event detection model may indicate a model trained to receive an image frame sequence and an audio signal as input, extract visual information related to the event from the image frame sequence, and extract auditory information related to the event from the audio signal and output it.
[0059] A communication unit (not shown) performs functions for transmitting and receiving signals through a wireless channel. All or part of the communication unit (not shown) may be referred to as a transmitting unit, a receiving unit, or a transmitting and receiving unit. The communication unit (not shown) may provide a function for the process path optimization device (150) and at least one other node to communicate with each other through a communication network. According to one embodiment of the present disclosure, when the control unit (240) of the process path optimization device (150) generates a request signal according to program code stored in a recording device such as a database (220), the request signal may be transmitted to at least one other node through a communication network under the control of the communication unit (not shown). Conversely, a control signal, command, content, file, etc. provided under the control of the processor of at least one other node may be received by the process path optimization device (150) through the communication unit (not shown).
[0060] An input / output interface (not shown) may be a means for interfacing with an input / output device (not shown). In this case, the input device may be provided in the form of a device such as a keyboard or a mouse, for example, and the output device may be provided in the form of a device such as a display unit for displaying images. As another example, the input / output interface (not shown) may be a means for interfacing with a device in which the functions for input and output are integrated into one, such as a touchscreen. Specifically, the control unit (240) of the process path optimization device (150) may display a service screen or content configured using data provided by a server in processing commands of a computer program loaded in the database (220) on a display through the input / output interface (not shown). According to one embodiment of the present disclosure, the input / output interface (not shown) may include a means for interfacing with a display unit (not shown). The input / output interface (not shown) may receive user input for a web browsing window displayed on the display unit (not shown) and may receive output data to be output through the display unit (not shown) in response to the aforementioned user input from the control unit (240).
[0061] A display unit (not shown) indicates a display module comprising one or more displays. Each of the one or more displays included in the display unit (not shown) may individually display independent content, and the one or more displays described above may be combined to display a single content. According to one embodiment of the present disclosure, the one or more displays included in the display unit (not shown) may include physically separated multiple displays, may be physically combined multiple displays, or may be displays capable of dividing a single screen for use.
[0062] The process path optimization device (150) performs reinforcement learning based on input data received from the input unit (210) and cutting history data stored in the database (220) to calculate an optimal cutting sequence and entry path that minimizes the non-cutting movement time of the processing head and disperses local heat accumulation of the metal to be processed.
[0063] That is, the input unit (210) indicates a component that receives input data including drawing data, material information, and thickness information of the metal to be processed.
[0064] For example, the input unit (210) receives material type and thickness information of the metal to be processed, along with CAD-based drawing data, and provides it so that it can be used to generate a path later.
[0065] The database (220) indicates a storage unit that stores cutting history data of past cutting processes.
[0066] The above cutting history data can be classified by material and thickness and may include cutting sequence, movement path of the processing head, time required, and thermal deformation error data. This cutting history data is used as training data when performing reinforcement learning by the path generation unit (230). At this time, the thermal deformation error data is an error value based on dimensional deviation, warping, bending, or quality inspection results after cutting, and can be used as training data / evaluation indicator for reinforcement learning described later.
[0067] The path generation unit (230) performs reinforcement learning based on input data received from the input unit (210) and cutting history data stored in the database (220) to calculate an optimal cutting sequence and entry path that minimizes the non-cutting movement time of the processing head and disperses local heat accumulation of the metal to be processed.
[0068] According to one embodiment of the present disclosure, the path generation unit (230) may model the metal to be processed into a plurality of regions and determine the cutting order through a neural network model learned using state information of each region and compensation information for performing cutting. In addition, the path may be generated to mitigate heat concentration by considering the thickness of the metal and the distance from the cut-completed part.
[0069] According to one embodiment of the present disclosure, the path generation unit (230) may model the area of the metal to be processed into a plurality of nodes and may receive the current estimated temperature of each node as a state. In addition, the path generation unit (230) may include a deep neural network that selects cutting at a specific node as an action and is trained with the amount of overall process time reduction and heat distribution uniformity according to the action as a reward.
[0070] Additionally, the path generation unit (230) may apply a heat dissipation logic that, when the thickness of the metal to be processed is greater than or equal to a preset threshold, excludes adjacent parts within a preset distance from the currently cut part from the next cutting sequence and assigns weights to prioritize the selection of parts located at a distance. At this time, the threshold and distance values may be set experimentally or empirically depending on the material, thickness, equipment conditions, etc.
[0071] According to one embodiment of the present invention, the reference value for the thickness of the metal to be processed, which is set in advance, may be 50 cm or more and 80 cm or less, and more specifically, 60 cm. In addition, the reference value for the distance from the currently cut part to the adjacent part, which is set in advance, may be 50 cm or more and 100 cm or less, and more specifically, 75 cm. As previously explained, when the thickness of the metal to be processed is 60 cm or more, which is the reference value set in advance, the path generation unit (230) may apply a heat distribution logic that assigns or adjusts weights to exclude adjacent parts within 75 cm of the currently cut part from the next cutting sequence and to prioritize the selection of parts located beyond 75 cm, which is the reference value set in advance. However, this is merely one embodiment of the invention and is not necessarily limited to these reference values.
[0072] Meanwhile, the path generation unit (230) can calculate an optimal path that considers the entire movement and column accumulation while adhering to the inner loop leading cut by searching for a path within a group that is phase-aligned in the preprocessing unit (not shown).
[0073] The control unit (240) instructs a component to convert the optimal cutting sequence and entry path calculated by the path generation unit (230) into a numerical control (NC) code and transmit it to the cutting device. Accordingly, the cutting device can perform the actual cutting process according to the NC code.
[0074] A preprocessing unit (not shown) identifies an inner loop and an outer loop from the drawing data and directs a component to perform topological sorting to cut the inner loop first. For example, the preprocessing unit (140) may form a group to cut the inner loop, such as a through hole or slot of a part, before the outer contour, and provide sorted information so that the path generation unit (150) searches for an optimal path within the group.
[0075] The feedback module (not shown) collects processing time data and quality inspection data measured after the actual cutting process is performed by the control unit (240), updates the database (220), and directs a component to retrain the neural network model of the path generation unit (230) based on this. Accordingly, the process path optimization system (100) can continuously improve performance in response to changes in equipment / material / thickness / environment.
[0077] FIG. 3 illustrates a schematic diagram of the process of a process path optimization device generating numerical control code information in a process path optimization system according to various embodiments of the present disclosure.
[0078] Referring to FIG. 3, the process path optimization system (100) receives input data (310) of a metal to be processed and can structure it into geometric shape information and physical property information. Subsequently, the process path optimization system (100) can search for an optimal path by performing reinforcement learning using cutting history data stored in a database (220), convert it into a numerical control (NC) code, and transmit it to a cutting device (not shown).
[0079] Specifically, the drawing data includes vector-based geometric information that defines the shape and arrangement of the metal to be processed. The input unit can extract drawing data from CAD file formats such as DXF (Drawing Exchange Format), DWG, and IGES.
[0080] Here, the drawing data includes object coordinate information such as lines, arcs, and polylines that form the outline of the part to be processed. In particular, the drawing data is utilized as foundational data to define the 'Environment' in which the path generation unit plans the cutting path. In this system, the drawing data may go beyond simple shape information to include information on closed loops constituting a closed shape, as well as spatial arrangement information such as the distance between parts within the raw material and the rotation angle of the parts. This serves as a reference coordinate system for the reinforcement learning model, which will be described later, to calculate the movement distance and entry angle between nodes.
[0081] Material information indicates the chemical composition of the metal to be processed or the type of material according to standard specifications. For example, material information may include broad classification information such as mild steel, stainless steel, and aluminum, and detailed grade information such as SS400, SUS304, and AL5052.
[0082] According to one embodiment of the present disclosure, material information can be linked to unique physical characteristic values of the material, going beyond a simple text identifier. That is, the input unit can load data on the thermal conductivity, specific heat, and thermal expansion coefficient of the metal by referencing an internal memory or database using the material information as a key value. This thermal characteristic data is utilized as a state variable in reinforcement learning, in which the path generation unit predicts the range of heat diffusion occurring during processing and calculates the possibility of deformation due to local heat accumulation.
[0083] Thickness information indicates the vertical dimension (T, Thickness) of the metal plate to be processed. The thickness information is entered as numerical data in millimeters (mm) or inches.
[0084] Thickness information serves as a key factor in determining the output parameters (e.g., beam output, cutting gas pressure) of a laser or plasma cutting device, while simultaneously acting as a threshold that modifies weights in process path optimization. Specifically, thickness information is used to estimate the total amount of heat input applied to the metal during cutting. For instance, if the input thickness information exceeds a preset threshold (e.g., for thick plates), the system determines that there is a high need for heat dissipation and can control the reward function of the path generation unit to increase the weight of the heat distribution uniformity item.
[0085] The database (220) is a storage medium that stores and manages big data, i.e., cutting history data, which is the result of numerous cutting processes performed in the past. The database (220) serves as an 'Experience Replay Memory' that the reinforcement learning model of the path generation unit refers to when performing training, and provides structured data indexed according to physical property variables such as material and thickness.
[0086] Specifically, the cutting order refers to chronological work list information indicating the sequence in which multiple parts placed within a single raw sheet were processed.
[0087] The cutting sequence data includes not only the sequence at the individual part level but also the execution sequence between inner closed curves (e.g., hole machining) and outer closed curves (outer cutting) within a single part. According to one embodiment of the present disclosure, the cutting sequence data may include pattern information indicating the degree of spatial dispersion of the cutting execution locations, rather than a simple sequence of numbers. For example, in the case of aluminum material with high thermal conductivity, the cutting sequence data may indicate whether a 'skip cutting' pattern is applied, which alternately processes the left and right sides of the raw material to avoid localized heating. This serves as basic data for an artificial intelligence model to learn a sequence strategy that minimizes heat accumulation in a specific material.
[0088] The movement path refers to the physical trajectory data of the processing head or nozzle moving from the start point to the end point.
[0089] The movement path is stored by distinguishing it into a cutting feed path, where the laser is irradiated to melt the actual metal, and a rapid traverse path, where cutting stops and the material moves quickly to the next processing position. Additionally, the movement path data includes shape (straight, arc, etc.) and length information for the lead-in and lead-out paths, which determine the edge quality of the metal. The path generation unit analyzes this data to evaluate and learn the efficiency of the shortest path algorithm that minimizes the 'air cut' time during the entire process.
[0090] Required time refers to the actual time data taken to complete the corresponding cutting process and is an indicator that quantifies productivity.
[0091] Time expenditure data can be stored not only as the total cycle time but also broken down into piercing time, cutting time, and head movement time. In particular, the database can store data by comparing theoretically calculated estimated times with actual times measured in the field. This is intended to reflect actual cutting speed reductions caused by equipment acceleration / deceleration performance or material characteristics, and helps the reinforcement learning model calculate rewards by reflecting physical constraints in the real world rather than a simulation environment.
[0092] Thermal deformation error data refers to data regarding the shape accuracy of a part measured after processing is completed, and is an indicator that quantifies quality.
[0093] Specifically, thermal deformation error data refers to the deviation between the ideal coordinate values of the input drawing data and the coordinate values of the workpiece measured by a vision sensor or a 3D measuring instrument after actual processing. According to one embodiment of the present disclosure, this data includes a value quantified in millimeters (mm) of the amount of dimensional error caused by warpage or shrinkage due to heat. By mapping and storing the correlation of thermal deformation errors according to the cutting order and path by material and thickness, the database predicts the expected defect rate when a path generation unit selects a specific path and induces the unit to learn in a direction that minimizes thermal deformation (assigning a penalty).
[0094] The path generation unit (230) is a core computational unit of the process path optimization system, and drives a reinforcement learning agent based on input data and past experience (cutting history data) to derive an optimal solution that simultaneously achieves two conflicting (trade-off) goals of productivity and quality.
[0095] Specifically, minimizing the rapid traverse time of the processing head means shortening the air cut, which is the period during which the laser or plasma beam simply moves its position while deactivated.
[0096] More specifically, the path generator calculates the Euclidean distance between the exit points of numerous parts placed within the workpiece metal and the entry points of the next part. While similar to the classic Traveling Salesman Problem (TSP), the path generator goes beyond simple distance calculation to consider the acceleration and deceleration profiles of the machining equipment. In other words, rather than a path that requires abrupt changes and causes severe deceleration even with a short travel distance, the path generator can define an inertia-friendly path that maintains high-speed movement, even if the travel distance is somewhat longer, as the 'optimal entry path.' This path can then be identified and selected as the 'minimum time path.' In short, the calculated optimal entry path can be the minimum time path. Through the 'minimum time path,' the non-cutting (rapid traverse) time of the machining head can be minimized, thereby shortening the overall process cycle time. More specifically, the minimum time path can be determined by the inertia-friendly path (w), the acceleration of the processing equipment (I_A), the deceleration of the processing equipment (I_DA), and the travel time (t).
[0097] Here, the minimum time path (D_min) is determined by the following mathematical formula 1.
[0099]
[0101] Referring to Equation 1, w represents an inertia-friendly path, I_A represents the acceleration of the processing equipment, I_DA represents the deceleration of the processing equipment, and t represents the travel time of the metal to be processed. The minimum time path (D_min) can be determined by considering an inertia-friendly path, which allows for high-speed movement even if the travel distance is somewhat longer, rather than a path where severe deceleration occurs due to the need for abrupt changes of direction even if the travel distance is short. Therefore, the minimum time path becomes longer as the acceleration of the processing equipment increases, and the minimum time path becomes shorter as the deceleration of the processing equipment increases. Furthermore, the minimum time path can become shorter as the inertia-friendly path, which involves relatively low abrupt changes of direction, increases.
[0102] Dispersing localized heat accumulation in the metal being processed refers to a logic that prevents thermal distortion—such as bending or twisting of the material—caused by high heat generated during the cutting process concentrating in specific areas of the metal sheet.
[0103] The path generation unit models the entire area of the metal to be processed as a virtual thermal grid and tracks the temperature of each grid as a state. When cutting is performed, the temperature of the corresponding coordinates and adjacent grids rises, and the system internally performs a simulation in which heat spreads to the surroundings or cools over time, depending on the material's thermal diffusivity.
[0104] When selecting the next part to cut, the reinforcement learning model imposes a heavy penalty on parts located in a 'Hot Zone' where the current temperature exceeds a preset threshold. Accordingly, instead of cutting adjacent parts consecutively, the path generator creates a 'Jumping Path' that moves to a distant part in a 'Cool Zone' to perform the cut, thereby ensuring that heat is not accumulated around specific parts but is evenly distributed across the entire sheet metal.
[0105] The process of calculating the optimal cutting sequence and entry path is the step of determining the final work order by combining the aforementioned time minimization and heat distribution.
[0106] Here, the cutting order refers to the macroscopic sequence of which part to cut first, and the lead-in path refers to the position and angle of entry of the piercing that penetrates from the outside to the inside when starting to cut individual parts.
[0107] The path generation unit not only determines the cutting order but also dynamically determines the optimal entry position that minimizes the travel distance to the next part. For example, since entry is possible from any point in 360 degrees for a circular part, the path generation unit shortens the travel path by setting the point closest to the end of the previous part's cutting as the entry point (Start Point). At the same time, if the entry point is a thermally vulnerable location, it generates the final NC code by calculating the order and entry path in an interdependent manner, such as by rerouting the entry point to a different angle.
[0108] This decision-making process is carried out through a deep neural network, and the path generation unit is trained to maximize a reward function such as <Equation 2>.
[0110]
[0112] Referring to mathematical formula 2, is the point in time It means compensation in, refers to the non-cutting travel time generated by the corresponding action. Also, ε is an indicator representing the uniformity of the heat distribution as the standard deviation of the temperature distribution over the entire processing area, and ε is a small constant to prevent the denominator from becoming zero. In other words, It is an indicator of the uniformity of the heat distribution, obtained by normalizing the standard deviation of the temperature distribution over the entire processing area to between 0 and 1. It refers to a penalty imposed for whether process constraints have been violated, such as violations of safety separation distances or exceeding heat concentration conditions, or for the extent of such violations. Meanwhile, , , These may represent weights for adjusting the importance of reducing non-cutting travel time, uniformizing heat distribution, and complying with process constraints, respectively. Based on such a compensation function, the path generation unit calculates the optimal cutting sequence and entry path to enable thermally stable cutting while reducing process time.
[0113] Specifically, the path generation unit is the compensation function of Equation 2 ( Using ), the optimal entry position is dynamically determined by the following mechanism.
[0114] First, the path generation unit analyzes the shape of the part to be cut next and generates a plurality of 'Candidate Lead-in Points' that can be entered. For example, in the case of a circular part, multiple points such as 0 degrees, 90 degrees, 180 degrees, and 270 degrees on the circumference can be candidates.
[0115] At this time, the path generation unit virtually calculates the movement path from the current machining head position (the cutting end point of the previous part) to each candidate entry point, and Calculate (non-cutting travel time).
[0116] The first term of <Mathematical Equation 2> If you refer to The smaller the value of the corresponding term (i.e., the shorter the travel distance), the larger the total reward ( It can have a structure where ) increases. Therefore, the artificial intelligence model has a 'distance-based incentive' to select the candidate entry point physically closest to the current location in order to maximize the reward.
[0117] However, if a point is selected based solely on distance, thermal deformation issues may occur. Therefore, the path generation unit [calculates] the expected temperature distribution change when cutting starts from the selected candidate entry point ( It simultaneously predicts... If the nearest entry point is near an already heated 'Hot Zone' and thus increases heat concentration during cutting, then the second term of Equation 2 is... The value of decreases, or the constraint violation term This is activated and causes the total reward to drop drastically.
[0118] Consequently, the path generator simulates competition between two variables—'shortest distance (maximization of the first term)' and 'lowest heat accumulation (maximization of the second term)'—within Equation 2. Through this process, the path generator performs 'Dynamic Determination,' which recalculates and selects a thermally safe entry point in real-time while minimizing travel time, whenever the cutting end position of the previous part changes.
[0120] FIG. 4 illustrates a flowchart (400) of the operation method of a process path optimization device (150) in a process path optimization system (100) according to various embodiments of the present disclosure.
[0121] Referring to FIG. 4, first, in step (410), the input unit of the process path optimization system receives input data containing processing information of the metal to be processed. The input data may include drawing data containing the shape and arrangement of the part to be processed (e.g., DXF, DWG file, etc.), material information indicating the type of material (e.g., mild steel, stainless steel, aluminum, etc.), and thickness information of the material. The material information may include not only the type of metal (e.g., SUS304, SS400, AL6061, etc.), but also physical property data such as the thermal conductivity, coefficient of thermal expansion, and melting point of the material. The thickness information indicates the thickness of the plate to be processed, and if the thickness is greater than or equal to a preset threshold (e.g., 6 mm), the system may classify it as a 'thick plate' and switch to a mode that applies a separate heat dissipation weight. These information serves as basic variables for the thermal deformation prediction and path calculation described later.
[0122] Next, in step (420), a process path optimization system according to an embodiment of the present invention stores or loads cutting history data of a past cutting process through a database. The cutting history data is big data that includes the cutting sequence of past processes performed under various material and thickness conditions, the movement path of the processing head, the actual time required, and thermal deformation error data that occurred during the execution of the process. This data is used as a criterion for setting a reward system when an AI model refers to past cases similar to current working conditions or performs reinforcement learning. The thermal deformation error data includes the difference value between the dimensions of the actual workpiece obtained through a vision sensor or a 3D measuring instrument after processing and the design drawing, and this can function as feedback data to correct the loss function of the AI model. The database can be implemented as a relational database (RDBMS) or a time-series database (Time-series DB).
[0123] Next, in step (430), the AI-based path generation unit calculates the optimal cutting sequence and entry path based on the received input data and the stored cutting history data. Specifically, in this step, two conflicting goals are simultaneously optimized using a reinforcement learning model or the like. The first is to increase overall process efficiency by minimizing the 'Rapid Traverse' time, during which the processing head moves without cutting, and the second is to prevent 'local heat accumulation' by spatially dispersing the cutting sequence so that heat is not concentrated on specific parts of the metal to be processed. Through this multi-purpose optimization, the AI path generation unit generates a path that minimizes thermal deformation while maintaining high productivity.
[0124] Finally, in step (440), the control unit converts the calculated optimal cutting sequence and entry path data into commands that can be recognized by the actual processing equipment. That is, the control unit converts the path data into numerical control (NC) codes, such as G-codes, and transmits them to a cutting device, such as a laser cutter, to control the actual processing to be performed. The control unit converts the calculated logical path data into numerical control (NC) codes, which are physical driving commands, and transmits them to the cutting device.
[0125] This process includes a 'post-processing' step. The generated path is converted into G-code and M-code that match the controller characteristics of specific cutting equipment (e.g., 5-axis laser cutter, plasma cutter).
[0126] Furthermore, this step may additionally include a step of performing a 'Virtual Machining Simulation' prior to transmitting the code. Through the simulation, the possibility of collision with the clamp on the head's movement path or whether the expected accumulated heat exceeds the allowable limit is verified in advance, and the final data is transmitted to the cutting device via TCP / IP, EtherCAT, or serial communication only if there are no abnormalities.
[0128] FIG. 5 illustrates an example that explains the process path optimization process in its entirety in a process path optimization system according to various embodiments of the present disclosure.
[0129] Referring to FIG. 5, an artificial intelligence-based process path optimization system (500) according to one embodiment of the present disclosure is largely composed of an input unit, a database, a path generator, and a controller, and performs a cyclic optimization process through linkage with actual processing equipment.
[0130] The Input module is a module that receives basic data for process execution. Drawing data refers to CAD file information (DXF, DWG, etc.) containing the shape, dimensions, and hole locations of the part to be machined. Material information includes physical property information such as the type of metal to be machined (e.g., steel, aluminum), thickness, and thermal conductivity. This information is utilized as an important parameter for predicting thermal deformation in the path generation unit, which will be described later.
[0131] The database stores and manages cutting history, which is big data of processes performed in the past. Specifically, it may include the following data.
[0132] The cutting order may include sequence data regarding the order in which parts were cut in past operations. The move path may include actual trajectory data of the movement of the machining head. The processing time may include actual time data taken when machining along the corresponding path. The above data can serve as a knowledge base that the artificial intelligence model of the path generation unit uses for learning or as a reference when establishing a new process plan.
[0133] The path generator is the core computing unit of this system and calculates the optimal path based on a reinforcement learning algorithm.
[0134] As illustrated by the icons in the drawing, Rapid Traverse optimization involves the path generator designing the path to minimize the section (dotted arrow) where the machining head moves to the next position without performing processing. In Reinforcement Learning, the AI agent analyzes the current machining state and learns to maximize rewards, such as reducing process time and dispersing heat accumulation. The arrow at the bottom of the drawing, returning from the machining equipment to the path generator, visually represents the structure where actual machining results or simulation results are fed back as training data for the AI model.
[0135] The controller converts the optimal path calculated by the path generation unit into a language that the actual machining equipment can understand. The NC code generates the optimized cutting sequence and entry coordinates as numerical control code, such as G-code, and transmits them to the machining equipment (bottom icon).
[0136] The system (500) of Fig. 5 is not a unidirectional process that simply receives and processes data, but rather implements an intelligent system in which path optimization performance is continuously improved as the process is repeated through a virtuous cycle structure leading from 'Input -> Optimization (PATH GENERATOR) -> Processing (CONTROLLER / Machine) -> Learning / Feedback (Reinforcement Learning)'.
[0138] Methods according to the embodiments described in the claims or specification of the present disclosure may be implemented in the form of hardware, software, or a combination of hardware and software. A method according to an embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium, or may be implemented as a computer program stored on a computer-readable recording medium in combination with hardware.
[0139] When implemented in software, a computer-readable storage medium may be provided for storing one or more programs (software modules). One or more programs stored in the computer-readable storage medium are configured for execution by one or more processors within an electronic device. One or more programs include instructions that cause the electronic device to execute methods according to the embodiments described in the claims or specification of this disclosure.
[0140] Such programs (software modules, software) may be stored in random access memory, non-volatile memory including flash memory, read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic disc storage devices, compact disc-ROM (CD-ROM), digital versatile discs (DVDs), or other forms of optical storage devices, magnetic cassettes. Alternatively, they may be stored in memory composed of some or all of these. Additionally, each constituent memory may include multiple units.
[0141] Additionally, the program may be stored on an attachable storage device that can be accessed via a communication network such as the Internet, Intranet, LAN (local area network), WAN (wide area network), or SAN (storage area network), or a combination thereof. Such a storage device may be connected to a device performing an embodiment of the present disclosure through an external port. Additionally, a separate storage device on a communication network may be connected to a device performing an embodiment of the present disclosure.
[0142] In the specific embodiments of the present disclosure described above, the components included in the disclosure are expressed in a singular or plural form according to the specific embodiments presented. However, the singular or plural expression is selected to suit the situation presented for convenience of explanation, and the present disclosure is not limited to singular or plural components; even if a component is expressed in the plural form, it may be composed of a singular form, and even if a component is expressed in the singular form, it may be composed of a plural form.
[0143] Meanwhile, although specific embodiments have been described in the detailed description of the present disclosure, it is understood that various modifications are possible within the scope of the present disclosure. Therefore, the scope of the present disclosure should not be limited to the described embodiments, but should be defined by the claims set forth below as well as equivalents thereof. Explanation of the symbols
[0145] 100, 200, 500: Process Path Optimization System 110: User terminal 130: External server 150: Process path optimizer 170: Network 210: Input section 220: Database 230: Path generation section 240: Control unit 310: Input data 400: Flowchart of Process Path Optimization Method 410: Step of receiving input data 420: Step to save or load cutting history data 430: Step to calculate the optimal cutting sequence and entry path 440: Step of converting the optimal path into an NC code and transmitting it
Claims
Claim 1 A process path optimization system for metal processing comprises: an input unit receiving input data including drawing data, material information, and thickness information of a metal to be processed; a database storing cutting history data including cutting sequences, movement paths, required times, and thermal deformation error data by material and thickness of previously performed cutting processes; and a path generation unit that performs reinforcement learning based on the input data and the cutting history data to calculate an optimal cutting sequence and entry path that minimizes the rapid traverse time of a processing head and disperses local heat accumulation of the metal to be processed. A control unit that converts the calculated optimal cutting sequence and entry path into a numerical control (NC) code and transmits it to a cutting device; wherein the path generation unit includes a deep neural network that models the area of the metal to be processed into a plurality of nodes, receives the current estimated temperature of each node as a state, selects cutting at a specific node as an action, and is trained with the total process time reduction amount and heat distribution uniformity according to the action as a reward; wherein, if the thickness of the metal to be processed is greater than or equal to a preset threshold, an adjacent part within a preset distance from the currently cut completed part is excluded from the next cutting sequence, and a heat distribution logic is applied to assign weights to prioritize the selection of a part located further away than the preset distance; and an AI-based process path optimization system that calculates a minimum time path (D_min) according to Equation 1. [Equation 1] (w may represent the inertia-friendly path, I_A the acceleration of the processing equipment, I_DA the deceleration of the processing equipment, and t the travel time of the metal to be processed.) Claim 2 delete Claim 3 delete Claim 4 The AI-based process path optimization system according to claim 1 further comprises a preprocessing unit that identifies an inner loop and an outer loop from the drawing data and performs phase alignment to prioritize cutting the inner loop, wherein the path generation unit searches for an optimal path within the group aligned by the preprocessing unit. Claim 5 An artificial intelligence-based process path optimization system according to claim 1, further comprising a feedback module that collects processing time data and quality inspection data measured after an actual cutting process is performed by the control unit, updates the database, and retrains the neural network model of the path generation unit based on the data. Claim 6 delete